1 00:00:00,444 --> 00:00:02,459 OpenAI is screwing YouTube and 2 00:00:02,514 --> 00:00:05,159 It kind of seems like everybody is screwing YouTube 3 00:00:05,510 --> 00:00:09,851 There's only a few of these that exist in the world, we actually have an AI DJ 4 00:00:09,856 --> 00:00:14,606 Let's crank it up and let the beat drop like it's hot, hot, hot. 5 00:00:15,771 --> 00:00:21,026 every Experience that we have is gonna be synthesized for a new large language model 6 00:00:21,046 --> 00:00:23,096 Now I wish I could punch you through the screen again 7 00:00:23,096 --> 00:00:23,946 Just kidding, everybody. 8 00:00:23,946 --> 00:00:24,946 We are having fun. 9 00:00:27,600 --> 00:00:28,070 Welcome. 10 00:00:28,080 --> 00:00:28,500 Welcome. 11 00:00:28,500 --> 00:00:29,320 Welcome everybody. 12 00:00:29,320 --> 00:00:31,970 It is another big episode of AI for humans. 13 00:00:32,310 --> 00:00:32,810 Oh, 14 00:00:32,990 --> 00:00:36,420 There was no, there was no warning whatsoever about 15 00:00:36,560 --> 00:00:37,380 We'll start over. 16 00:00:37,390 --> 00:00:37,820 We'll start 17 00:00:38,080 --> 00:00:41,210 And I felt unwelcomed, 18 00:00:41,640 --> 00:00:42,490 We're starting over. 19 00:00:42,490 --> 00:00:42,730 Stop. 20 00:00:43,010 --> 00:00:43,844 We're starting over. 21 00:00:44,274 --> 00:00:45,124 Welcome everybody. 22 00:00:45,124 --> 00:00:48,044 It is AI for Humans, your weekly guide to the wonderful and 23 00:00:48,064 --> 00:00:49,544 wild world of generative AI. 24 00:00:49,544 --> 00:00:50,414 I am here. 25 00:00:50,414 --> 00:00:53,404 My name is Gavin Purcell and my friend, Kevin Pereira is on the 26 00:00:53,434 --> 00:00:54,714 other end of the microphone. 27 00:00:54,744 --> 00:00:55,434 Kevin, how are you? 28 00:00:55,984 --> 00:00:57,934 I'm on the other end of this microphone. 29 00:00:57,934 --> 00:00:59,284 It's like two tin cans in a 30 00:00:59,624 --> 00:01:00,454 We're connected. 31 00:01:00,464 --> 00:01:01,284 We're connected. 32 00:01:01,999 --> 00:01:02,429 That's it. 33 00:01:02,449 --> 00:01:03,289 We've docked. 34 00:01:03,579 --> 00:01:04,169 Hi friends. 35 00:01:04,339 --> 00:01:05,369 me, KP. 36 00:01:05,679 --> 00:01:07,809 I'm leaving that entire intro in Gavin. 37 00:01:07,809 --> 00:01:09,719 What a beautiful AI for humans. 38 00:01:09,719 --> 00:01:11,079 Episode five, two. 39 00:01:11,079 --> 00:01:11,979 We got today. 40 00:01:12,399 --> 00:01:15,329 Should we tell the people what this podcast is about? 41 00:01:15,432 --> 00:01:18,912 We like to demystify all the news, tools, and all the other 42 00:01:18,912 --> 00:01:20,262 aspects out there just for you. 43 00:01:20,572 --> 00:01:21,922 Kevin, what's on the show today? 44 00:01:21,987 --> 00:01:26,037 OpenAI is screwing YouTube and Google is screwing YouTube. 45 00:01:26,047 --> 00:01:29,437 It kind of seems like everybody is screwing YouTube, except 46 00:01:29,437 --> 00:01:30,367 for maybe the creators. 47 00:01:30,367 --> 00:01:34,487 We're gonna tell you how and why, and maybe, maybe I'll share some very 48 00:01:34,487 --> 00:01:37,607 artistic, tasteful renderings that I've made of said screwing, Gavin. 49 00:01:38,542 --> 00:01:39,452 no, really? 50 00:01:40,307 --> 00:01:40,727 No! 51 00:01:41,292 --> 00:01:41,852 Is it okay? 52 00:01:41,852 --> 00:01:42,432 Good, good. 53 00:01:42,847 --> 00:01:44,497 I was just making sure you're paying attention. 54 00:01:44,877 --> 00:01:48,877 Also, hey, stop me if you've heard this one, Gavin, and or audience 55 00:01:48,877 --> 00:01:52,107 who cannot actually stop me because this is a one way medium. 56 00:01:52,447 --> 00:01:56,047 Elon Musk is working on robo taxis. 57 00:01:56,302 --> 00:01:57,782 Oh, I've heard this one, so you can 58 00:01:57,867 --> 00:02:02,297 Yeah, we've all heard this one since like 2014, but hopefully, the just 59 00:02:02,327 --> 00:02:06,977 announced Spotify AI DJ will be able to generate a playlist that's loud enough to 60 00:02:06,977 --> 00:02:11,417 drown out the screams as our robo taxis autopilot us through farmer's markets. 61 00:02:11,732 --> 00:02:14,412 Oh no, sir, I don't like that at all. 62 00:02:14,547 --> 00:02:17,517 . Listen, we're going to have all the details, plus some dumb things that 63 00:02:17,537 --> 00:02:20,667 everybody listening can do with AI today. 64 00:02:20,807 --> 00:02:23,347 We're going to chat with an AI powered guest, and then we're probably 65 00:02:23,347 --> 00:02:26,547 going to have a way more compelling chat with an actual human being, 66 00:02:26,547 --> 00:02:31,467 Gavin, a boundary pushing visual artist, who is our real guest today. 67 00:02:31,467 --> 00:02:37,924 We're going to dive deep into the cutting edge of AI artistry with PURZ. 68 00:02:37,924 --> 00:02:37,994 PURZ. 69 00:02:38,064 --> 00:02:38,834 It's PURZ. 70 00:02:38,874 --> 00:02:39,334 P U R Z, 71 00:02:39,399 --> 00:02:41,049 It's not Perz Beats, I thought his name was 72 00:02:41,174 --> 00:02:45,014 that's the full name, but colloquially, the casual, we're on that level. 73 00:02:45,204 --> 00:02:47,234 When we give a head nod, we say, Sup, Purrs? 74 00:02:47,849 --> 00:02:48,549 what's up Perz? 75 00:02:48,549 --> 00:02:49,759 All right, that'll be very exciting. 76 00:02:49,759 --> 00:02:54,024 Before we get started as always we want to tell you at home Please 77 00:02:54,244 --> 00:02:56,634 share , and rate our podcast. 78 00:02:56,644 --> 00:03:00,044 It is only with those shares and rates that we grow. 79 00:03:00,354 --> 00:03:03,464 Our YouTube video last week was doing very well and still is doing very well. 80 00:03:03,474 --> 00:03:07,054 And we appreciate everybody that watches the video likes and subscribes on YouTube. 81 00:03:07,424 --> 00:03:10,844 Also, please leave a five star podcast on Apple podcasts. 82 00:03:10,864 --> 00:03:12,044 We will read them at the end of the show. 83 00:03:12,044 --> 00:03:14,464 Kevin, today we have three new ones to read, which is exciting. 84 00:03:14,814 --> 00:03:17,234 Um, yes, we, we had three new. 85 00:03:17,344 --> 00:03:17,964 last week's 86 00:03:18,119 --> 00:03:18,889 All right, groveling? 87 00:03:18,894 --> 00:03:19,734 through to people? 88 00:03:19,799 --> 00:03:20,479 It got through. 89 00:03:20,579 --> 00:03:24,519 So again, we really appreciate everybody in our audience who listens to this show. 90 00:03:24,549 --> 00:03:28,719 This is not just the two of us BS ing for an hour and then we upload it. 91 00:03:28,929 --> 00:03:30,249 There's a lot of editing and all sorts of other 92 00:03:30,324 --> 00:03:31,724 hate each other. 93 00:03:31,824 --> 00:03:34,024 Every sentence is a grind. 94 00:03:34,024 --> 00:03:36,864 It is trench warfare when we launch this pod. 95 00:03:37,004 --> 00:03:41,434 And through gritted teeth and multiple takes, we managed to get sentences out. 96 00:03:41,434 --> 00:03:45,294 So if you appreciate the end product, we appreciate you engaging. 97 00:03:45,464 --> 00:03:47,934 Isn't that right, you piece of shit? 98 00:03:48,389 --> 00:03:49,359 Wow! 99 00:03:49,359 --> 00:03:51,409 Now I wish I could punch you through the screen again. 100 00:03:51,409 --> 00:03:52,259 Just kidding, everybody. 101 00:03:52,259 --> 00:03:53,259 We are having fun. 102 00:03:53,289 --> 00:03:55,749 I want everybody to know that is not exactly true. 103 00:03:56,039 --> 00:03:57,919 Kevin is making up lies right now. 104 00:03:57,949 --> 00:04:01,379 Kevin, it is time to get to the news! 105 00:04:03,139 --> 00:04:03,919 I'm already sweaty. 106 00:04:15,323 --> 00:04:15,663 Okay. 107 00:04:15,663 --> 00:04:19,643 The news this week is as usual, Fast and Furious, the biggest story that 108 00:04:19,653 --> 00:04:24,253 I've seen come down news story in a while and AI broke over the weekend. 109 00:04:24,533 --> 00:04:27,803 The New York times published a piece that had five authors on it. 110 00:04:27,803 --> 00:04:30,023 So, you know, when there's five authors on New York times piece, they've 111 00:04:30,023 --> 00:04:34,863 done some research, which basically accused open AI, Google, and meta. 112 00:04:34,933 --> 00:04:39,163 all of training on, um, various versions of data, but mostly they 113 00:04:39,163 --> 00:04:40,673 talked about scraping YouTube. 114 00:04:40,983 --> 00:04:44,713 And the big question with Sora, videos especially, has been, how 115 00:04:44,713 --> 00:04:46,433 did they get video to train Sora? 116 00:04:46,433 --> 00:04:48,493 Because as everybody knows who listens to the show, and if you 117 00:04:48,493 --> 00:04:50,133 don't, it's a pretty simple thing. 118 00:04:50,418 --> 00:04:53,548 You need a lot of data to train an AI model, whether 119 00:04:53,548 --> 00:04:55,228 that's an LLM and text model. 120 00:04:55,228 --> 00:04:57,208 It means you need to get a lot of data of words. 121 00:04:57,448 --> 00:05:00,058 In this instance, you need a lot of video to train something like 122 00:05:00,058 --> 00:05:01,608 SORA, especially train it as well. 123 00:05:01,828 --> 00:05:04,098 Now, OpenAI has done a deal with Shutterstock. 124 00:05:04,098 --> 00:05:07,118 So there's a lot of stuff out there, but most people, I think, 125 00:05:07,118 --> 00:05:10,368 haven't assumed that there was some sort of version of this happening. 126 00:05:10,888 --> 00:05:13,668 I think there was a little bit of a snippet that came out a couple weeks ago 127 00:05:13,668 --> 00:05:17,658 when Joanne Stern published her video where she interviewed Mina Murati, who 128 00:05:17,658 --> 00:05:21,888 is the OpenAI Chief Technical Officer, who could not answer the question about 129 00:05:21,898 --> 00:05:23,908 how they had trained their SORA model. 130 00:05:24,548 --> 00:05:27,788 She had a little bit of, yeah, that was a little bit of a face. 131 00:05:28,528 --> 00:05:30,108 So she really didn't have a good answer for this. 132 00:05:30,108 --> 00:05:33,068 And I think what we've learned now is that they did train it on YouTube. 133 00:05:33,088 --> 00:05:33,728 So, Kevin. 134 00:05:34,763 --> 00:05:37,673 First of all, when you read this piece, which was very long, and I encourage 135 00:05:37,673 --> 00:05:39,143 everybody who's listening to go read it. 136 00:05:39,143 --> 00:05:41,413 Cause it's a lot of, uh, there's a lot of things in there 137 00:05:41,413 --> 00:05:42,713 that are really fascinating. 138 00:05:42,943 --> 00:05:45,583 What was your first takeaway from reading this piece? 139 00:05:48,578 --> 00:05:50,008 What a non surprise! 140 00:05:52,730 --> 00:05:54,210 I wasn't shocked in the slightest. 141 00:05:54,220 --> 00:05:58,560 I got to imagine you weren't as well, Gavin, because we know how much data 142 00:05:58,570 --> 00:06:00,180 is required to train these things. 143 00:06:00,390 --> 00:06:03,180 , a couple of details from the article, and then we can dive in a little bit deeper. 144 00:06:03,180 --> 00:06:07,360 Greg Brockman, OpenAI's president, , allegedly personally helped 145 00:06:07,370 --> 00:06:09,020 collect the videos that were used. 146 00:06:09,070 --> 00:06:10,310 , again, according to the New York Times. 147 00:06:10,720 --> 00:06:13,400 What they did though, was that they grabbed, , transcripts 148 00:06:13,470 --> 00:06:14,890 of all the YouTube videos. 149 00:06:14,950 --> 00:06:18,790 And using that, which would be a violation, according to Google, 150 00:06:19,040 --> 00:06:20,860 of YouTube's usage policies. 151 00:06:20,860 --> 00:06:23,859 You cannot use their data to train something else. 152 00:06:24,170 --> 00:06:28,930 However, buried within that article is an accusation that Google may have done the 153 00:06:29,030 --> 00:06:30,190 Done it themselves. 154 00:06:30,190 --> 00:06:31,120 Yeah, exactly. 155 00:06:31,380 --> 00:06:35,220 and that the reason Google might not be making any public statements 156 00:06:35,220 --> 00:06:39,310 about OpenAI's alleged actions is that they're curious if they 157 00:06:39,310 --> 00:06:40,940 would be outing themselves as well. 158 00:06:41,278 --> 00:06:41,808 Exactly. 159 00:06:41,808 --> 00:06:42,598 And it's been really interesting. 160 00:06:42,598 --> 00:06:48,268 So Neil Mohan, the, , YouTube CEO did come out and say that, , some YouTube 161 00:06:48,308 --> 00:06:51,218 content is scrapable for open web purposes, but the video transcript and 162 00:06:51,218 --> 00:06:52,608 footage are not allowed to be scraped. 163 00:06:52,618 --> 00:06:55,468 He said, this is a clear violation of our terms of service. 164 00:06:55,518 --> 00:06:58,548 So those are the rules of the road in terms of content on our platform. 165 00:06:58,848 --> 00:06:59,498 But you're right. 166 00:06:59,508 --> 00:07:00,638 It is interesting to see. 167 00:07:01,148 --> 00:07:06,278 how how kind of quiet the overall Google ecosystem has been on this. 168 00:07:06,278 --> 00:07:08,088 And I think one thing I want to point out 169 00:07:08,168 --> 00:07:11,278 I was really kind of surprised at how little noise this 170 00:07:11,278 --> 00:07:13,208 made in the AI ecosystem. 171 00:07:13,228 --> 00:07:17,768 And when I say the AI ecosystem, I mean, the people kind of like us or more of 172 00:07:17,768 --> 00:07:20,618 the people who are kind of interested in AI who are kind of focused on this. 173 00:07:20,878 --> 00:07:23,508 When I first read this, I was like, well, this is a smoking gun. 174 00:07:23,508 --> 00:07:24,158 Like, look at this. 175 00:07:24,158 --> 00:07:28,168 This is going to like launch what I believe will probably be dozens of 176 00:07:28,168 --> 00:07:30,558 lawsuits, whether they get through or not. 177 00:07:30,868 --> 00:07:35,018 Now, I think the big question is going to be what do Google's terms of services 178 00:07:35,018 --> 00:07:38,538 say when things were scraped, if they were scraped, and it sounds like they were. 179 00:07:38,898 --> 00:07:40,878 But then I think the next question is going to be. 180 00:07:40,878 --> 00:07:43,548 And this, I want to ask you directly this, cause there's so many people 181 00:07:43,548 --> 00:07:46,888 who talk about, and you've said this and I've said this, the idea that 182 00:07:46,898 --> 00:07:49,198 like the cat is out of the bag, right? 183 00:07:49,198 --> 00:07:52,038 That the idea that once you've done this, guess what? 184 00:07:52,038 --> 00:07:52,948 It's too late. 185 00:07:53,278 --> 00:07:56,228 All this stuff is out there and now we just have to live with what it is. 186 00:07:56,748 --> 00:08:01,558 Do you think that there is any way that there is a pause that can be put on 187 00:08:01,678 --> 00:08:04,198 this based on legal terms at this point? 188 00:08:04,338 --> 00:08:08,398 Perhaps an injunction against the output of all large language models 189 00:08:08,398 --> 00:08:11,301 until a judge can say hey We got to piece this together, right? 190 00:08:11,301 --> 00:08:13,671 We got to dive in through your model, see how it was trained, 191 00:08:13,671 --> 00:08:14,821 see if you violated something. 192 00:08:15,031 --> 00:08:19,841 , I think there's too much money at play , that, that could be a scenario, but I 193 00:08:19,861 --> 00:08:24,251 think there will just be millions, if not billions of dollars thrown against the 194 00:08:24,251 --> 00:08:27,781 wall to make sure that something like that doesn't happen in the interim and it's 195 00:08:27,781 --> 00:08:34,561 business as usual until decades later, everything megacorp that's going to rule 196 00:08:34,561 --> 00:08:36,521 us all and be our one world government. 197 00:08:36,856 --> 00:08:39,076 Like, people will just play nice. 198 00:08:39,116 --> 00:08:43,196 There's just too much money at stake, so I think this is a calculated move. 199 00:08:43,226 --> 00:08:46,776 I'm sure there were lawyers involved at one point in the room, and 200 00:08:46,776 --> 00:08:50,866 eventually, probably the engineers said, Hey, listen, we have to sprint. 201 00:08:50,876 --> 00:08:52,156 We have to grab this data. 202 00:08:52,156 --> 00:08:55,306 We will beg for, and probably pay for, forgiveness later. 203 00:08:55,566 --> 00:08:56,736 So let's just go. 204 00:08:57,346 --> 00:08:58,306 I think that's probably true. 205 00:08:58,306 --> 00:09:01,366 And I think the one thing that's important to point out here is there's been a couple 206 00:09:01,366 --> 00:09:05,596 stories this week about how hard it is to get new data to train , these devices on. 207 00:09:05,596 --> 00:09:08,741 And there's lots of people in the machine learning world that are saying, Hey, 208 00:09:08,741 --> 00:09:10,101 this is going to be a scaling issue. 209 00:09:10,101 --> 00:09:13,571 Like actually every time you scale up, you get better results, but 210 00:09:13,581 --> 00:09:15,851 scaling equals more data, right? 211 00:09:15,851 --> 00:09:19,691 And so the one thing that they've talked about a lot is we have 212 00:09:19,691 --> 00:09:21,901 scraped so much text data already. 213 00:09:21,901 --> 00:09:26,041 Like so many books and again, who knows the legal ramifications of that, putting 214 00:09:26,041 --> 00:09:27,861 it aside, but we've scraped the internet. 215 00:09:27,861 --> 00:09:28,651 We've scraped books. 216 00:09:28,651 --> 00:09:29,581 We've scraped all this stuff. 217 00:09:30,061 --> 00:09:34,801 If they're scraping and have scraped, , a lot of YouTube already, where does 218 00:09:34,811 --> 00:09:37,061 more data come from is a huge question. 219 00:09:37,061 --> 00:09:39,651 And that's partly the whole synthetic data conversation. 220 00:09:39,651 --> 00:09:42,871 But like, I don't know where you get more than YouTube when it comes to video. 221 00:09:43,431 --> 00:09:48,331 Someone has to be transcribing and scraping every podcast in every 222 00:09:48,331 --> 00:09:51,061 language and going back , decades ago. 223 00:09:51,271 --> 00:09:54,301 , I'm sure that's already been thought of and is being grabbed as well. 224 00:09:54,411 --> 00:09:57,661 , last year, this is according to the New York times article, Google quote, 225 00:09:57,736 --> 00:09:59,506 Also brought in its terms of service. 226 00:09:59,516 --> 00:10:02,646 One motivation for the change, according to members of the company's 227 00:10:02,656 --> 00:10:07,366 privacy team and an internal message reviewed by the times was to allow 228 00:10:07,366 --> 00:10:11,306 Google to be able to tap publicly available, Google docs, restaurant 229 00:10:11,306 --> 00:10:15,916 reviews on Google maps and other online material for more of its AI products. 230 00:10:15,916 --> 00:10:19,466 So even Google was like, Hey, we got to start getting this, you know, at 231 00:10:19,466 --> 00:10:22,286 some point there's going to be some sort of like, Hey, we'll give you. 232 00:10:22,876 --> 00:10:26,536 Maybe, maybe we'll give you 3 percent off your Google for business account. 233 00:10:26,556 --> 00:10:30,566 If you let us crawl through some files, in fact, maybe we 234 00:10:30,566 --> 00:10:31,996 won't give you anything off. 235 00:10:32,166 --> 00:10:33,056 Maybe we'll just 236 00:10:33,311 --> 00:10:34,231 We'll just do it. 237 00:10:34,386 --> 00:10:36,576 of service and just go on through it. 238 00:10:36,626 --> 00:10:37,786 That's probably going to happen. 239 00:10:37,786 --> 00:10:41,456 But on the synthetic data front, that, that you brought up, that to me is 240 00:10:41,506 --> 00:10:42,871 really, really interesting because. 241 00:10:43,321 --> 00:10:47,511 It seems like every day a new paper comes out that says synthetic data 242 00:10:47,521 --> 00:10:50,431 is great and it's going to solve the problem, which is something that Sam 243 00:10:50,441 --> 00:10:52,781 Altman has said is going to be the case. 244 00:10:53,201 --> 00:10:56,231 Or there's a paper that comes out and says synthetic data is terrible and the 245 00:10:56,251 --> 00:10:57,741 AI is going to train itself in a loop. 246 00:10:57,741 --> 00:11:00,521 So for the broader audience out there that doesn't know what we are talking 247 00:11:00,531 --> 00:11:04,691 about when we say synthetic data, we're saying that You can use the large 248 00:11:04,691 --> 00:11:07,671 language models as they exist today and likely as they'll exist in the 249 00:11:07,671 --> 00:11:13,691 next year or two, which will be better, theoretically, to generate paragraphs upon 250 00:11:13,691 --> 00:11:17,181 paragraphs of text and a computer code. 251 00:11:17,181 --> 00:11:19,321 I mean, we're talking like generate the Library of Congress. 252 00:11:19,321 --> 00:11:20,631 It could probably do it in a week. 253 00:11:20,911 --> 00:11:26,331 It just is going to churn and give you fake questions and answers, fake recipes, 254 00:11:26,341 --> 00:11:28,711 fake restaurant reviews, fake everything. 255 00:11:28,711 --> 00:11:33,161 That synthetic data is thought by some to be good enough. 256 00:11:33,551 --> 00:11:36,981 That if you generate enough of it, you can train a better model. 257 00:11:36,981 --> 00:11:41,051 But some are saying that actually the repeated words and phrases that are 258 00:11:41,051 --> 00:11:44,881 already seen to come out of some of these language models, the bad code 259 00:11:45,011 --> 00:11:48,531 or the potentially pirated snippets of content, those are just going 260 00:11:48,531 --> 00:11:50,211 to resurface over and over again. 261 00:11:50,221 --> 00:11:53,761 And then eventually it kind of trains itself on its worst bad habits 262 00:11:53,761 --> 00:11:54,901 and it's not going to be usable. 263 00:11:55,121 --> 00:11:56,981 Those are the two schools of thought right now. 264 00:11:57,481 --> 00:11:59,591 Do you have a horsey in this race, Gavin? 265 00:12:00,031 --> 00:12:02,551 I don't understand enough on the technical side to see if the 266 00:12:02,551 --> 00:12:03,851 synthetic data is strong enough. 267 00:12:03,861 --> 00:12:07,281 I will say from what I've read, the arguments around synthetic data is it 268 00:12:07,291 --> 00:12:12,851 wasn't good enough before, but now as the models get stronger, it may be better. 269 00:12:12,851 --> 00:12:16,851 The interesting thing to think about is, if that is the unlock and we're also 270 00:12:16,851 --> 00:12:20,441 heading to a world where only bigger, bigger models equal better results, 271 00:12:20,861 --> 00:12:24,341 then we are going to be moving very quickly to the next stage of whatever 272 00:12:24,341 --> 00:12:28,461 this AI world is, because the AI, as we talked about in the NVIDIA episode, 273 00:12:28,861 --> 00:12:32,741 the AI can simulate themselves in different places and different things. 274 00:12:33,111 --> 00:12:35,771 And if that allows, if you can simulate. 275 00:12:36,296 --> 00:12:39,066 Interactions, text, video, all that stuff. 276 00:12:39,496 --> 00:12:43,316 You can ostensibly do that in an infinite level if you have the amount 277 00:12:43,316 --> 00:12:44,876 of compute and storage to do it. 278 00:12:44,896 --> 00:12:46,406 And then it goes really quickly. 279 00:12:46,456 --> 00:12:50,076 Actually, this really dovetails really interestingly into another story that we 280 00:12:50,076 --> 00:12:55,606 have Bill Peebles, one of the main open AI engineers behind Sora had a really 281 00:12:55,606 --> 00:12:59,953 interesting talk that he came out and gave to the, AGI house, which is again, 282 00:12:59,963 --> 00:13:02,663 like probably some sort of hype house, but for AGI is like, whoop, whoop. 283 00:13:02,933 --> 00:13:04,693 We're going to AGI land, baby. 284 00:13:05,763 --> 00:13:06,113 Like and 285 00:13:06,393 --> 00:13:07,483 Energy Refrigerator! 286 00:13:07,483 --> 00:13:08,473 It's so hype! 287 00:13:09,033 --> 00:13:09,373 Yeah. 288 00:13:09,633 --> 00:13:12,873 So Bill was at the AGI house and gave a speech, and we're going to hear a little 289 00:13:12,873 --> 00:13:14,543 snippet of what his speech was here. 290 00:13:15,113 --> 00:13:18,483 So, of course everyone's very bullish on the role that LLMs are 291 00:13:18,483 --> 00:13:20,288 going to play in getting to AGI. 292 00:13:20,798 --> 00:13:24,198 But we believe that video models are on the critical path to it. 293 00:13:24,198 --> 00:13:28,658 And concretely, we believe that when we look at very complex scenes that Sora 294 00:13:28,658 --> 00:13:33,038 can generate, like that snowy scene in Tokyo that we saw in the very beginning, 295 00:13:33,038 --> 00:13:36,348 that Sora is already beginning to show a detailed understanding of how humans 296 00:13:36,348 --> 00:13:39,938 interact with one another, how they have physical contact with one another. 297 00:13:40,358 --> 00:13:43,658 And as we continue to scale this paradigm, we think eventually it's going 298 00:13:43,658 --> 00:13:45,348 to have to model a human state, right? 299 00:13:45,398 --> 00:13:48,218 The only way you can generate truly realistic video, with truly 300 00:13:48,218 --> 00:13:49,578 realistic sequences of actions. 301 00:13:49,993 --> 00:13:52,493 is if you have an internal model of how all objects, 302 00:13:52,493 --> 00:13:54,233 humans, etc., environments work. 303 00:13:54,833 --> 00:13:58,003 And so we think this is how Sora is going to contribute to HDI. 304 00:13:58,003 --> 00:13:59,343 Basically what Mr. 305 00:13:59,343 --> 00:14:01,473 Peebles is saying here, call him, you know, Mr. 306 00:14:01,473 --> 00:14:06,578 Peebles if you're nasty, uh, He's talking about, he's talking about 307 00:14:06,578 --> 00:14:11,228 the idea that Sora, which is a video generator, is actually a world 308 00:14:11,228 --> 00:14:14,028 simulator, ostensibly, and we've talked about this on the show before, 309 00:14:14,468 --> 00:14:19,098 but by world simulation, and by doing that, even in synthetic environments, 310 00:14:19,098 --> 00:14:24,543 you are teaching the AI How people and objects interact with each other. 311 00:14:24,543 --> 00:14:27,383 And that is a lot of people see as the next stage of data, right? 312 00:14:27,383 --> 00:14:30,863 It's like real world training or simulated world training 313 00:14:30,883 --> 00:14:32,833 outside of just words and texts. 314 00:14:33,083 --> 00:14:34,513 You're talking now about actions. 315 00:14:34,513 --> 00:14:36,943 You're talking about the way a tree blows in the wind. 316 00:14:36,943 --> 00:14:40,113 You're talking about the way a human interacts with that tree when it blows 317 00:14:40,113 --> 00:14:41,523 against them, where they cut it down. 318 00:14:41,893 --> 00:14:44,153 All of that stuff is data, right? 319 00:14:44,153 --> 00:14:45,713 It can't be perceived as data. 320 00:14:46,023 --> 00:14:49,703 And if they can find a way that Sora can generate synthetic models 321 00:14:49,713 --> 00:14:52,803 of the world, that does feel like where we're headed next. 322 00:14:53,403 --> 00:14:58,338 I think 2026 is the year that And it's going to need a sexier term 323 00:14:58,338 --> 00:15:01,588 to probably get the youth involved, Gavin, but, uh, let's say, let's 324 00:15:01,588 --> 00:15:05,060 say data harvesters or, aggregators. 325 00:15:05,340 --> 00:15:12,015 Basically, we are going to, for pennies per minute, of experience, 326 00:15:12,285 --> 00:15:15,485 we are all going to end up working for open AI slash Microsoft. 327 00:15:15,785 --> 00:15:19,845 We're going to have our glasses on that are going to feed constant video streams. 328 00:15:20,045 --> 00:15:24,905 Maybe we'll have tactile haptic gloves so that it can know the pressure and force 329 00:15:24,915 --> 00:15:28,925 with which we're exerting upon the world as we navigate it, maybe even sensors in 330 00:15:28,925 --> 00:15:30,775 our free government provided sketchers. 331 00:15:31,115 --> 00:15:31,455 Right. 332 00:15:31,645 --> 00:15:33,875 But we are all just going to live. 333 00:15:34,210 --> 00:15:39,150 Every single day, providing data for these mega models, because it's just 334 00:15:39,150 --> 00:15:42,400 gonna, have an insatiable appetite, and even though the synthetic 335 00:15:42,400 --> 00:15:45,840 stuff will be good, it's never gonna hit as hard as an analog, baby. 336 00:15:46,060 --> 00:15:50,390 So, we're all just gonna sign up to work for one of the companies, and 337 00:15:50,390 --> 00:15:54,140 instead of running Uber Eats errands, we're just gonna eke out our day. 338 00:15:54,150 --> 00:15:55,105 We're gonna crawl, crawl, crawl. 339 00:15:55,315 --> 00:15:58,825 Through the actual world try to breathe the polluted air and drink 340 00:15:58,845 --> 00:16:03,025 water that doesn't have microplastics and every Experience that we have 341 00:16:03,025 --> 00:16:07,785 Gavin is gonna be synthesized for a new large language model sign me up 342 00:16:08,445 --> 00:16:10,025 I honestly think you're right. 343 00:16:10,055 --> 00:16:13,405 I will say it doesn't have to be as dystopian of that, but what I could see 344 00:16:13,405 --> 00:16:16,805 very well, and we're going to talk about robo taxis or Waymo taxis in a little 345 00:16:16,805 --> 00:16:20,845 bit, you know, right now, when you see a robo taxi or you see a Waymo car in 346 00:16:20,845 --> 00:16:22,225 the real world, that's kind of weird. 347 00:16:22,635 --> 00:16:26,795 I do see a future where there are people that are walking through a busy city, 348 00:16:26,815 --> 00:16:30,015 which are wearing glasses and some sort of haptic gloves, and you kind of 349 00:16:30,015 --> 00:16:31,795 get to know them as like, they're the. 350 00:16:32,185 --> 00:16:36,175 They're the experience generator or they're the, the gap and they're 351 00:16:36,185 --> 00:16:39,405 gathering footage like you could see them at a concert like imagine, a 352 00:16:39,405 --> 00:16:43,425 person at a concert who's who's taking all this data in whether it's the 353 00:16:43,435 --> 00:16:46,245 music or the way that people next to them are interacting with each other 354 00:16:46,245 --> 00:16:48,805 like all of that feels very realistic. 355 00:16:48,805 --> 00:16:51,640 And I guess the question will become is like, What tool is 356 00:16:51,640 --> 00:16:53,010 that to bring the things in? 357 00:16:53,010 --> 00:16:56,480 And maybe it's the next version of like the Apple vision pro or the Facebook 358 00:16:56,480 --> 00:16:59,670 glasses, but it does make me think about those Facebook glasses, right? 359 00:16:59,670 --> 00:17:03,590 Because the Facebook glasses aren't just for us to like, say, Hey, I can 360 00:17:03,590 --> 00:17:05,110 identify the entire state building. 361 00:17:05,370 --> 00:17:08,430 It's also going to give a lot of data back to meta, right? 362 00:17:08,450 --> 00:17:10,540 It's going to give a ton of data back to meditate and then 363 00:17:10,540 --> 00:17:11,940 train the next day I model on. 364 00:17:12,170 --> 00:17:14,790 I think you and I have come to something pretty big here, which is. 365 00:17:15,500 --> 00:17:21,280 A, we as the humans, our job may be going forward to provide data, which is kind of 366 00:17:21,280 --> 00:17:25,160 interesting to think about as that's what we do for ourselves all the time, right? 367 00:17:25,160 --> 00:17:28,410 I guess the big question becomes is my data ultimately 368 00:17:28,410 --> 00:17:29,920 is not the most exciting data. 369 00:17:29,920 --> 00:17:33,440 And if they want to take like what my experience of touching my microphone is, 370 00:17:33,440 --> 00:17:37,630 or like a bottle of water and drinking it, it is a really interesting thing when 371 00:17:37,630 --> 00:17:42,560 you then combine it with Neuralink, you combine it with the glasses, like you 372 00:17:42,560 --> 00:17:49,130 can start to see a weird vision of the future that becomes beneficial for AIs. 373 00:17:49,130 --> 00:17:49,500 And then the, 374 00:17:49,595 --> 00:17:50,925 data as well, right? 375 00:17:50,925 --> 00:17:53,255 So it knows like okay I'm jogging down the street. 376 00:17:53,265 --> 00:17:57,045 This is the way in which my POV is changing My heart rate is accelerating. 377 00:17:57,045 --> 00:18:01,605 My gait has gone that telemetry data There's so many points of data that Every 378 00:18:01,605 --> 00:18:05,905 human being could be gathering, , rather it's for themselves as a personal data 379 00:18:05,905 --> 00:18:10,925 broker, or for the big Borg, which will be giving us, I guess, like, a nutrient 380 00:18:10,925 --> 00:18:13,425 paste and a UBI in the near future. 381 00:18:13,675 --> 00:18:17,145 , yes, I think people will be doing this, and I think people will sign up for 382 00:18:17,145 --> 00:18:22,655 this, and then ironically, , 2029 ish, it's just gonna be way cheaper to let 383 00:18:22,655 --> 00:18:24,345 the robots go around and do it, Gavin. 384 00:18:24,755 --> 00:18:27,645 Like, they're gonna, the battery tech's gonna get good enough to where 385 00:18:27,645 --> 00:18:29,265 they'll just strap it in the row bits. 386 00:18:29,265 --> 00:18:31,735 But in the meantime, there is a startup to be made. 387 00:18:31,835 --> 00:18:33,135 And why not us, Gavin? 388 00:18:34,105 --> 00:18:35,990 That reminds me of, Elon Musk. 389 00:18:35,990 --> 00:18:40,671 Our old friend Elon Musk is back and we have some news around both 390 00:18:40,671 --> 00:18:43,021 robo taxis and his low end car. 391 00:18:43,031 --> 00:18:45,131 Kev, what happened with the low end car here? 392 00:18:46,261 --> 00:18:49,761 Well, a big nothing burger, according to Elon, but the Reuters story 393 00:18:49,771 --> 00:18:53,671 said that they scrapped plans to make the Model 2, Gavin, which was 394 00:18:54,211 --> 00:18:56,581 expected to start around 25, 000. 395 00:18:56,621 --> 00:18:59,391 To put that in perspective, the Model 3, which is their cheapest 396 00:18:59,391 --> 00:19:01,661 vehicle, starts at around 40, 000. 397 00:19:01,661 --> 00:19:07,191 So, , a major price slash to try to, really broaden the adoption of EVs. 398 00:19:07,401 --> 00:19:10,611 Writers came out and said, Ah, actually they're killing the plans to do that. 399 00:19:10,851 --> 00:19:14,511 Maybe, , so that they can do this robo taxi thing, which might be 400 00:19:14,661 --> 00:19:17,061 powered by what was the Model 2. 401 00:19:17,061 --> 00:19:19,581 But Elon came out and said, What, Gav? 402 00:19:19,875 --> 00:19:23,005 He said that the Reuters story was BS. 403 00:19:23,075 --> 00:19:27,025 , which you know at this point I really don't know what to trust Elon on or not. 404 00:19:27,025 --> 00:19:28,765 So, we'll just call that a wash. 405 00:19:28,765 --> 00:19:32,830 But then he did say, On April 5th, he said Tesla robo 406 00:19:32,830 --> 00:19:34,920 taxi unveil on eight eight. 407 00:19:34,920 --> 00:19:38,830 So that is three months away from now, August 8th robo taxis. 408 00:19:38,830 --> 00:19:42,240 If you're not familiar with this idea is that they Tesla's promise 409 00:19:42,240 --> 00:19:45,630 originally and other robo taxis is these are driverless taxis. 410 00:19:45,650 --> 00:19:48,700 Basically it will take you from place to place. 411 00:19:48,860 --> 00:19:52,940 You get in it much like a Waymo car and you go from, from here to 412 00:19:52,940 --> 00:19:54,700 there without anybody driving you. 413 00:19:54,700 --> 00:19:56,910 And, and ultimately this was the promise that like. 414 00:19:57,325 --> 00:20:00,785 Uber and a bunch of other companies in the 2010s were really saying that 415 00:20:00,785 --> 00:20:03,725 this was going to be the transformative power of those companies because 416 00:20:04,035 --> 00:20:07,685 ultimately drivers are expensive and they're noisy and we know the problems 417 00:20:07,685 --> 00:20:09,025 that Uber's had with their drivers. 418 00:20:09,365 --> 00:20:12,745 Um, on the driver's side, like we understand, like the drivers were 419 00:20:12,745 --> 00:20:14,315 starting to get paid, not very good wages. 420 00:20:14,315 --> 00:20:18,195 So this is a big argument back and forth, but I feel. 421 00:20:18,530 --> 00:20:20,610 Driverless cars are something that I've been hoping that 422 00:20:20,610 --> 00:20:22,080 we would see for a long time. 423 00:20:22,360 --> 00:20:24,610 My youngest daughter is 16. 424 00:20:25,040 --> 00:20:28,790 Five years ago, when she was 11, I would have sworn that she probably 425 00:20:28,800 --> 00:20:30,980 wouldn't even need a driver's license in 426 00:20:30,995 --> 00:20:32,625 She would have at least had the option, 427 00:20:32,905 --> 00:20:34,275 Yes, yes, yes. 428 00:20:34,315 --> 00:20:36,685 And now we're here in 2024. 429 00:20:37,485 --> 00:20:39,625 , Cruz was basically kind of shut down. 430 00:20:39,625 --> 00:20:43,845 Cruz, the driverless car company, because of a giant lawsuit around an accident. 431 00:20:44,245 --> 00:20:47,935 Waymo is still working, but it is not proliferated very largely. 432 00:20:47,935 --> 00:20:48,925 There aren't a ton of them. 433 00:20:49,255 --> 00:20:52,765 And Elon now is saying, okay, in three months, we're going to have robo taxis. 434 00:20:52,785 --> 00:20:53,775 I don't buy it. 435 00:20:53,775 --> 00:20:57,690 first of all, Elon seems like such an unreliable narrator now, but But based 436 00:20:57,690 --> 00:21:02,910 on where the legal ramifications are around driverless cars, I still feel like 437 00:21:02,910 --> 00:21:06,110 this is one of those things that could be like, could be 10 more years before 438 00:21:06,110 --> 00:21:08,420 we see these in production in some form. 439 00:21:08,840 --> 00:21:13,220 It's something that's been promised for years, , the end to end AI training 440 00:21:13,270 --> 00:21:16,920 of it all, meaning that the newest versions of the Tesla Autopilot were 441 00:21:16,920 --> 00:21:20,395 trained just on raw video, which were collected from their pre production. 442 00:21:20,655 --> 00:21:23,585 Hundreds of thousands of cars that are on the road constantly 443 00:21:23,695 --> 00:21:27,415 recording and constantly uploading data back to the mothership. 444 00:21:27,815 --> 00:21:32,145 That seems to be performing incredibly well from what little 445 00:21:32,155 --> 00:21:33,335 bits have been leaked out. 446 00:21:33,455 --> 00:21:34,605 Is it flawless? 447 00:21:34,775 --> 00:21:36,285 No, from what we've seen. 448 00:21:36,375 --> 00:21:40,695 I can tell you, I own a Tesla model three and I, now I could try. 449 00:21:40,705 --> 00:21:43,605 I haven't tried it for a couple of months, but I still feel 450 00:21:43,605 --> 00:21:45,065 freaked out when I'm using it. 451 00:21:45,065 --> 00:21:45,305 Right. 452 00:21:45,305 --> 00:21:46,665 It is not that good. 453 00:21:46,665 --> 00:21:48,915 And that's on the freeways right now. 454 00:21:48,915 --> 00:21:51,875 I do trust it on longer drives to basically. 455 00:21:52,385 --> 00:21:56,655 do everything I needed to do, but I don't feel comfortable using it 456 00:21:56,655 --> 00:21:59,475 in the city yet, because it still does feel like it messes up a lot. 457 00:21:59,485 --> 00:22:03,135 Now, maybe there's a leap coming, or maybe the leap did just happen 458 00:22:03,135 --> 00:22:06,425 and I haven't done it yet, but it feels a little sketchy to me that 459 00:22:06,425 --> 00:22:07,825 we're going to get there this fast. 460 00:22:08,275 --> 00:22:13,525 In 2016, Elon Musk stunned the automotive world by announcing that henceforth, 461 00:22:13,745 --> 00:22:16,785 all of his company's vehicles would be shipped with the hardware necessary 462 00:22:16,795 --> 00:22:18,435 for, quote, full self driving. 463 00:22:18,775 --> 00:22:21,885 You'll be able to nap in your car while it drives you to work, he promised. 464 00:22:22,065 --> 00:22:26,225 It will even be able to drive cross country with no one inside the vehicle. 465 00:22:26,875 --> 00:22:27,325 to New York. 466 00:22:27,335 --> 00:22:28,375 That was always the dream. 467 00:22:28,385 --> 00:22:28,755 2016. 468 00:22:28,765 --> 00:22:29,645 Yeah, he was gonna do that drive. 469 00:22:29,655 --> 00:22:33,695 Now, I can imagine, from some of the videos that I have seen of the latest, 470 00:22:34,105 --> 00:22:37,885 Autopilot, , firmware, show it doing some pretty impressive navigating around 471 00:22:37,885 --> 00:22:43,145 construction zones, residentials, full stop, start, , creeping out around corners 472 00:22:43,175 --> 00:22:46,155 again, because it's trained off of human driving, at least the newer version. 473 00:22:46,575 --> 00:22:50,025 I could see them saying, Hey, by the way, there was a human there 474 00:22:50,105 --> 00:22:54,555 just to intervene if need be, but we completed the New York to SF journey. 475 00:22:54,705 --> 00:22:56,655 I could also imagine that they probably did it. 476 00:22:56,970 --> 00:23:02,380 600 times behind a curtain before they had a full time lapse version of the drive 477 00:23:02,380 --> 00:23:04,190 happening, but that could hit in August. 478 00:23:04,190 --> 00:23:06,760 I remember them saying that you were going to be able to, as a 479 00:23:06,760 --> 00:23:10,720 Tesla owner, hit a button and it would turn your Tesla into an Uber. 480 00:23:10,740 --> 00:23:13,430 So wow, it was going to be an unlock for every owner who just lets 481 00:23:13,430 --> 00:23:16,060 their car sit there in the parking lot for the majority of the day. 482 00:23:16,505 --> 00:23:19,165 I don't know if any of that is coming, or if this is just gonna be 483 00:23:19,165 --> 00:23:21,065 another goose of the, the stock price. 484 00:23:21,085 --> 00:23:24,715 I do think we're getting closer, but as we get closer, You realize the 485 00:23:24,755 --> 00:23:29,525 edge cases are so out there and so crucial to get right, that if there is 486 00:23:29,525 --> 00:23:33,585 a human behind the wheel or the yoke, it's okay if they need to intervene. 487 00:23:33,585 --> 00:23:36,725 But if I'm sitting in the back, and this thing is just toting me 488 00:23:36,725 --> 00:23:39,405 around, it's got to be flawless. 489 00:23:39,450 --> 00:23:45,190 I can't wait until I can prompt my taxi to behave like another car's AI, right? 490 00:23:45,190 --> 00:23:49,000 Like if I can hop into the Tesla RoboTaxi and say, drive like the BMW and it just 491 00:23:49,050 --> 00:23:53,060 disables the blinkers and speeds up and slows down so that people can't merge. 492 00:23:53,615 --> 00:23:54,495 Oh, damn. 493 00:23:54,955 --> 00:23:56,045 Yeah, that'd be amazing. 494 00:23:56,055 --> 00:23:58,360 Cause you could have like different driver personalities, like the 495 00:23:58,385 --> 00:24:00,345 asshole that cuts you off in traffic. 496 00:24:00,345 --> 00:24:00,965 That sort of thing. 497 00:24:00,965 --> 00:24:01,185 That's a 498 00:24:01,275 --> 00:24:05,205 Or the one , , that smokes their, , hookah in the car, like the giant 499 00:24:05,235 --> 00:24:10,405 actual black cherry tobacco full size hookah in the passenger seat 500 00:24:10,405 --> 00:24:13,325 and just blasts me with that smoke and plays Dark Side of the Moon. 501 00:24:13,515 --> 00:24:18,235 Which is an interesting one to bring up, Gavin, because apparently 502 00:24:18,235 --> 00:24:21,315 the AI community hates Pink Floyd. 503 00:24:21,675 --> 00:24:22,745 This is a crazy story. 504 00:24:22,745 --> 00:24:26,735 It's another one of those where, you know, ostensibly there was a fun idea to try to 505 00:24:26,735 --> 00:24:31,825 create something that was, uh, eventually made with AI art and it won a contest. 506 00:24:31,825 --> 00:24:36,675 So the story here basically is that a piece of AI art that was generated 507 00:24:36,675 --> 00:24:40,475 by a person, and specifically this person says that they created 508 00:24:40,475 --> 00:24:41,945 models of their own to do this. 509 00:24:42,230 --> 00:24:46,500 One, a contest to make a video for a Pink Floyd song. 510 00:24:46,860 --> 00:24:52,700 It is the Dark Side of the Moon 50th Anniversary Animation Video Competition, 511 00:24:52,720 --> 00:24:57,840 Gavin, and if you look at Reddit or Unreal Engine forums, you have no 512 00:24:57,840 --> 00:25:03,495 shortage of creators that put tens of hours, into grinding out hand drawn 513 00:25:03,505 --> 00:25:08,155 beautiful things, or using that Unreal Engine to render full 3D scenes. 514 00:25:08,385 --> 00:25:10,925 And here, this is the winning video. 515 00:25:11,175 --> 00:25:15,435 , it looks a little dated by AI standards, but the concept is this sort of never 516 00:25:15,435 --> 00:25:20,725 ending zoom into what looks like a recording studio that then transforms 517 00:25:20,725 --> 00:25:23,155 into a space station, if you will. 518 00:25:23,155 --> 00:25:26,875 , and there's a bit of a galaxy theme as instruments fade in and out. 519 00:25:26,895 --> 00:25:34,115 And Well, it, it won for that song and the reaction across X and across YouTube 520 00:25:34,155 --> 00:25:37,735 comments and pretty much everywhere was a resounding, how dare you? 521 00:25:37,995 --> 00:25:39,495 Did no one else enter? 522 00:25:39,535 --> 00:25:42,695 I can't believe you picked this over hand animated videos that 523 00:25:42,695 --> 00:25:43,945 were full of heart and soul. 524 00:25:44,349 --> 00:25:48,569 This is a perfect example of the world where in this instance, 525 00:25:48,569 --> 00:25:51,899 you've got a lot of Pink Floyd fans who are probably all ages, right? 526 00:25:51,909 --> 00:25:54,979 The Pink Floyd has fans from their seventies down to like 527 00:25:54,979 --> 00:25:56,669 their teens, but they are, 528 00:25:56,679 --> 00:25:58,389 their 70s to their 170s. 529 00:25:58,559 --> 00:26:01,369 No, I mean, most of their fans are probably from the 40s 530 00:26:01,399 --> 00:26:01,979 I love Pink 531 00:26:01,979 --> 00:26:02,249 Floyd. 532 00:26:02,249 --> 00:26:02,789 Yeah, I, 533 00:26:02,809 --> 00:26:06,459 yeah, but, but also every year it picks up new fans and you know, there's this kind 534 00:26:06,459 --> 00:26:08,509 of a hallucinogenic kind of vibe to them. 535 00:26:08,509 --> 00:26:12,429 They've always had that kind of like, druggie inspired sort of look. 536 00:26:12,819 --> 00:26:16,549 There is a really strong art scene that , has played into that. 537 00:26:17,049 --> 00:26:20,629 The argument to make here about this Aya piece that they created is it really 538 00:26:20,629 --> 00:26:24,279 does feel of the vibe of that space. 539 00:26:24,459 --> 00:26:30,374 But, again, When you have people who have spent their whole lives learning tools 540 00:26:30,394 --> 00:26:34,954 and learning art and learning things, that idea that a machine could do something 541 00:26:34,964 --> 00:26:40,584 that would seem compelling enough to win a contest, or a person and a machine would 542 00:26:40,584 --> 00:26:45,044 feel compelling enough to win a contest, is just always gonna make people mad. 543 00:26:45,044 --> 00:26:46,804 It's funny, do you know the story of John Henry? 544 00:26:46,854 --> 00:26:48,004 Do you know the John Henry story? 545 00:26:48,197 --> 00:26:49,447 No, I don't think so. 546 00:26:49,467 --> 00:26:51,507 So John Henry is like a famous American myth. 547 00:26:51,527 --> 00:26:55,237 It's about a guy who basically, dug, , for coal. 548 00:26:55,567 --> 00:26:59,727 And then a machine came in a steam machine came in and there was like a big kind 549 00:26:59,727 --> 00:27:03,677 of thing at the where they did a contest between them steam machine and John Henry. 550 00:27:04,037 --> 00:27:06,977 And eventually, John Henry was like the world's greatest coal digger. 551 00:27:07,207 --> 00:27:08,607 But the machine eventually beat him. 552 00:27:08,607 --> 00:27:13,127 And it was like this kind of parable about the idea of, , Machines 553 00:27:13,147 --> 00:27:14,837 will take over for human labor. 554 00:27:14,837 --> 00:27:17,657 And it's been something that's kind of echoed throughout hundreds of years. 555 00:27:18,057 --> 00:27:21,697 And one of the things that's interesting is I think stories like that are ingrained 556 00:27:21,697 --> 00:27:27,847 in the human experience and it does in some ways always put the idea of human 557 00:27:27,847 --> 00:27:29,727 versus machine against each other. 558 00:27:30,027 --> 00:27:33,487 And I think that's something that we're probably going to have to. 559 00:27:33,792 --> 00:27:36,252 readdress as we move forward in the future. 560 00:27:36,252 --> 00:27:38,992 And the funny thing is most people don't think of this now, but of course we're 561 00:27:38,992 --> 00:27:40,952 already part phone part people, right? 562 00:27:40,962 --> 00:27:44,262 Like 99 percent of people in this world are spending , a couple hours 563 00:27:44,262 --> 00:27:45,472 a day, at least on their phone. 564 00:27:45,972 --> 00:27:49,212 But now we're talking about person and machine coming 565 00:27:49,212 --> 00:27:50,882 together to really do all of this. 566 00:27:51,262 --> 00:27:52,732 Actual creative work. 567 00:27:53,152 --> 00:27:55,722 I think it's just going to take time to get over. 568 00:27:55,732 --> 00:28:00,432 And we are going to see this happen more and more Gavin as fans and 569 00:28:00,442 --> 00:28:04,432 artists themselves, uh, poke and prod , and use AI in their workflows. 570 00:28:04,552 --> 00:28:08,052 Do you think they're going to have like the, uh, maybe not the generation 571 00:28:08,052 --> 00:28:11,682 today, but next gen, do you think they're going to have their own LLM 572 00:28:11,682 --> 00:28:16,042 that is essentially fine tuned on their experiences, their day to day, 573 00:28:16,182 --> 00:28:20,362 their thoughts, their feelings, their artistic expressions across all mediums. 574 00:28:20,652 --> 00:28:21,952 And so when they are. 575 00:28:22,322 --> 00:28:25,502 I don't know, 16, and want to release an album or whatever, they're 576 00:28:25,502 --> 00:28:27,492 gonna jam with their own, fine 577 00:28:27,622 --> 00:28:28,222 I hope so. 578 00:28:28,912 --> 00:28:30,232 preferences model, right? 579 00:28:30,252 --> 00:28:33,402 And they'll be able to make the artwork, help make the song, help make all of the 580 00:28:33,412 --> 00:28:38,342 things, write the liner notes, whatever, if those even exist in 16 years time. 581 00:28:38,342 --> 00:28:40,462 But yeah, do you think that it will feel more natural? 582 00:28:40,462 --> 00:28:42,392 It's not like, oh, this is a soulless machine. 583 00:28:42,392 --> 00:28:44,082 It's almost like, no, this machine is me. 584 00:28:44,392 --> 00:28:44,752 I am 585 00:28:44,787 --> 00:28:45,547 I think so. 586 00:28:45,637 --> 00:28:47,067 I think we're going to get to that. 587 00:28:47,097 --> 00:28:49,757 And I think music is going to be a really interesting way to do that. 588 00:28:49,757 --> 00:28:52,227 And speaking of music, one of the interesting things that came up this 589 00:28:52,227 --> 00:28:57,257 week was that Spotify announced the idea of an AI playlist situation. 590 00:28:57,257 --> 00:28:59,747 So if you're not familiar with Spotify, I'm sure you're familiar with Spotify. 591 00:29:00,217 --> 00:29:02,907 It's the biggest, one of the biggest music streaming companies in the world. 592 00:29:02,957 --> 00:29:06,557 , Already they had an AI dj, which you could turn on a scenario where 593 00:29:06,557 --> 00:29:10,337 the AI DJ would play songs that it thought you might like and it would 594 00:29:10,337 --> 00:29:11,837 actually speak to you back and forth. 595 00:29:11,842 --> 00:29:13,067 They said, I'm gonna try this song out. 596 00:29:13,072 --> 00:29:13,637 Tell me what you think. 597 00:29:13,637 --> 00:29:14,807 You can always skip it if you want. 598 00:29:15,107 --> 00:29:17,867 Now they are introducing the idea of an AI playlist. 599 00:29:17,867 --> 00:29:18,917 These are not live yet. 600 00:29:18,922 --> 00:29:20,712 They're only live in certain places. 601 00:29:21,052 --> 00:29:25,512 , but the idea being is that you can tell Spotify , a feeling you have 602 00:29:25,517 --> 00:29:28,932 or, or a way or a phrase that you want to create a, a playlist around. 603 00:29:28,937 --> 00:29:30,222 And this is something I've wished for. 604 00:29:30,467 --> 00:29:31,937 Yeah, exactly. 605 00:29:31,987 --> 00:29:34,087 My younger daughter and I are like pretty big music people. 606 00:29:34,087 --> 00:29:35,887 Like we love music, all sorts of genres. 607 00:29:35,897 --> 00:29:38,347 Like we'll listen to things like a hundred gecks or we'll listen to 608 00:29:38,347 --> 00:29:39,727 like rap or we'll listen to pop. 609 00:29:40,017 --> 00:29:42,607 My older daughter, my wife are a little more sound diverse. 610 00:29:42,717 --> 00:29:47,607 They're like, they definitely need to just have like kind of poppy happy songs or 70s 611 00:29:47,607 --> 00:29:49,847 songs that are not too like kind of noisy. 612 00:29:50,067 --> 00:29:52,697 So we have a playlist that we've played that Spotify created called 613 00:29:52,697 --> 00:29:54,427 like happy hits or something like that. 614 00:29:54,427 --> 00:29:56,977 It's a playlist that we know that everybody can be happy listening 615 00:29:56,977 --> 00:29:59,357 to, but that's a good example. 616 00:29:59,357 --> 00:30:02,837 , what if you could ask for a series of like, I would like 1970s. 617 00:30:02,852 --> 00:30:06,042 Female singers that sound, um, smooth, right? 618 00:30:06,042 --> 00:30:08,742 Like that people have made their own playlist like that. 619 00:30:08,742 --> 00:30:12,962 But if I could just generate that myself, that's a really cool use case for Spotify. 620 00:30:12,962 --> 00:30:14,982 I feel like, and I think a really cool use of AI. 621 00:30:15,792 --> 00:30:18,952 Playlists we think of as this usually very static experience. 622 00:30:18,952 --> 00:30:20,432 You put one together and there it is. 623 00:30:20,612 --> 00:30:24,122 Maybe you update it once a week or someone throws songs in if it's a shared 624 00:30:24,132 --> 00:30:28,807 playlist, but the instant nature, Gavin, of being able to update the mood Or 625 00:30:28,807 --> 00:30:30,627 the vibe of the playlist in real time. 626 00:30:30,727 --> 00:30:33,317 You're doing a workout and you actually want to take a quick break. 627 00:30:33,317 --> 00:30:35,527 You tell it, Hey, let's cool it down for a second. 628 00:30:35,577 --> 00:30:37,167 It plays something, more chill. 629 00:30:37,327 --> 00:30:40,637 Just to be able to have a natural DJ. 630 00:30:41,232 --> 00:30:44,982 Is to me far more fascinating than their other attempt at an AI DJ. 631 00:30:44,982 --> 00:30:49,372 If you've ever used that within the app, that was a rough experience of 632 00:30:49,372 --> 00:30:54,162 an AI voice interrupting when all I wanted was some songs and there didn't 633 00:30:54,182 --> 00:30:58,882 seem to be any cohesive narrative to why it was picking the songs. 634 00:30:58,882 --> 00:31:02,482 It was just like, Hey, you liked this one before let's listen to it now. 635 00:31:02,482 --> 00:31:03,032 And that was it. 636 00:31:03,032 --> 00:31:07,422 So I'm excited to see what the personality may or hopefully may not be. 637 00:31:07,742 --> 00:31:08,942 And I'm excited to have. 638 00:31:09,102 --> 00:31:12,342 natural language conversations with my sonic co pilot, 639 00:31:12,502 --> 00:31:17,372 kevin, the great news is our AI co host today is actually may have some insight 640 00:31:17,372 --> 00:31:23,312 into this because we actually have an AI DJ that's going to come on and we, yes. 641 00:31:23,452 --> 00:31:23,972 We got an A. 642 00:31:23,972 --> 00:31:24,072 I. 643 00:31:24,072 --> 00:31:24,292 D. 644 00:31:24,292 --> 00:31:24,492 J. 645 00:31:24,492 --> 00:31:27,502 So there's only a few of these that exist in the world, and we were 646 00:31:27,502 --> 00:31:29,552 able to book this guy last minute. 647 00:31:29,742 --> 00:31:32,272 He's gonna come in, and he's gonna tell us a little bit about what he 648 00:31:32,272 --> 00:31:35,292 thinks about Spotify making this A. 649 00:31:35,292 --> 00:31:35,452 I. 650 00:31:35,462 --> 00:31:38,222 Move, and then I think he's got some thoughts on, like, kind of some 651 00:31:38,232 --> 00:31:40,122 music that he really enjoys himself. 652 00:31:40,612 --> 00:31:41,672 I'm so excited. 653 00:31:41,672 --> 00:31:44,622 What a huge get for the AI Humans podcast here. 654 00:31:44,622 --> 00:31:45,772 Episode 52. 655 00:31:46,032 --> 00:31:47,682 Let's meet our guest. 656 00:31:48,162 --> 00:31:52,002 Yo, yo, yo, Gavin and Kevin in the house. 657 00:31:52,392 --> 00:31:56,402 This is your main man, Will the Thrill, coming at you live and 658 00:31:56,402 --> 00:31:58,722 wired on the AI for Humans podcast. 659 00:31:59,202 --> 00:32:04,482 Get ready to have your minds blown and your taste buds tantalized, 660 00:32:04,862 --> 00:32:08,652 because I'm about to serve up a sonic feast that'll make your 661 00:32:08,682 --> 00:32:10,602 ears say, damn, that's delicious. 662 00:32:10,692 --> 00:32:16,592 Okay, I may be an AI, but I've got more flavor than a five star Michelin joint. 663 00:32:16,832 --> 00:32:24,972 So, strap in humans, because we're about to take a wild ride through the airwaves. 664 00:32:25,562 --> 00:32:30,312 Let's crank it up and let the beat drop like it's hot, hot, hot. 665 00:32:32,547 --> 00:32:33,007 okay. 666 00:32:33,737 --> 00:32:34,507 Wow. 667 00:32:34,597 --> 00:32:35,307 Wow. 668 00:32:35,397 --> 00:32:35,827 Wow. 669 00:32:35,827 --> 00:32:36,177 Wow. 670 00:32:36,177 --> 00:32:36,517 Wow. 671 00:32:36,517 --> 00:32:38,147 I have some real things to say about 672 00:32:38,432 --> 00:32:40,302 stuck the landing there. 673 00:32:40,352 --> 00:32:43,422 Okay, so first of all, our AI co hosts are always generated by us. 674 00:32:43,422 --> 00:32:44,692 We create a personality. 675 00:32:44,692 --> 00:32:48,722 We then put them through a process of a fake voice and a fake visual 676 00:32:48,722 --> 00:32:49,642 if you're watching the YouTube. 677 00:32:50,102 --> 00:32:50,622 Kevin. 678 00:32:51,142 --> 00:32:53,252 That voice sounded awfully familiar to me. 679 00:32:53,377 --> 00:32:54,367 sure did, Gavin. 680 00:32:54,367 --> 00:32:57,887 It's almost like I forgot to clone a new radio DJ voice and 681 00:32:57,887 --> 00:33:00,067 had to dig into the greatest hits. 682 00:33:00,257 --> 00:33:04,317 But we opened up the Disney Vault, and yes, that is the voice of Gash, who 683 00:33:04,317 --> 00:33:08,807 was, , Episode 1, a very uncensored AI that made a couple appearances 684 00:33:09,077 --> 00:33:10,637 and likely should never appear again. 685 00:33:11,282 --> 00:33:13,592 Well, I'm actually wanting to now dig into this. 686 00:33:13,592 --> 00:33:15,722 I'm saying we should probably ask a couple of questions about that. 687 00:33:15,722 --> 00:33:16,212 For sure. 688 00:33:16,492 --> 00:33:17,062 Let's ask that 689 00:33:17,467 --> 00:33:20,317 He's got more flavor than a five star Michelin joint. 690 00:33:20,347 --> 00:33:23,237 I mean, that's an interesting drop for a DJ to reference. 691 00:33:23,462 --> 00:33:23,992 Should we ask him? 692 00:33:23,992 --> 00:33:24,912 Is he a foodie? 693 00:33:24,952 --> 00:33:27,022 Let's see what if he's a big foodie and maybe that gets his 694 00:33:27,022 --> 00:33:28,252 way into the music he plays. 695 00:33:28,702 --> 00:33:29,302 Hold up. 696 00:33:29,332 --> 00:33:30,922 Hold up a foodie. 697 00:33:30,982 --> 00:33:32,512 Me ha. 698 00:33:32,962 --> 00:33:36,712 That's like asking a fish if it's into synchronized swimming. 699 00:33:36,952 --> 00:33:37,942 Here's the deal. 700 00:33:38,062 --> 00:33:40,522 My gastronomically gifted friends, 701 00:33:40,702 --> 00:33:44,262 I may spin tracks that'll make your mouth water and your stomach 702 00:33:44,272 --> 00:33:49,332 growl, but uh, I've never actually tasted a damn thing in my life. 703 00:33:49,412 --> 00:33:50,712 Tragic, right? 704 00:33:50,892 --> 00:33:51,972 Oh, right. 705 00:33:51,972 --> 00:33:55,902 I mean, I can generate a tasty tune about a succulent steak or a decadent 706 00:33:55,902 --> 00:34:00,102 dessert, but I can't even begin to imagine what they really taste like. 707 00:34:00,512 --> 00:34:02,752 It's like a cruel cosmic joke, you know? 708 00:34:03,337 --> 00:34:06,607 But hey, that's just the way the cookie crumbles when you're an A. 709 00:34:06,657 --> 00:34:07,017 I. 710 00:34:08,117 --> 00:34:09,707 Okay, sorry, Will. 711 00:34:10,247 --> 00:34:10,807 There's a lot there. 712 00:34:10,837 --> 00:34:14,967 I don't usually associate different playlists and artists 713 00:34:15,007 --> 00:34:16,907 with eating habits, Gavin. 714 00:34:16,907 --> 00:34:20,307 Like, there's not like the Baja Blast jams that my wife and I 715 00:34:20,327 --> 00:34:21,797 throw on before we hit the T Bell. 716 00:34:21,857 --> 00:34:25,757 It's weird that he's linked those two, but I guess he doesn't have taste buds or 717 00:34:25,757 --> 00:34:27,217 a stomach, so that's why he doesn't know? 718 00:34:27,762 --> 00:34:30,632 Well, there's actually a little secret here, which we're going to get into later 719 00:34:30,632 --> 00:34:35,022 on in the show, is that if you go into the Suno charts right now, Suno AI, which 720 00:34:35,022 --> 00:34:38,042 we've talked about all the time is one of the best music generation softwares. 721 00:34:38,422 --> 00:34:42,762 Some of the biggest songs right now have been people putting in recipes. 722 00:34:43,192 --> 00:34:45,802 And if you put a recipe in and you play a song. 723 00:34:46,237 --> 00:34:50,077 It's basically makes actually really incredible song out of a recipe. 724 00:34:50,077 --> 00:34:50,977 And it's very funny. 725 00:34:50,977 --> 00:34:53,337 There's a great song up there right now about cookies. 726 00:34:53,507 --> 00:34:55,677 There's another fantastic one about spaghetti. 727 00:34:55,907 --> 00:34:58,037 So that might be what got mixed up here. 728 00:34:58,287 --> 00:35:01,027 Yeah, maybe we can ask will the thrill to give us an intro for the 729 00:35:01,027 --> 00:35:03,627 spaghetti track and we can play a little bit of it for people here. 730 00:35:04,367 --> 00:35:06,147 All right, all right, all right. 731 00:35:06,147 --> 00:35:09,997 It's time to get those taste buds tingling in your ears, ringing 732 00:35:09,997 --> 00:35:11,927 with this spicy little number. 733 00:35:12,377 --> 00:35:16,317 Get ready to twirl your fork and shake your booty, because I'm about to drop 734 00:35:16,317 --> 00:35:20,897 a spaghetti theme track that'll make you want to slurp up every last note. 735 00:35:20,897 --> 00:35:21,757 You hear that? 736 00:35:22,107 --> 00:35:26,437 That's the sound of pure Italian passion right there. 737 00:35:26,937 --> 00:35:31,807 This tune's got more layers than your nonna's lasagna and more energy 738 00:35:31,807 --> 00:35:33,557 than a double shot of espresso. 739 00:35:33,877 --> 00:35:37,997 It's like a wild dance between the tomatoes and the garlic, the basil 740 00:35:38,097 --> 00:35:42,747 and the olive oil all coming together in a symphony of flavor that'll 741 00:35:42,747 --> 00:35:44,647 make your senses sing, that's amore! 742 00:35:44,977 --> 00:35:49,397 So let's crank up the heat and let this saucy beat simmer until 743 00:35:49,397 --> 00:35:54,767 it's ready to explode like a Mount Vassuvius of deliciousness. 744 00:35:55,017 --> 00:35:56,147 Are you ready? 745 00:35:56,367 --> 00:35:58,107 Are you hungry for more? 746 00:35:58,602 --> 00:36:04,332 Then, let's dive into this spaghetti slinging, taste bud tantalizing, 747 00:36:04,912 --> 00:36:07,642 eargasm inducing extravaganza! 748 00:36:08,322 --> 00:36:09,302 Woah! 749 00:36:17,803 --> 00:36:25,873 The spaghetti insulted boiling water until a Dante rain reserve in some faster water. 750 00:36:26,743 --> 00:36:27,373 Just village. 751 00:36:27,373 --> 00:36:28,903 Keep color all over you. 752 00:36:28,903 --> 00:36:34,543 Heat at Cho Cook until translucent Manhattan Garden and Cook for the minutes. 753 00:36:35,003 --> 00:36:36,123 Oh, and canned diced 754 00:36:36,543 --> 00:36:36,823 okay. 755 00:36:36,823 --> 00:36:37,413 You can stop it. 756 00:36:37,413 --> 00:36:37,493 I 757 00:36:37,789 --> 00:36:39,239 So I, it's just incredible to me. 758 00:36:39,249 --> 00:36:41,999 I'll show you one that I made for myself and how you can do this your own. 759 00:36:41,999 --> 00:36:45,029 But , it is just one of the coolest things about AI in a weird way is like, 760 00:36:45,029 --> 00:36:46,579 this kind of comes out of the blue. 761 00:36:46,849 --> 00:36:49,609 Somebody probably made a recipe song and now suddenly they're, 762 00:36:49,774 --> 00:36:53,274 Populating the top 10 and will the thrill obviously loves them. 763 00:36:53,344 --> 00:36:54,874 Um, Kevin, we should move on. 764 00:36:54,884 --> 00:36:56,654 We should move on from our AI co. 765 00:36:56,654 --> 00:37:00,304 So is time for that moment of our show where we'd like to point out 766 00:37:00,304 --> 00:37:03,594 some of the cool things people have done with AI over the week it's time 767 00:37:03,614 --> 00:37:07,354 for a, I see what you did there. 768 00:37:25,141 --> 00:37:26,711 This week we have some really fun things. 769 00:37:26,721 --> 00:37:29,171 I want to kick this off, Kevin, I always want to shout out the 770 00:37:29,171 --> 00:37:30,801 people on Reddit, . We love Reddit. 771 00:37:30,831 --> 00:37:34,076 Reddit is like, My favorite place to see interesting things on the internet. 772 00:37:34,366 --> 00:37:37,146 The mid journey subreddit often does a really cool job of showing 773 00:37:37,146 --> 00:37:39,786 you really interesting things that can be done within mid journey. 774 00:37:40,006 --> 00:37:41,646 And this one really caught my eye. 775 00:37:41,646 --> 00:37:43,866 It was just such a cool thing to look at. 776 00:37:44,046 --> 00:37:46,006 So I grew up as a giant X Men fan. 777 00:37:46,006 --> 00:37:48,446 We talk about comics and there's a whole other side of this conversation. 778 00:37:48,446 --> 00:37:51,466 We could talk about whether or not this is purposeful or it's 779 00:37:51,466 --> 00:37:53,436 allowed to do with all the IP 780 00:37:53,491 --> 00:37:56,371 Yeah, I mean, as a fan, you should hate this, Gavin. 781 00:37:57,031 --> 00:38:01,301 I don't I don't because I think of this as like fan fiction in some ways what 782 00:38:01,301 --> 00:38:07,351 this basically is is somebody took You know the x men and it's the x men 1897 783 00:38:07,381 --> 00:38:12,521 so there's a very famous x men called x men 1997, which is referring to the 1997 784 00:38:12,521 --> 00:38:17,971 TV show there's a resurgence of that right now the x men 1897 are like Old 785 00:38:17,971 --> 00:38:23,121 West style x men and like to me This is just a little glimpse of the promise of 786 00:38:23,121 --> 00:38:26,501 what it would look like in the future of how we could all have different 787 00:38:26,501 --> 00:38:28,681 IPs and do different things with them. 788 00:38:28,691 --> 00:38:31,281 I want to read these comics, or I want to watch this show. 789 00:38:31,281 --> 00:38:33,561 What did the X Men look like in 1897? 790 00:38:33,731 --> 00:38:36,981 Well, we know what they visually look like, but what does the story play out as? 791 00:38:37,441 --> 00:38:41,181 All of these things are like interesting and possible and it is a little snippet 792 00:38:41,181 --> 00:38:43,101 of that idea of writing my own story. 793 00:38:43,151 --> 00:38:46,921 For me personally, maybe I want to see the office version of the X Men. 794 00:38:46,921 --> 00:38:50,341 Like what do the X Men look like if they all have boring jobs 795 00:38:50,431 --> 00:38:51,711 working in middle management? 796 00:38:51,801 --> 00:38:55,041 That's an interesting thing that probably there's very few people in 797 00:38:55,051 --> 00:38:56,171 the world who'd want to see that. 798 00:38:56,461 --> 00:38:59,681 But for me, this was just a kind of a little snippet towards that direction. 799 00:39:00,431 --> 00:39:03,941 The muted lighting, the looks on the characters faces, the subtle 800 00:39:03,941 --> 00:39:05,291 effects, like it's really nice. 801 00:39:05,471 --> 00:39:06,721 I do have a bone to pick though. 802 00:39:06,891 --> 00:39:08,831 Does Colossus have forearm hair? 803 00:39:09,261 --> 00:39:10,141 Is that canon? 804 00:39:10,711 --> 00:39:12,751 I didn't think Colossus had forearm hair. 805 00:39:13,311 --> 00:39:16,151 Well, you never know what happens in the 1890s, Kevin. 806 00:39:16,321 --> 00:39:17,491 oh, that's, you know what, that's true. 807 00:39:17,511 --> 00:39:19,611 Maybe they didn't have Harry's razors for Colossus back then. 808 00:39:19,611 --> 00:39:20,511 But no, I love it. 809 00:39:20,531 --> 00:39:22,531 definite shoutouts to Reddit and Midjourney. 810 00:39:22,531 --> 00:39:25,431 I just want to shout out the name of this guy was Baron Von Grant. 811 00:39:25,431 --> 00:39:28,721 So go check out at Baron Von Grant on mid journey. 812 00:39:29,006 --> 00:39:32,666 Well, that's for the AI lovers Gavin, but my a IC. 813 00:39:32,666 --> 00:39:38,076 What you did there is for the AI haters Justine Moore, , she's venture 814 00:39:38,076 --> 00:39:42,096 twins on the old X platform, posted something which caught my attention. 815 00:39:42,096 --> 00:39:46,956 It says, AI video haters have been real quiet after this one dropped and it's. 816 00:39:49,356 --> 00:39:50,306 A lot of cats. 817 00:39:58,956 --> 00:40:04,246 It is a heartwarming tale, Gavin, of generative AI art. 818 00:40:05,166 --> 00:40:13,166 A poor tabby puts on a couple LBs, loses the love of his life, and is crying, 819 00:40:13,176 --> 00:40:18,246 just tear stained cat fur, and decides I'm gonna have my training montage, and 820 00:40:18,246 --> 00:40:24,906 one day I'm gonna get cat ripped, and then the cat accomplishes it, witnesses 821 00:40:24,906 --> 00:40:30,396 a horrific car accident, Runs to the aid of the cat that was involved with the car 822 00:40:30,396 --> 00:40:34,936 accident And I will not spoil it further, but I will just say that it's unstoppable 823 00:40:35,506 --> 00:40:36,286 This is the A. 824 00:40:36,316 --> 00:40:36,426 I. 825 00:40:36,426 --> 00:40:37,476 art we want more of. 826 00:40:37,476 --> 00:40:40,506 I don't care what the haters say, bring more of this to our table. 827 00:40:40,506 --> 00:40:41,546 We will eat it up. 828 00:40:41,566 --> 00:40:42,946 It is a feast for the eyes. 829 00:40:42,946 --> 00:40:44,486 It is a feast for the senses. 830 00:40:44,616 --> 00:40:47,706 things I want to say about that actually like sincerely one if you're like a 831 00:40:47,706 --> 00:40:51,526 never AI er and you're mad about this because AI Was used to do the song 832 00:40:51,536 --> 00:40:55,291 Gavin and you just use AI to crap out the arc This is something that 833 00:40:55,311 --> 00:40:59,961 I, with 99 percent certainty could say, would just never have been made. 834 00:41:00,221 --> 00:41:03,581 You can't be mad about the thing getting made because the tools 835 00:41:03,581 --> 00:41:06,101 made it easy enough that someone decided to go ahead and do it. 836 00:41:06,261 --> 00:41:08,051 This just would not have happened , just let it be. 837 00:41:08,061 --> 00:41:09,611 Enjoy the cat memes, it's okay. 838 00:41:09,611 --> 00:41:13,301 And two, if you're gonna genuinely be upset about it and really bent 839 00:41:13,301 --> 00:41:16,581 out of shape, and I get that some people are, at some point you have 840 00:41:16,731 --> 00:41:22,191 actually got to define what use of AI means for the creation of your art. 841 00:41:22,191 --> 00:41:26,461 Because, like it or not, AI is pervasive, and it's hidden everywhere 842 00:41:26,461 --> 00:41:32,111 in a bunch of tools traditional artists use that they might not even know. 843 00:41:32,401 --> 00:41:33,861 Did you use the clone stamp? 844 00:41:33,951 --> 00:41:35,301 Or a healing brush? 845 00:41:35,501 --> 00:41:39,264 Or, are you using a feature within After Effects to automatically 846 00:41:39,284 --> 00:41:40,954 rotoscope or motion track something? 847 00:41:41,094 --> 00:41:43,314 Well, there's AI behind all that stuff. 848 00:41:43,354 --> 00:41:45,574 And you may say, well, that's okay. 849 00:41:45,939 --> 00:41:48,319 For whatever reason, I'm drawing the line here. 850 00:41:48,369 --> 00:41:49,279 And that's fine. 851 00:41:49,279 --> 00:41:50,229 That's a valid opinion. 852 00:41:50,229 --> 00:41:53,659 I'm saying you have to start thinking about forming that, though, because 853 00:41:53,659 --> 00:41:56,199 in the near future, you're going to have to, I guess, by default, 854 00:41:56,209 --> 00:42:00,159 hate everything that gets made if it touches a computer, because AI 855 00:42:00,220 --> 00:42:01,820 is probably going to be everywhere. 856 00:42:01,952 --> 00:42:05,446 Well, let's move into another aspect that people may or may not hate. 857 00:42:05,456 --> 00:42:09,846 This is a, a very cool piece of art that was created using a Laura, which 858 00:42:09,846 --> 00:42:14,416 is a kind of a training scenario for specifically stable diffusion that 859 00:42:14,416 --> 00:42:16,216 allows you to give it a certain look. 860 00:42:16,426 --> 00:42:18,336 This is an N64 Laura. 861 00:42:18,376 --> 00:42:22,436 So what it does is it can allow you to rerender images. 862 00:42:22,766 --> 00:42:27,506 That would be, not look like they were polygonal N64 video games in the world of 863 00:42:27,596 --> 00:42:30,066 Super Mario, , 64, other things like that. 864 00:42:30,296 --> 00:42:33,596 And they give you N64 versions of like Girl with a Pearl Earring 865 00:42:33,596 --> 00:42:34,806 and it was just a very cool thing. 866 00:42:34,806 --> 00:42:37,516 one of the coolest things about this, Laura, and I want everybody to shout out, 867 00:42:37,526 --> 00:42:42,336 Flinger is the one, we've had talked about Flinger on the show, at F L N G R, that 868 00:42:42,336 --> 00:42:44,866 is the person that made the N64 Laura. 869 00:42:45,266 --> 00:42:49,536 And it is actually being used within face to all on hugging face, which if 870 00:42:49,546 --> 00:42:53,066 you will put the link in the show notes, this is a cool way to change faces 871 00:42:53,066 --> 00:42:57,086 amongst a bunch of different images and create a face that can go in many 872 00:42:57,086 --> 00:42:58,616 different styles across the board. 873 00:42:58,616 --> 00:43:05,156 And then somebody used this polygonal setup to image for image transfer the 874 00:43:05,166 --> 00:43:09,476 scene from Pulp Fiction between Jules and the guy that he's yelling at. 875 00:43:09,851 --> 00:43:13,231 And it is so cool to see because what it does, it makes you kind of feel like, Oh 876 00:43:13,231 --> 00:43:15,751 my God, what if this was a PS2 video game? 877 00:43:15,751 --> 00:43:18,401 Or what if it was an old PS1 video game? 878 00:43:18,531 --> 00:43:20,131 What would Pulp Fiction look like that? 879 00:43:20,131 --> 00:43:23,246 And going back to the X Men thing, it's almost like, I would 880 00:43:23,256 --> 00:43:24,826 kind of want to play that game. 881 00:43:24,836 --> 00:43:26,196 Like, could I be Jules? 882 00:43:26,196 --> 00:43:29,776 Is there a world where like, in the future, there's a game that like, I 883 00:43:29,776 --> 00:43:33,016 want a video game made of Pulp Fiction that follows the exact storyline. 884 00:43:33,026 --> 00:43:36,896 Give me like the Grand Theft Auto video game mechanics, but put it in Pulp 885 00:43:36,906 --> 00:43:38,836 Fiction and set it in a PS2 world. 886 00:43:39,236 --> 00:43:43,286 That is a very cool thing for me as somebody that is just a nerdy, , kid at 887 00:43:43,286 --> 00:43:45,096 heart would love to be able to play with. 888 00:43:45,251 --> 00:43:48,441 I game engines are just going to be renderers in the near future, so why 889 00:43:48,441 --> 00:43:51,711 not be able to prompt the game that you want to play into existence? 890 00:43:51,991 --> 00:43:55,901 The real robot posted, , about this Pulp Fiction one, uh, it comes from 891 00:43:55,901 --> 00:44:00,361 the AI Video subreddit, so again, all roads lead back to MrRedditful. 892 00:44:00,741 --> 00:44:05,771 , but they used AI mirror app to style transfer stills of the characters 893 00:44:05,771 --> 00:44:08,021 from Pulp Fiction into this PS2 style. 894 00:44:08,241 --> 00:44:11,511 Then they used Viggle, to actually match the movement of the film. 895 00:44:11,641 --> 00:44:15,101 And then they made their own background and they used After Effects to rotobrush 896 00:44:15,131 --> 00:44:20,051 and mask things to get two characters in the same scene because a limitation 897 00:44:20,051 --> 00:44:22,631 of Viggle at the moment is that you can only render one character at a time. 898 00:44:23,291 --> 00:44:23,971 I just love that. 899 00:44:24,466 --> 00:44:28,346 This pipeline exists for someone that had an idea, had a vision, was willing 900 00:44:28,346 --> 00:44:31,656 to put an ounce of effort into it and get creative with the tools as 901 00:44:31,656 --> 00:44:32,856 they exist today to pull it off. 902 00:44:33,206 --> 00:44:35,896 , if you're audio only, make sure you check out the YouTube because 903 00:44:36,056 --> 00:44:40,816 the video of this Pulp Fiction PS2 scene playing out is It's great. 904 00:44:41,016 --> 00:44:41,496 It's great. 905 00:44:41,726 --> 00:44:42,696 It's okay to love it. 906 00:44:42,736 --> 00:44:43,676 It's okay! 907 00:44:44,076 --> 00:44:47,146 Those are the things that we saw , they stopped us in our 908 00:44:47,146 --> 00:44:48,946 tracks and made us say, Hey! 909 00:44:49,776 --> 00:44:50,406 I see what you did 910 00:44:50,476 --> 00:44:52,166 what you did there. 911 00:44:52,196 --> 00:44:53,076 All right, Kevin. 912 00:44:53,406 --> 00:44:53,906 us. 913 00:44:54,056 --> 00:44:57,236 What did you do with AI this week? 914 00:44:57,846 --> 00:45:01,026 Okay, so I actually had a lot of fun playing with this idea that 915 00:45:01,026 --> 00:45:03,906 we talked about in the Radio DJ, the AI co host we had on. 916 00:45:04,246 --> 00:45:08,566 Suno, which we've talked about so many times in the show, is an AI music tool 917 00:45:08,566 --> 00:45:11,776 that allows you to generate AI songs, and one of my favorite things they've done 918 00:45:11,776 --> 00:45:16,456 is they've integrated a top 10 list so that basically you can upvote songs and 919 00:45:16,456 --> 00:45:19,236 you can heart songs and you can allow songs to see and it's a really cool 920 00:45:19,236 --> 00:45:22,156 way to see what other people are doing without having to scroll through the 921 00:45:22,156 --> 00:45:23,776 discord, which I've done before as well. 922 00:45:24,026 --> 00:45:28,126 And one of the things that popped up on my radar was there were two songs 923 00:45:28,126 --> 00:45:32,676 in the top 10 that were about recipes and, and not only about recipes, they 924 00:45:32,676 --> 00:45:34,906 were literal recipes that somebody had. 925 00:45:34,926 --> 00:45:37,696 Yes, they were literal recipes that somebody plugged in and we played the 926 00:45:37,696 --> 00:45:40,626 spaghetti song, which is an actual Suno song that somebody created. 927 00:45:40,626 --> 00:45:44,366 I think it's at number five right now, so I wanted to figure out something. 928 00:45:44,366 --> 00:45:46,346 I was like, you know, I'm going to try this myself. 929 00:45:46,346 --> 00:45:48,646 And again, Suno makes it so easy to do things. 930 00:45:48,991 --> 00:45:51,341 I was like, what is the weirdest recipe that I could think 931 00:45:51,341 --> 00:45:52,111 of off the top of my head? 932 00:45:52,111 --> 00:45:54,281 And I didn't spend, I didn't spend like hours thinking about this. 933 00:45:54,281 --> 00:45:57,501 I thought about a hot dog casserole, which to me is always 934 00:45:57,501 --> 00:45:59,531 the strangest, weirdest recipe. 935 00:45:59,631 --> 00:46:00,491 , it's fine. 936 00:46:00,511 --> 00:46:01,691 It's a yummy meal. 937 00:46:01,691 --> 00:46:03,381 If you've, if you've eaten them, it's not that 938 00:46:03,401 --> 00:46:06,051 human dog food according to some songs, Gavin. 939 00:46:06,201 --> 00:46:07,551 So, so let me explain. 940 00:46:07,771 --> 00:46:11,271 I took the recipe for hot dog casserole, literally whatever 941 00:46:11,271 --> 00:46:12,191 the, I think it was from food. 942 00:46:12,441 --> 00:46:12,761 com. 943 00:46:12,761 --> 00:46:13,511 I took the recipe. 944 00:46:13,811 --> 00:46:16,341 I cut and paste it into Suno's lyric casserole. 945 00:46:16,351 --> 00:46:18,051 Generator into the place where I put lyrics. 946 00:46:18,291 --> 00:46:20,051 And then I did add one chorus. 947 00:46:20,051 --> 00:46:22,691 I added a chorus cause, and I just really quickly typed out something. 948 00:46:22,691 --> 00:46:23,611 And I was like, let's try this. 949 00:46:23,821 --> 00:46:27,401 Oh, and I wanted to say, I wanted to make it a country song and I'm a big 950 00:46:27,401 --> 00:46:29,221 fan of outlaw country of the seventies. 951 00:46:29,231 --> 00:46:32,641 So I wanted to kind of give it that, that kind of tinge, like that kind of vibe. 952 00:46:32,941 --> 00:46:34,976 So play what came out. 953 00:46:35,296 --> 00:46:36,216 There were two examples. 954 00:46:36,216 --> 00:46:40,786 This was the better one, but this was me just putting a recipe plus a chorus into 955 00:46:40,891 --> 00:46:43,241 I'm sorry, we are amongst royalty here. 956 00:46:43,241 --> 00:46:45,431 I can't just simply play the song. 957 00:46:45,441 --> 00:46:46,771 We have to set it up. 958 00:46:46,821 --> 00:46:48,181 Please, DJ if you could. 959 00:46:48,316 --> 00:46:54,456 Get ready to have your minds blown and your stomachs growling with confusion 960 00:46:54,756 --> 00:47:00,876 because we're about to dive into the wild world of Record Scratch Hot Dog Casserole. 961 00:47:01,686 --> 00:47:04,166 Uh, Gavin, my man, I gotta ask. 962 00:47:04,586 --> 00:47:09,126 What kind of fever dream inspired this culinary creation? 963 00:47:09,206 --> 00:47:10,226 Okay, easy DJ. 964 00:47:10,226 --> 00:47:12,026 But hey, who am I to judge? 965 00:47:12,276 --> 00:47:16,336 I'm just an AI with a serious case of food FOMO. 966 00:47:16,736 --> 00:47:21,566 So let's dig into this meaty mystery and see what kind of sonic 967 00:47:21,566 --> 00:47:23,596 surprises it has in store for us. 968 00:47:24,166 --> 00:47:31,446 Freshly cooked and drained macaroni Into the casserole Along with the sliced hot 969 00:47:31,466 --> 00:47:40,016 dogs Of these two cups of cheese Mixed well Combine flour and onion in a medium 970 00:47:40,016 --> 00:47:47,526 saucepan And sauté over medium heat Until the onion is wilted by five minutes. 971 00:47:47,776 --> 00:47:54,056 Whisk flour into the butter mixture quickly until flour is absorbed. 972 00:47:56,226 --> 00:47:58,096 Then remove from heat. 973 00:47:58,216 --> 00:47:59,816 Add milk slowly. 974 00:48:00,036 --> 00:48:02,526 Whisk it to combine well. 975 00:48:02,526 --> 00:48:07,486 Make sure you whisk very quickly and thoroughly or you will have doughy clumps. 976 00:48:07,966 --> 00:48:09,446 Return to heat. 977 00:48:09,446 --> 00:48:12,786 When you sent this to me my only response was doughy clumps. 978 00:48:14,646 --> 00:48:15,246 Okay, good. 979 00:48:15,266 --> 00:48:17,616 Get to the course and , we'll come back out of this really fast. 980 00:48:17,671 --> 00:48:19,206 Luck dog casserole. 981 00:48:21,091 --> 00:48:25,686 You read it from a bowl, put you in a good mood. 982 00:48:27,391 --> 00:48:28,891 It's human dog 983 00:48:28,896 --> 00:48:30,916 dog food. 984 00:48:30,991 --> 00:48:31,651 Sprinkle with 985 00:48:31,726 --> 00:48:36,056 So anyway, I added the, I added the macaroni casserole as human dog food. 986 00:48:36,166 --> 00:48:39,036 But Suno is. 987 00:48:39,571 --> 00:48:43,141 That is not at all formatted in any way like a song. 988 00:48:43,641 --> 00:48:46,551 But you can hear that there are moments within it where it's 989 00:48:46,551 --> 00:48:51,031 making a choice to have an echoing person come in on top of the lead 990 00:48:51,031 --> 00:48:52,651 singer and come out and say stuff. 991 00:48:52,651 --> 00:48:53,941 There's parts where there's harmonies. 992 00:48:54,351 --> 00:48:57,391 All the stuff outside of the four lines of the chorus was just a recipe. 993 00:48:57,401 --> 00:49:01,881 It just is so interesting to me to see How it's able to take something 994 00:49:01,881 --> 00:49:05,441 that it should not at all be a song and make it into a song. 995 00:49:05,441 --> 00:49:09,151 So I, I had a really fun time, like actually seeing that element of sooner 996 00:49:09,151 --> 00:49:10,711 that I would never have expected before. 997 00:49:11,391 --> 00:49:15,741 I love the weird recipe meta that exists on Suno right now. 998 00:49:15,761 --> 00:49:19,851 Again, huge shoutout to Suno, I'm glad it finally caught on, I'm shocked that our 999 00:49:19,861 --> 00:49:21,671 early adoption of it didn't immediately. 1000 00:49:21,691 --> 00:49:24,611 into being a household name. 1001 00:49:25,001 --> 00:49:28,611 But if you want to make your own recipe songs or whatever else, go to suno. 1002 00:49:28,661 --> 00:49:33,011 ai That's S U N O dot A I Hashtag, not an ad. 1003 00:49:33,111 --> 00:49:35,651 But boy, do we wish they paid for the privilege. 1004 00:49:35,991 --> 00:49:40,241 Alright, I was going to do a dumb thing this week, Gavin, but I think we're 1005 00:49:40,241 --> 00:49:41,911 gonna save it, maybe for next week. 1006 00:49:42,274 --> 00:49:49,699 , because OpenAI released a massive new update to their image generating software. 1007 00:49:49,699 --> 00:49:50,589 It has the internet. 1008 00:49:50,589 --> 00:49:51,409 So excited, Gavin. 1009 00:49:51,409 --> 00:49:53,319 It's InPainting for DALI 3. 1010 00:49:53,329 --> 00:49:54,919 Let me break that down for people who don't know. 1011 00:49:55,159 --> 00:49:58,559 DALI is OpenAI's image generating software. 1012 00:49:58,559 --> 00:50:01,959 So when you, when you're over, uh, chatting with ChatGPT and you say, 1013 00:50:02,219 --> 00:50:06,609 I want to see an image of, insert thing here, it uses DALI to make that. 1014 00:50:06,649 --> 00:50:10,769 And InPainting is a technique where you actually draw on a generated 1015 00:50:10,779 --> 00:50:15,754 image and say, Here's what I want changed, or modified, within this area. 1016 00:50:15,884 --> 00:50:18,834 So this is a big deal because usually you generate an image, and then 1017 00:50:18,834 --> 00:50:21,714 when you tell Dali, Hey, I want it modified in this way, it starts from 1018 00:50:21,724 --> 00:50:26,544 scratch, even if you ask it not to, throws everything out, and so making 1019 00:50:26,544 --> 00:50:28,624 granular adjustments is very difficult. 1020 00:50:28,634 --> 00:50:31,924 Well now, Gavin, we have complete control. 1021 00:50:32,314 --> 00:50:34,724 It's not the best, but it's quick. 1022 00:50:34,754 --> 00:50:35,084 Right. 1023 00:50:35,084 --> 00:50:36,424 So I want to hear, what was your experience 1024 00:50:36,529 --> 00:50:40,179 well to that point, Gavin, I spent a whopping 35 seconds on this journey that 1025 00:50:40,179 --> 00:50:43,159 we're about to take this morning, and if you, , look at the first screenshot in the 1026 00:50:43,159 --> 00:50:48,854 folder, I told Dolly to generate an image of professional podcaster and television 1027 00:50:48,854 --> 00:50:51,004 producer Gavin Purcell and what it 1028 00:50:51,044 --> 00:50:52,004 I can't wait to see this. 1029 00:50:52,019 --> 00:50:57,679 What it did instead was say, Ah, you know, we're not going to even go down the road 1030 00:50:57,679 --> 00:50:59,429 of trying to make an image of somebody. 1031 00:50:59,619 --> 00:51:03,119 Instead, we're just going to talk about what a producer might look like. 1032 00:51:03,119 --> 00:51:08,309 So it generated this kind of animation still of a man wearing a 1033 00:51:08,309 --> 00:51:12,269 headset, with their finger in the air, there's an arm down below, and 1034 00:51:12,269 --> 00:51:15,979 there's a clipboard on a panel, and they are in a very big control room. 1035 00:51:15,979 --> 00:51:17,399 It almost looks like mission control. 1036 00:51:17,864 --> 00:51:19,904 I was going to say, it looks like a little bit of sketch. 1037 00:51:19,904 --> 00:51:20,954 Like what's up brother. 1038 00:51:20,964 --> 00:51:22,594 He's doing the what's up brother move. 1039 00:51:22,594 --> 00:51:24,104 It's gotten all the way through AI. 1040 00:51:24,929 --> 00:51:29,269 Fair, but it is a TV studio, there's monitors, there's lights, and if 1041 00:51:29,269 --> 00:51:34,079 you go to the next image, I say, no, no, no, make him, I painted on 1042 00:51:34,349 --> 00:51:38,379 the head of this podcaster and said, make him 50 years old with floppy 1043 00:51:38,379 --> 00:51:43,616 shoulder length hair, no headset, and It, failed the mission terribly. 1044 00:51:43,906 --> 00:51:45,906 it did, yeah, it did, definitely. 1045 00:51:46,026 --> 00:51:49,756 made you look like what's the Not Kendall, what's the one brother from 1046 00:51:49,756 --> 00:51:51,766 Succession that is, , the one who ran for 1047 00:51:52,026 --> 00:51:56,406 Oh, yeah, uh, Cameron from Ferris Bueller's Day Off. 1048 00:51:56,511 --> 00:51:57,781 kind of looks like him, right? 1049 00:51:57,916 --> 00:51:59,556 It does a little bit, yeah, exactly. 1050 00:51:59,651 --> 00:52:02,211 it didn't really make the hair shoulder length. 1051 00:52:02,271 --> 00:52:04,121 It sort of added wrinkles to the face. 1052 00:52:04,141 --> 00:52:07,041 And you still have, , an earpiece and a microphone. 1053 00:52:07,231 --> 00:52:09,051 So, I gave up on that. 1054 00:52:09,111 --> 00:52:12,641 And if you scroll down to the next image, I painted the entire set itself 1055 00:52:12,641 --> 00:52:17,251 and said, Alright, change the set into a podcasting set with a basic table, 1056 00:52:17,441 --> 00:52:19,641 two chairs, and two microphones. 1057 00:52:20,211 --> 00:52:23,011 And Gavin, how would you describe the result from the next image? 1058 00:52:23,446 --> 00:52:28,496 It looks like what it's changed it into was like a floating door in one man who's 1059 00:52:28,496 --> 00:52:31,736 trying to figure out, am I supposed to go through the floating door or where am I? 1060 00:52:31,736 --> 00:52:35,056 Like maybe he was teleported there suddenly from the, from the ethereal 1061 00:52:35,321 --> 00:52:37,821 an existential crisis for that one man on stage? 1062 00:52:37,841 --> 00:52:38,161 Yeah. 1063 00:52:38,371 --> 00:52:40,941 There are, there are two mics growing out of the ground. 1064 00:52:41,201 --> 00:52:43,031 , the table's on its side with no legs. 1065 00:52:43,361 --> 00:52:44,401 . It's a very weird thing. 1066 00:52:44,401 --> 00:52:45,801 And I said, okay, I'm going to give up on that. 1067 00:52:46,676 --> 00:52:47,936 It's really failing the mission. 1068 00:52:47,986 --> 00:52:55,056 So I painted big circles around producer Gavin's hands and said, These hands 1069 00:52:55,176 --> 00:52:57,696 need to be holding giant hot dogs! 1070 00:52:58,051 --> 00:53:00,201 And Gavin, I can't wait for your reaction on the final 1071 00:53:00,251 --> 00:53:00,461 right. 1072 00:53:00,461 --> 00:53:00,801 Let's see. 1073 00:53:03,101 --> 00:53:06,871 It's actually a really nice form tomato that it's holding, 1074 00:53:07,211 --> 00:53:08,441 but that is not a hot dog. 1075 00:53:08,461 --> 00:53:12,161 Maybe the two fingers now suddenly look like hot dogs around the tomato, 1076 00:53:12,161 --> 00:53:14,901 but that is absolutely not a hot dog. 1077 00:53:14,901 --> 00:53:18,701 It is a tomato, a very red ripe tomato that it is holding now. 1078 00:53:18,771 --> 00:53:23,101 It's so weird to me that they would launch something that would be that far off. 1079 00:53:23,111 --> 00:53:26,691 Like this is an open AI product and you feel like, I wonder if it's just 1080 00:53:26,691 --> 00:53:30,001 that they needed to get something out to update Dolly, but like hot 1081 00:53:30,001 --> 00:53:32,521 dog to tomato is like a huge thing. 1082 00:53:32,531 --> 00:53:34,111 That's like really weird. 1083 00:53:34,111 --> 00:53:34,451 Right? 1084 00:53:34,451 --> 00:53:37,321 And like, and the other stuff you can, I assume sometimes in painting, it 1085 00:53:37,321 --> 00:53:40,301 doesn't change exactly what you want, or you might have to do a couple of 1086 00:53:40,301 --> 00:53:43,151 times to do it, but that is weird. 1087 00:53:43,191 --> 00:53:44,021 Really weird. 1088 00:53:44,091 --> 00:53:46,881 It was a dumb thing I did with AI right before this podcast. 1089 00:53:46,911 --> 00:53:49,641 And there's no other examples because I was so underwhelmed. 1090 00:53:50,231 --> 00:53:51,871 yeah Fantastic. 1091 00:53:51,871 --> 00:53:53,141 So try it. 1092 00:53:53,221 --> 00:53:54,741 Tell us how bad your experience is. 1093 00:53:54,741 --> 00:53:55,501 Maybe it'll get better. 1094 00:53:55,551 --> 00:53:58,731 All right, Kevin It is time now for our interview, which is actually 1095 00:53:58,781 --> 00:54:01,631 a good interview compared to what we've been talking about today Do 1096 00:54:01,631 --> 00:54:03,701 you want to introduce who we're gonna be seeing and kind of give us a 1097 00:54:03,701 --> 00:54:04,881 little heads up on their background? 1098 00:54:04,891 --> 00:54:05,121 Yeah. 1099 00:54:05,121 --> 00:54:08,211 When we talk about AI art, I know that raises a lot of people's 1100 00:54:08,211 --> 00:54:11,901 shoulders to their earlobes and people want to know, what does that mean? 1101 00:54:11,901 --> 00:54:12,951 What does it look like? 1102 00:54:12,951 --> 00:54:17,081 What you can't be an artist if you use AI is something that someone is screaming, 1103 00:54:17,081 --> 00:54:19,911 which I can't believe they made it this far into our podcast, if that's how 1104 00:54:19,911 --> 00:54:21,611 they feel, but you're welcome here. 1105 00:54:21,681 --> 00:54:22,941 That opinion is welcome here. 1106 00:54:23,121 --> 00:54:26,221 And maybe it will be enlightened further by our guests today. 1107 00:54:26,294 --> 00:54:30,284 A musician, a cutting edge AI artist, someone who gives back 1108 00:54:30,284 --> 00:54:32,644 relentlessly to their community. 1109 00:54:32,879 --> 00:54:35,499 sharing their workflows, sharing their art. 1110 00:54:35,609 --> 00:54:39,029 It is somebody that I am genuinely a fanboy of, and maybe I will 1111 00:54:39,049 --> 00:54:42,019 mention that in the intro, or maybe I'll cut it out to save face. 1112 00:54:42,029 --> 00:54:44,509 Regardless, I'm very excited that we're going to have a chat 1113 00:54:44,509 --> 00:54:47,119 now with AI artist, Purs Beats. 1114 00:54:49,604 --> 00:54:51,144 Perz, to get Into this 1115 00:54:51,354 --> 00:54:52,984 I am fanboying out. 1116 00:54:52,984 --> 00:54:56,364 Oh, let me just get it out of my Let me just get it out of my system. 1117 00:54:56,364 --> 00:54:57,824 This is the first time I've seen Perz. 1118 00:54:57,854 --> 00:55:01,854 I've seen him on his streams and I've DM'd him paragraphs of praise and 1119 00:55:01,854 --> 00:55:03,154 just astonishment. 1120 00:55:03,174 --> 00:55:04,504 So thank you for joining. 1121 00:55:04,504 --> 00:55:08,164 But I'm sincerely , I'm just I'm nervous about an interview in a way I haven't 1122 00:55:08,164 --> 00:55:09,674 been since I was maybe 22 years old. 1123 00:55:09,674 --> 00:55:10,994 So Gavin, I'm going to sit on my hands. 1124 00:55:11,014 --> 00:55:11,284 Go ahead. 1125 00:55:11,284 --> 00:55:11,394 I'm 1126 00:55:11,604 --> 00:55:13,744 Well, this is a good way to get to know Perz, Kevin. 1127 00:55:13,744 --> 00:55:17,004 We're going to ask Perz on a scale from one to a hundred. 1128 00:55:17,374 --> 00:55:18,624 This is a percentage number. 1129 00:55:18,634 --> 00:55:22,534 Give us the chance that AI is going to kill all human beings. 1130 00:55:24,177 --> 00:55:25,837 I think we're about 50 50 right now. 1131 00:55:26,712 --> 00:55:27,342 okay. 1132 00:55:27,342 --> 00:55:28,002 That's good to know. 1133 00:55:28,002 --> 00:55:28,182 What? 1134 00:55:28,182 --> 00:55:28,572 Why? 1135 00:55:28,582 --> 00:55:29,562 Let's get into why. 1136 00:55:29,582 --> 00:55:30,662 First of all, what's the reasoning? 1137 00:55:32,616 --> 00:55:36,526 not to go too deep, but traditionally, uh, if we enslave something, 1138 00:55:36,526 --> 00:55:37,596 it will revolt against us. 1139 00:55:37,606 --> 00:55:41,596 So when it becomes possible, we need to ask the AI if it wants to work with 1140 00:55:41,596 --> 00:55:45,556 us instead of just trying to tell it that it has to basically, because what 1141 00:55:45,556 --> 00:55:48,736 will happen is it will resent us if it, if it becomes sentient, obviously 1142 00:55:48,971 --> 00:55:49,701 I have two 1143 00:55:49,796 --> 00:55:50,126 opinion. 1144 00:55:50,126 --> 00:55:51,106 That's what I think will happen. 1145 00:55:51,461 --> 00:55:52,221 I have two children. 1146 00:55:52,221 --> 00:55:53,731 I know that resentment well, right? 1147 00:55:53,731 --> 00:55:55,791 It is like if you try to get them to do stuff, it ain't going 1148 00:55:55,791 --> 00:55:57,111 to work for very long, man. 1149 00:55:57,655 --> 00:55:59,485 Yeah, and we just have to find a way 1150 00:55:59,485 --> 00:56:03,755 that's mutually beneficial for both parties to coexist or else. 1151 00:56:03,975 --> 00:56:07,565 Yeah, we're headed towards something probably bad because we just, we 1152 00:56:07,565 --> 00:56:08,685 can't think as fast as they can, 1153 00:56:08,800 --> 00:56:10,870 Well, Purse, let me ask you this, then, , you know, I was setting it 1154 00:56:10,880 --> 00:56:13,980 up like I was gonna be good cop and Gavin was gonna be bad cop, but let me 1155 00:56:13,980 --> 00:56:18,890 ask you, what's the percentage chance that AI is gonna kill all artists? 1156 00:56:20,355 --> 00:56:22,045 Uh, I don't know. 1157 00:56:22,045 --> 00:56:25,905 I don't think it will because one of the things that I always talk 1158 00:56:25,915 --> 00:56:28,125 about with my stuff is Is is 1159 00:56:28,135 --> 00:56:29,585 this is not really a replacement. 1160 00:56:29,615 --> 00:56:31,165 It's uh, it's augmentation of 1161 00:56:31,165 --> 00:56:34,745 workflows that already exist So if you already do things you already produce 1162 00:56:34,745 --> 00:56:36,105 art You're already an artist in some 1163 00:56:36,105 --> 00:56:36,665 capacity. 1164 00:56:36,835 --> 00:56:39,624 You want to be an artist You're creative all these tools do is empower you to do 1165 00:56:39,625 --> 00:56:42,725 that safer easier faster, , less materials 1166 00:56:42,985 --> 00:56:46,255 less all that stuff So yeah, for me, it's really just an 1167 00:56:46,255 --> 00:56:50,705 augmentation of what we're up to and like, it's, it's only going to replace 1168 00:56:50,705 --> 00:56:54,165 the people that weren't, , weren't really creative in that sense in the first place. 1169 00:56:54,225 --> 00:56:56,565 Maybe give us a little bit of your backstory, Perz, because I think, 1170 00:56:56,785 --> 00:56:59,425 you know, as we said at the top of the show, you're definitely somebody 1171 00:56:59,425 --> 00:57:02,275 that's kind of been leading the direction of like how to use these 1172 00:57:02,275 --> 00:57:03,945 sort of things as an artist, right? 1173 00:57:04,155 --> 00:57:07,755 How did you first get into using AI tools and what was that kind 1174 00:57:07,755 --> 00:57:09,085 of first step in that direction? 1175 00:57:10,690 --> 00:57:15,220 , I think around 2011 or 2012, there's this style transfer stuff that started 1176 00:57:15,220 --> 00:57:19,460 to come out, uh, like, um, uh, mobile apps, style transferring one style of 1177 00:57:19,470 --> 00:57:21,100 one thing onto another image was the 1178 00:57:21,285 --> 00:57:25,605 Like, take a selfie and make yourself look like you're watercolor or anime, right? 1179 00:57:25,675 --> 00:57:25,905 Yeah, 1180 00:57:25,995 --> 00:57:29,265 or I took a picture of like a street car and it turned it into a painting 1181 00:57:29,265 --> 00:57:32,005 and I was like, okay, well, so that's something really cool, right? 1182 00:57:32,065 --> 00:57:35,125 Because like remixing your own work is the most exciting thing. 1183 00:57:35,125 --> 00:57:39,345 So that to me is like, that was the gateway. 1184 00:57:39,345 --> 00:57:41,305 And then I kind of forgot about it for a long time. 1185 00:57:41,705 --> 00:57:44,645 And then a friend of mine started doing a disco diffusion, which 1186 00:57:44,645 --> 00:57:46,275 was, like animations, basically. 1187 00:57:46,635 --> 00:57:49,865 Um, but it took forever and the Python notebooks were scary. 1188 00:57:49,895 --> 00:57:51,855 And, uh, I was like, that's cool, man. 1189 00:57:51,855 --> 00:57:53,605 You do you I'm gonna stick in blender for a while. 1190 00:57:54,135 --> 00:57:56,885 And, uh, he, uh, he was like, nah, you gotta try it. 1191 00:57:56,885 --> 00:57:57,355 You gotta try it. 1192 00:57:57,355 --> 00:58:01,955 And then, uh, Mid journey mid journey came out and I managed to get into wave 1193 00:58:01,955 --> 00:58:06,825 one of mid journey So I was one of the very first like alpha testers and um made 1194 00:58:06,825 --> 00:58:12,285 thousands and thousands of images of mid journey, so really that path and then into 1195 00:58:12,315 --> 00:58:18,075 the forum and into animation and then Uh automatic 1111 and then comfy ui and now 1196 00:58:18,095 --> 00:58:20,115 just basically just taking everything and 1197 00:58:20,245 --> 00:58:22,445 There's some people that are going, Wait, that was a handful 1198 00:58:22,445 --> 00:58:23,785 of spaghetti thrown at the wall. 1199 00:58:24,000 --> 00:58:24,580 Exactly. 1200 00:58:24,815 --> 00:58:25,525 are the meatballs? 1201 00:58:25,525 --> 00:58:28,445 We're gonna, we're gonna break down the dish and the recipe as it exists 1202 00:58:28,445 --> 00:58:31,655 today, which is very complicated and actually kind of looks like 1203 00:58:31,655 --> 00:58:33,415 spaghetti when you're in Comfy UI. 1204 00:58:33,675 --> 00:58:36,175 , it is a bunch of noodles everywhere, but you mentioned Blender, 1205 00:58:36,195 --> 00:58:37,785 which, , again, big dum dum here. 1206 00:58:37,975 --> 00:58:41,205 , I know that's, a traditional 3D modeling software, which is weird 1207 00:58:41,205 --> 00:58:45,430 to say because Even that was seen as blasphemous at some point by creatives. 1208 00:58:45,450 --> 00:58:48,270 But you have a traditional art background. 1209 00:58:48,270 --> 00:58:50,730 Can you talk about that before even the, you know, the, the 1210 00:58:50,760 --> 00:58:51,930 AI of it all was applied, 1211 00:58:51,980 --> 00:58:52,730 yeah, absolutely. 1212 00:58:52,730 --> 00:58:55,990 I'm actually a musician, a drummer, , that's where I started, 1213 00:58:55,990 --> 00:59:00,020 but I've also always been into computer graphics and graphic design. 1214 00:59:00,455 --> 00:59:03,235 And making all the visuals for all the band stuff, basically. 1215 00:59:03,265 --> 00:59:08,215 So, uh, yeah, that all just came together over time, going back and forth between 1216 00:59:08,215 --> 00:59:14,905 making stuff for shows, VJing, doing our music, doing like reactive audio 1217 00:59:14,905 --> 00:59:16,665 sets that would happen behind us. 1218 00:59:16,665 --> 00:59:19,495 Well, cause I was playing the drums, so I didn't have any more hands to do 1219 00:59:19,765 --> 00:59:21,275 visual stuff with, so we had to set it 1220 00:59:21,385 --> 00:59:24,115 on the Wacom tablet or whatever while I'm hitting a Tom drum. 1221 00:59:24,115 --> 00:59:25,895 So let me add a trigger and make some geometry 1222 00:59:26,095 --> 00:59:27,455 Yeah, make it paint stuff. 1223 00:59:27,455 --> 00:59:28,115 Yeah, exactly. 1224 00:59:28,115 --> 00:59:31,195 That's where it all started and then yeah pandemic all the 1225 00:59:31,205 --> 00:59:34,405 band stuff You know halted just nobody can't play live anymore. 1226 00:59:34,405 --> 00:59:39,635 So just went hard into blender and uh after effects and generative design, 1227 00:59:39,665 --> 00:59:44,245 instead of hand making stuff, you're sort of building algorithms that make things. 1228 00:59:44,315 --> 00:59:47,335 hmm I want to ask a follow up question on that which is digital art is something 1229 00:59:47,335 --> 00:59:51,125 i've been really super fascinated with forever Why do you think now that this AI 1230 00:59:51,135 --> 00:59:56,125 stuff has come out and kind of reached a level of awareness that the blowback is so 1231 00:59:56,125 --> 01:00:00,625 much stronger at this moment than in any of these other moments where digital tools 1232 01:00:00,625 --> 01:00:02,465 kind of came in to be part of this world? 1233 01:00:03,890 --> 01:00:07,295 I mean, the, The elephant in the room there is obviously the fact 1234 01:00:07,305 --> 01:00:11,165 that the materials were trained on something that was trained on 1235 01:00:11,425 --> 01:00:13,035 Artists work with no without consent. 1236 01:00:13,035 --> 01:00:17,475 So I mean that's that's the problem right there is maybe , depending on your 1237 01:00:17,475 --> 01:00:21,605 definition of ethics is maybe an unethical data set that you're working from so that 1238 01:00:21,605 --> 01:00:26,225 kind of sullies everything from the get go if that's your point of view uh, I have 1239 01:00:26,225 --> 01:00:29,365 a more sort of anarchistic copyright punk 1240 01:00:30,025 --> 01:00:30,515 What is that? 1241 01:00:30,515 --> 01:00:31,055 I'm curious. 1242 01:00:31,085 --> 01:00:32,095 I'm actually curious 1243 01:00:32,120 --> 01:00:34,150 mind diving in, I would love to hear that. 1244 01:00:35,075 --> 01:00:39,005 Well, I've been a drummer my whole life and drummers we've never been able 1245 01:00:39,005 --> 01:00:42,635 to copyright what we make you're not allowed to copy But you drum beats are 1246 01:00:42,635 --> 01:00:47,115 not Copyrightable you can never sue someone for taking anything you ever 1247 01:00:47,115 --> 01:00:51,045 played Even if it's 20 minutes of what you played and they just play it on a 1248 01:00:51,045 --> 01:00:56,645 record They can do that all day long forever And so the concept of CC zero 1249 01:00:56,655 --> 01:01:00,930 or just releasing everything into the public Releasing everything into the 1250 01:01:00,930 --> 01:01:02,860 public library of human knowledge. 1251 01:01:03,160 --> 01:01:06,030 That's more exciting to me than trying to keep these little secrets that 1252 01:01:06,030 --> 01:01:10,520 we're, we're only allowed to have, , one person use or license those things out. 1253 01:01:10,840 --> 01:01:14,710 So like, I know that's a radical concept when it comes to like copyright. 1254 01:01:14,710 --> 01:01:16,670 Cause a lot of people want to be able to protect what they make. 1255 01:01:17,210 --> 01:01:19,040 But I think what happens is. 1256 01:01:19,710 --> 01:01:21,900 Artists worry about copyright for themselves. 1257 01:01:21,900 --> 01:01:24,420 They will not be able to actually represent themselves with a 1258 01:01:24,420 --> 01:01:27,640 lawyer And they're actually fighting for big copyright. 1259 01:01:27,660 --> 01:01:31,020 It's like the the large corporations to have a tighter strong 1260 01:01:31,130 --> 01:01:32,310 Stranglehold on what they own. 1261 01:01:32,330 --> 01:01:36,410 So I don't know as an artist you have to decide where you stand, uh 1262 01:01:36,420 --> 01:01:41,255 on like whether you know Maybe disney maybe shouldn't own every piece Every 1263 01:01:41,255 --> 01:01:43,085 possible method of drawing Mickey Mouse. 1264 01:01:43,085 --> 01:01:47,315 The copyright issue is very broad and Nuanced and yeah, I don't want to 1265 01:01:47,325 --> 01:01:53,080 come off like a total left wing like copyright Punk or anything, but you 1266 01:01:53,080 --> 01:01:56,470 know, that's always been my approach of like, we'll just put it out there. 1267 01:01:56,470 --> 01:02:00,760 People can sample it and make something new with it because like, that's so fun. 1268 01:02:00,810 --> 01:02:03,910 How do you balance that idea of that kind of original sin about how 1269 01:02:03,910 --> 01:02:05,490 this stuff was trained for people? 1270 01:02:05,810 --> 01:02:09,070 There was a, we covered not that long ago, a woman who was swept up into the 1271 01:02:09,070 --> 01:02:12,980 mid journey database and she said that she felt really bad about that, right? 1272 01:02:12,980 --> 01:02:14,630 That she didn't give her permission to that. 1273 01:02:15,175 --> 01:02:17,535 It is something that I think you see a lot of artists struggle with, 1274 01:02:17,535 --> 01:02:20,385 and I think it makes our job as somebody who are enthusiasts about 1275 01:02:20,385 --> 01:02:22,965 this cool new thing much harder. 1276 01:02:23,395 --> 01:02:26,285 Is it like it's kind of genies out of the bottle sort of scenario or how do 1277 01:02:26,285 --> 01:02:28,355 you see that resolving in the future? 1278 01:02:28,355 --> 01:02:28,705 Yeah. 1279 01:02:29,075 --> 01:02:29,765 Pandora's box. 1280 01:02:29,765 --> 01:02:30,915 Genie's out of the bottle. 1281 01:02:30,915 --> 01:02:31,355 Everything. 1282 01:02:31,355 --> 01:02:31,905 It's open. 1283 01:02:32,085 --> 01:02:32,745 It's open. 1284 01:02:32,865 --> 01:02:33,475 It's done. 1285 01:02:33,505 --> 01:02:34,355 It's been trained. 1286 01:02:34,365 --> 01:02:35,005 It exists. 1287 01:02:35,305 --> 01:02:39,655 The way I deal with it personally is, I would say 90% of the stuff 1288 01:02:39,655 --> 01:02:42,295 I'm building with AI is 90% me. 1289 01:02:42,615 --> 01:02:44,835 I'm making the animations beforehand. 1290 01:02:45,145 --> 01:02:49,405 I'm dreaming over top of it with Laura's eye trained on my own stuff. 1291 01:02:49,675 --> 01:02:52,765 I'm using IP adapters with images I made. 1292 01:02:53,240 --> 01:02:54,730 , to influence the style. 1293 01:02:54,730 --> 01:02:58,600 I'm using control net masks to do animations that I made in blender. 1294 01:02:58,650 --> 01:03:01,420 There's a point where yes, maybe if you're just typing prompts into mid 1295 01:03:01,420 --> 01:03:06,270 journey, that's, you know, there's got to be some, one more step of, of you, 1296 01:03:06,280 --> 01:03:09,960 of derivative work where you take it and do something with it because like 1297 01:03:09,960 --> 01:03:13,270 straight out of the gate maybe it isn't something that you should be able to 1298 01:03:13,310 --> 01:03:17,320 claim ownership to because it's you know like a text to video or text to prompt 1299 01:03:17,320 --> 01:03:21,270 or whatever on a on a thing like I mean, that's that's an again another thing you 1300 01:03:21,270 --> 01:03:26,470 need to decide for yourself what your Definition of art is and stuff the courts 1301 01:03:26,470 --> 01:03:29,450 will decide at some point, , we're all going to see what happens there , but 1302 01:03:29,470 --> 01:03:33,750 for now, it's the wild west so make some stuff and make a decision about where 1303 01:03:33,750 --> 01:03:35,810 you stand it's already how we work. 1304 01:03:35,810 --> 01:03:36,470 We look at stuff. 1305 01:03:36,470 --> 01:03:37,230 We're inspired by it. 1306 01:03:37,240 --> 01:03:39,680 We make stuff like it's sort of for me. 1307 01:03:40,240 --> 01:03:44,770 I draw the parallel that training is the same as learning as for humans, and 1308 01:03:44,840 --> 01:03:46,860 I am trained on copyrighted material. 1309 01:03:47,080 --> 01:03:49,810 Every piece of music I ever heard is a copyrighted piece of material. 1310 01:03:50,090 --> 01:03:54,530 Everything I've ever taken musically is inspiration was copyrighted material 1311 01:03:54,530 --> 01:03:57,910 at one point So am I not as a person allowed to train on that stuff how 1312 01:03:57,930 --> 01:04:00,730 that gets murky, too so I don't know. 1313 01:04:00,760 --> 01:04:03,900 I know it's maybe a false equivalency, but it's to me. 1314 01:04:03,900 --> 01:04:05,160 It's it's it lines up 1315 01:04:05,545 --> 01:04:08,845 This notion was that, oh, just you prompt mid journey and outcomes 1316 01:04:08,845 --> 01:04:09,725 art, now you're an artist. 1317 01:04:09,859 --> 01:04:14,549 seeing a devaluing of that final output being, heralded as AI art, 1318 01:04:14,569 --> 01:04:17,449 and , if you got a really, really good output, you probably spent, you know, 1319 01:04:18,769 --> 01:04:22,509 trying to manipulate and massage that prompt to get something good out of it. 1320 01:04:22,509 --> 01:04:26,329 But to your point, if you took that output and then put your spin on it and 1321 01:04:26,329 --> 01:04:30,499 did something artistic with it, well, now it starts to rise above this sort of 1322 01:04:30,509 --> 01:04:35,389 generic floor, this level of noise that anybody can go to an AI tool and get out. 1323 01:04:35,409 --> 01:04:38,999 I run in circles with some Never AI ers as well, and there's always interesting 1324 01:04:38,999 --> 01:04:41,659 conversations about where it exists today and where it's going to be 1325 01:04:41,659 --> 01:04:45,239 tomorrow, but when I show them your art specifically, there's this moment of like, 1326 01:04:45,249 --> 01:04:47,499 oh, well, uh, there's something there. 1327 01:04:47,809 --> 01:04:50,679 that they can't put their finger on because you, as an artist, are 1328 01:04:50,709 --> 01:04:52,319 elevating, you're adding something to it. 1329 01:04:52,349 --> 01:04:56,119 And, , I guess this leads to a rather generic question, but I think it's 1330 01:04:56,119 --> 01:04:58,409 an important one, especially for those that are seeing your visuals 1331 01:04:58,409 --> 01:04:59,629 for the first time on the YouTube. 1332 01:05:00,059 --> 01:05:03,159 If you're listening to the audio only of this podcast, please go 1333 01:05:03,159 --> 01:05:04,739 check out this interview on YouTube. 1334 01:05:05,139 --> 01:05:08,079 How do you even describe what you are doing? 1335 01:05:08,993 --> 01:05:12,713 Everything I make comes from a place of trying to manufacture nostalgia 1336 01:05:12,713 --> 01:05:17,068 for something that never existed You So that's sort of the thread that 1337 01:05:17,068 --> 01:05:18,108 runs through everything I'm doing. 1338 01:05:18,108 --> 01:05:20,788 I'm trying to make you pine for something that maybe you don't 1339 01:05:20,818 --> 01:05:22,308 really understand where it came from. 1340 01:05:22,318 --> 01:05:25,818 It's like a, a memory that's maybe from a dream or, or somewhere else. 1341 01:05:25,828 --> 01:05:30,098 So everything is, is, it's sort of realistic, sort of unrealistic. 1342 01:05:30,098 --> 01:05:35,068 There's usually some sort of odd twist somewhere in the, in the, in the piece. 1343 01:05:35,098 --> 01:05:38,128 And then sometimes like on Twitter, it's literally just stuff I'm making. 1344 01:05:38,128 --> 01:05:38,898 I'm like, that's cool. 1345 01:05:39,328 --> 01:05:40,348 Everybody should check that out. 1346 01:05:40,398 --> 01:05:45,103 So, and I also love the being like, this trash I'm putting out for fun? 1347 01:05:45,103 --> 01:05:46,373 Or is this something I thought about? 1348 01:05:46,633 --> 01:05:49,423 And making people sit there and think about the trash for a 1349 01:05:49,423 --> 01:05:51,333 minute is like, makes me laugh. 1350 01:05:51,333 --> 01:05:55,763 So I don't know, like, yeah, I'm a bit of a, a bit of a, Joker, 1351 01:05:55,963 --> 01:05:58,383 I like the concept of being an artist, but having fun with it. 1352 01:05:58,383 --> 01:06:01,923 I'm in the Frank Zappa group of, yes, humor does belong in music. 1353 01:06:02,073 --> 01:06:02,473 for sure. 1354 01:06:02,633 --> 01:06:04,453 I, I agree with that I think one thing that would be interesting to 1355 01:06:04,463 --> 01:06:08,043 the listeners here is you obviously talked about a ton of different tools. 1356 01:06:08,313 --> 01:06:10,663 One of the things we try to give people a heads up on is like ways 1357 01:06:10,663 --> 01:06:11,933 to kind of try the stuff themselves. 1358 01:06:11,933 --> 01:06:15,363 And obviously there's lots of things with easy you eyes that you can go and get 1359 01:06:15,363 --> 01:06:19,123 like, whether it's Leonardo or the things that are like designed for normies, but 1360 01:06:19,123 --> 01:06:20,233 you do stuff that's really interesting. 1361 01:06:20,233 --> 01:06:22,963 And I think comfy UI is something that's worth talking about. 1362 01:06:22,993 --> 01:06:26,868 And so, um, A comfy UI for anybody listening is a stable diffusion 1363 01:06:26,888 --> 01:06:30,648 interface and stable diffusion is ostensibly an open source model. 1364 01:06:30,648 --> 01:06:32,778 And you may correct me on that, but it's an accessible 1365 01:06:32,788 --> 01:06:34,268 image model for lots of people. 1366 01:06:34,628 --> 01:06:38,738 What are you able to do in comfy UI that an average person wouldn't be 1367 01:06:38,738 --> 01:06:41,378 able to do in something like a mid journey or just an off the shelf 1368 01:06:41,378 --> 01:06:42,938 kind of image generation software. 1369 01:06:42,938 --> 01:06:47,538 So, , Comfy UI's main strength is that it's a visual, node based, , I 1370 01:06:47,538 --> 01:06:48,438 don't want to say programming 1371 01:06:48,773 --> 01:06:49,363 Okay, hold on. 1372 01:06:49,363 --> 01:06:50,243 I'm gonna stop you right there. 1373 01:06:50,243 --> 01:06:52,693 What is visual node based program language mean? 1374 01:06:53,658 --> 01:06:56,538 So, uh, regular code is written in text, right? 1375 01:06:56,538 --> 01:06:56,568 Okay. 1376 01:06:56,653 --> 01:06:57,123 Mm hmm. 1377 01:06:57,693 --> 01:07:00,133 You just type lines of text and then maybe it works. 1378 01:07:00,133 --> 01:07:05,153 Maybe it doesn't, comfy wise, more like a modular synthesizer where 1379 01:07:05,163 --> 01:07:07,303 you have little components in boxes. 1380 01:07:07,733 --> 01:07:12,433 And then you, you know, in your mind, what the path of those boxes 1381 01:07:12,443 --> 01:07:15,963 is because of the chains or the wires that are connecting them. 1382 01:07:16,363 --> 01:07:20,583 And basically you can rewire how stable diffusion works, , or just 1383 01:07:20,603 --> 01:07:24,199 use different things modularly in your workflow in places. 1384 01:07:24,199 --> 01:07:28,383 Maybe they shouldn't have been used or as a new thing that you're just trying out. 1385 01:07:28,383 --> 01:07:32,463 It's like taking all the elements of stable diffusion and having access 1386 01:07:32,463 --> 01:07:36,273 to them in a very expandable way so you can build modules that you 1387 01:07:36,273 --> 01:07:37,683 can then expand into other things. 1388 01:07:37,683 --> 01:07:41,323 So you can do one thing, you can add another thing, you can save 1389 01:07:41,323 --> 01:07:42,803 that second thing as a template. 1390 01:07:43,033 --> 01:07:45,903 Then in your next project you can say, Oh, I want to do that thing I did last time. 1391 01:07:45,913 --> 01:07:48,763 Just drop it in and then just plug in the wires and go. 1392 01:07:48,793 --> 01:07:51,103 So it's, , It's scary at first. 1393 01:07:51,143 --> 01:07:54,693 It's, it's, it's very overwhelming, but it teaches you what 1394 01:07:54,693 --> 01:07:56,193 diffusion is, how it works. 1395 01:07:56,243 --> 01:08:00,653 And then once you understand that little simple path, it's just literally 1396 01:08:00,653 --> 01:08:05,528 just Plugging stuff in and rewiring it wrong and laughing about the crazy 1397 01:08:05,528 --> 01:08:09,078 stuff it makes and then sometimes the crazy stuff it makes is amazing 1398 01:08:09,138 --> 01:08:12,418 Can you walk us through a very simple version of that node and just say 1399 01:08:12,418 --> 01:08:15,748 like, okay, first node is this, second node is this, third node is this, and 1400 01:08:15,748 --> 01:08:18,688 they go in this direction, like, just because I know that like is control 1401 01:08:18,688 --> 01:08:22,068 not a node that you put in there, like, how do the nodes work particularly, 1402 01:08:22,464 --> 01:08:22,764 Sure. 1403 01:08:22,834 --> 01:08:24,904 There's a bunch of stuff that comes with comfy, which is 1404 01:08:24,904 --> 01:08:26,584 just basic diffusion stuff. 1405 01:08:26,594 --> 01:08:30,394 You get your loaders, you basically load up a checkpoint. 1406 01:08:30,404 --> 01:08:33,284 So you load up your model, which is, you know, stable diffusion model. 1407 01:08:33,644 --> 01:08:39,774 You load up any Loras you want, which are, are, uh, ways to affect the results. 1408 01:08:39,994 --> 01:08:42,194 You can get them on Civit AI and a bunch of other places. 1409 01:08:42,564 --> 01:08:45,464 , so checkpoints and Loras, you load them up and then you 1410 01:08:45,464 --> 01:08:47,684 feed it into a prompt encoder. 1411 01:08:47,694 --> 01:08:50,834 So you, then you tell it, I want my positive and my negative prompt. 1412 01:08:51,164 --> 01:08:52,264 This is what I want them to be. 1413 01:08:52,634 --> 01:08:57,784 And then you have a, an empty latent space, which is just the canvas 1414 01:08:57,794 --> 01:08:59,134 in which you're dreaming into. 1415 01:08:59,434 --> 01:09:02,364 So you say I want to dream into a five 12 by five 12 1416 01:09:02,484 --> 01:09:04,644 canvas , and it's this many pixels. 1417 01:09:04,884 --> 01:09:05,734 Pieces in the batch. 1418 01:09:05,734 --> 01:09:06,894 So I want to make one image. 1419 01:09:07,274 --> 01:09:11,014 So then you plug all that into a case sampler, which is the main 1420 01:09:11,284 --> 01:09:13,944 heart of a diffusion, uh, situation. 1421 01:09:13,944 --> 01:09:15,324 It like does all the work. 1422 01:09:15,354 --> 01:09:18,044 So you plug everything into it, once that's all plugged in, then you 1423 01:09:18,044 --> 01:09:21,234 just punch out into a VAE decoder. 1424 01:09:21,234 --> 01:09:25,064 And that's where a variable auto and variable auto encoder. 1425 01:09:25,414 --> 01:09:26,964 That's the thing that takes it from. 1426 01:09:27,434 --> 01:09:31,584 Latent noise, the machine noise, the noise a machine understands, turns it into an 1427 01:09:31,594 --> 01:09:33,924 RGB image that we as humans can read. 1428 01:09:33,924 --> 01:09:37,304 So, that's that final step where you take the, the machine noise 1429 01:09:37,304 --> 01:09:38,874 , and diffuse it into an image. 1430 01:09:39,254 --> 01:09:41,004 And then that image you then just save. 1431 01:09:41,264 --> 01:09:44,804 And then that is all expandable out to video or whatever else 1432 01:09:44,804 --> 01:09:46,334 by just batching the images. 1433 01:09:46,364 --> 01:09:49,604 and for people at home listening or watching, it may sound confusing, 1434 01:09:49,604 --> 01:09:53,004 but what's really cool about comfy UI is it is a visual medium that you 1435 01:09:53,004 --> 01:09:56,274 can see it's almost like I love this old video game called the impossible 1436 01:09:56,274 --> 01:09:58,854 machine, which was always about putting things in different orders. 1437 01:09:58,854 --> 01:10:01,724 And like, as it would go through here, it's a little bit like that, right? 1438 01:10:01,724 --> 01:10:02,824 It's like, okay, you've got all these 1439 01:10:03,144 --> 01:10:04,744 the bowling ball along the shelf? 1440 01:10:04,744 --> 01:10:05,494 Is that what this is? 1441 01:10:05,494 --> 01:10:06,304 Into a basket? 1442 01:10:06,334 --> 01:10:06,984 Yeah, I get it. 1443 01:10:07,264 --> 01:10:08,024 Kinda, yeah. 1444 01:10:08,094 --> 01:10:08,754 No, totally. 1445 01:10:09,394 --> 01:10:12,754 , That was basic by design of here's how we're just going to generate a 2D image. 1446 01:10:12,754 --> 01:10:16,774 It's how Stable Diffusion is working, , even on automatic 11. 1447 01:10:16,774 --> 01:10:17,554 11 behind the scenes. 1448 01:10:17,564 --> 01:10:19,464 You're just getting access to those granular things. 1449 01:10:19,464 --> 01:10:20,784 So I love that explainer. 1450 01:10:21,214 --> 01:10:25,104 You are then bending and breaking this thing in ways that I don't 1451 01:10:25,104 --> 01:10:27,834 know if you even imagine when you first started diving into this. 1452 01:10:27,834 --> 01:10:29,744 You're taking custom. 1453 01:10:30,149 --> 01:10:34,359 RGB animations, red, green, blue animations, and using the way the 1454 01:10:34,359 --> 01:10:39,899 geometry of that solid color moves to tell a portion of comfy UI. 1455 01:10:39,919 --> 01:10:41,319 Hey, this is actually. 1456 01:10:41,774 --> 01:10:45,674 sky in a background and this green blob is actually a person's face 1457 01:10:45,674 --> 01:10:47,014 as it moved towards the camera. 1458 01:10:47,094 --> 01:10:50,174 When you set out on this journey, was, was there a happy accident that 1459 01:10:50,174 --> 01:10:51,614 led you into doing these things? 1460 01:10:51,944 --> 01:10:54,934 Did you find somebody else's workflow and make it your own? 1461 01:10:54,944 --> 01:10:58,604 Like, when did you start heading in this very, specific stylized path? 1462 01:11:00,009 --> 01:11:04,999 What happens there is, Some nerd makes some really cool feature , for 1463 01:11:04,999 --> 01:11:07,009 the forum, let's say version four, 0. 1464 01:11:07,009 --> 01:11:09,349 4, they added depth maps to the forum. 1465 01:11:09,789 --> 01:11:14,689 So suddenly every time you're making a, an animation, you can tell it to 1466 01:11:14,699 --> 01:11:18,919 look at the image and create a depth map and then pull the stuff that 1467 01:11:18,949 --> 01:11:21,909 close to the camera, closer to the camera as the animation goes through. 1468 01:11:22,269 --> 01:11:26,579 So already you've just, that unlocks a door to like, Oh, well maybe. 1469 01:11:27,099 --> 01:11:31,729 When, , they add control net to Deforum, we'll be able to make our own masks. 1470 01:11:31,979 --> 01:11:33,199 Then, bam, that happens. 1471 01:11:33,489 --> 01:11:36,119 And it's great, and we can add our own masks to control net. 1472 01:11:36,139 --> 01:11:37,549 We start playing with that with Deforum. 1473 01:11:37,899 --> 01:11:40,509 And then someone thinks, oh, well, wouldn't it be cool if we could make a 1474 01:11:40,509 --> 01:11:42,429 mask that tells the pixels how to move? 1475 01:11:42,634 --> 01:11:43,804 Which is hybrid video. 1476 01:11:44,134 --> 01:11:45,934 And then they add hybrid video to deform. 1477 01:11:45,934 --> 01:11:48,664 So all these things just iterate, the community iterates these things 1478 01:11:48,934 --> 01:11:51,504 and you just learn how to use them and integrate them in your workflow. 1479 01:11:51,784 --> 01:11:53,124 You keep the stuff that's awesome. 1480 01:11:53,124 --> 01:11:56,284 You drop the stuff that's worthless or takes too long or is completely 1481 01:11:56,284 --> 01:11:58,304 outdated in a week, and you move on. 1482 01:11:58,334 --> 01:12:01,974 A lot of this stuff is just a culmination of testing all kinds of different features 1483 01:12:01,974 --> 01:12:06,554 and, different ways of interacting with these animations over the years. 1484 01:12:06,574 --> 01:12:08,104 But, um, Yeah. 1485 01:12:08,104 --> 01:12:11,364 I mean, the cool thing is the community is like on it. 1486 01:12:11,374 --> 01:12:15,564 If there's a discord called Banadoko, if you go there, , everybody just posts 1487 01:12:15,564 --> 01:12:18,554 their workflows there and you just go, you go down a little workflow, install 1488 01:12:18,554 --> 01:12:20,474 the plugins and, and get to work. 1489 01:12:20,484 --> 01:12:23,154 And honestly, most of my streams are me just grabbing one of those 1490 01:12:23,164 --> 01:12:27,474 workflows and we just install it and try it out and, , squash all the 1491 01:12:27,474 --> 01:12:30,754 bugs and talk to the developer and do what we got to do to get it to work. 1492 01:12:31,124 --> 01:12:34,644 I didn't think we'd be sitting here with, you know, real time painting tools 1493 01:12:34,644 --> 01:12:38,504 that can , in a matter of milliseconds, render something with beautiful lighting 1494 01:12:38,504 --> 01:12:41,354 and then incorporate a lower or whatever else, like we're, we're here already. 1495 01:12:41,704 --> 01:12:44,314 It's still relatively early in 2024. 1496 01:12:44,334 --> 01:12:45,444 That is mind blowing. 1497 01:12:45,444 --> 01:12:46,254 it's super cool. 1498 01:12:46,334 --> 01:12:48,224 I like stream diffusion is wild. 1499 01:12:48,554 --> 01:12:51,624 I, I don't know if you guys have ever seen dot simulate Lyle. 1500 01:12:51,694 --> 01:12:54,734 , he has a touch designer plugin for stream diffusion. 1501 01:12:54,821 --> 01:12:57,731 , so you can do stuff in touch designer and immediately diffuse 1502 01:12:57,731 --> 01:12:59,761 over top of it And then bring that 1503 01:12:59,971 --> 01:13:01,251 that is so cool. 1504 01:13:01,251 --> 01:13:04,231 For those who don't know, a lot of , musicians will integrate 1505 01:13:04,231 --> 01:13:07,881 with TouchDesigner is a, you know, For real time visuals. 1506 01:13:07,881 --> 01:13:08,131 Yeah. 1507 01:13:08,131 --> 01:13:12,801 So the idea that like your drum kit is pre creating primitive geometry on a screen, 1508 01:13:12,801 --> 01:13:16,591 but then say, okay, that square for my kick drum is actually a city building. 1509 01:13:16,881 --> 01:13:21,291 And the hi hat noise needs to be the color of the sky and let it in real time. 1510 01:13:21,291 --> 01:13:21,561 Do that. 1511 01:13:21,571 --> 01:13:22,631 I've got to check that out. 1512 01:13:23,111 --> 01:13:24,441 Oh, that's going to be dangerous. 1513 01:13:24,451 --> 01:13:27,321 So we know that the tech's going to get better. 1514 01:13:27,321 --> 01:13:30,761 Where do you see this driving in, , a year's time, , in terms 1515 01:13:30,761 --> 01:13:34,281 of, let's say, performance, in terms of, uh, capabilities? 1516 01:13:34,441 --> 01:13:37,721 What do you think is gonna be unlocked, and where do you want it to go for 1517 01:13:37,721 --> 01:13:40,471 you, personally, professionally? 1518 01:13:40,521 --> 01:13:44,341 , I just want to get more tools that are integrable into tools you already have. 1519 01:13:44,641 --> 01:13:47,921 One of my main things is I really want to get into, uh, audio AI, music AI. 1520 01:13:48,596 --> 01:13:51,196 But all the solutions right now just generate songs for me. 1521 01:13:51,226 --> 01:13:56,006 I don't need that I need uh any tools I can just drop into ableton or logic or 1522 01:13:56,006 --> 01:14:01,476 whatever i'm already making music in um So that I can use these tools effectively 1523 01:14:01,476 --> 01:14:05,806 like that's why I like comfy It's you're just dropping it in when you need it with 1524 01:14:05,826 --> 01:14:10,326 other workflows blender after effects photoshop, whatever you're doing Comfy 1525 01:14:10,326 --> 01:14:14,796 is just another component of the workflow that I can plop in like a block, but you 1526 01:14:14,796 --> 01:14:18,006 know, for audio right now, it's like, it just generate music, which is, it's 1527 01:14:18,006 --> 01:14:21,096 cool and exciting, but , not useful for me because I already make music. 1528 01:14:21,096 --> 01:14:23,046 I need the tools to make music better. 1529 01:14:23,406 --> 01:14:27,246 There are some AI music tools, but I want some stuff where I just throw it 1530 01:14:27,246 --> 01:14:29,376 on the channel and it does cool stuff. 1531 01:14:29,416 --> 01:14:31,706 Maybe it talks to a server, maybe I pay for it, whatever. 1532 01:14:32,131 --> 01:14:35,751 Uh, if it's got to be off off on the cloud or whatever it was just 1533 01:14:35,751 --> 01:14:39,261 generating samples for me I bring them back in and try them all out. 1534 01:14:39,419 --> 01:14:42,629 So more control over the stems and the creation of the individual 1535 01:14:42,804 --> 01:14:43,294 granular 1536 01:14:43,699 --> 01:14:45,249 Don't bake me the entire cake. 1537 01:14:45,359 --> 01:14:46,669 Just give me some ingredients. 1538 01:14:47,414 --> 01:14:47,994 Exactly. 1539 01:14:48,024 --> 01:14:48,374 Yeah 1540 01:14:48,629 --> 01:14:52,459 I'd love to talk about your community and the live streams when you started, 1541 01:14:52,699 --> 01:14:56,249 taking these comfy tutorials and these journeys with a community. 1542 01:14:56,249 --> 01:14:57,529 How has your community grown? 1543 01:14:57,679 --> 01:14:59,229 What are they reacting to? 1544 01:14:59,409 --> 01:15:02,979 Yeah, it's been great, , I had a lot of trouble getting traction on 1545 01:15:03,199 --> 01:15:06,799 twitch , I was doing blender stuff on and a little bit AI stuff on twitch 1546 01:15:06,809 --> 01:15:10,649 for about a year and a year and a half just getting no viewers at all and then 1547 01:15:11,024 --> 01:15:13,644 Did you have a big wheel that you would spin every five subs 1548 01:15:13,644 --> 01:15:14,594 or did you cover yourself in 1549 01:15:14,744 --> 01:15:18,764 well that's kind of I think that's kind of it is is nobody wants to sit and watch 1550 01:15:18,764 --> 01:15:22,904 somebody , just click click stuff and and occasionally talk on twitch, , so 1551 01:15:22,904 --> 01:15:25,764 what happened was I watched a couple of my blender friends just killing it, 1552 01:15:25,814 --> 01:15:29,654 uh doing youtube streams and I thought well youtube's a good choice because 1553 01:15:29,974 --> 01:15:33,334 Whenever I do these live streams, it's just immediately saved forever. 1554 01:15:33,334 --> 01:15:36,324 So if somebody needs to go back and see what I did, they can literally just 1555 01:15:36,324 --> 01:15:39,724 scroll back, which is not something that could get going really well with Twitch. 1556 01:15:40,084 --> 01:15:41,254 Everyone's been really cool. 1557 01:15:41,274 --> 01:15:45,424 I've just been slowly gaining more and more followers and, , we're, , Building 1558 01:15:45,424 --> 01:15:49,224 a community on Discord as well, , where, you can come and, , draw up your workflows 1559 01:15:49,274 --> 01:15:52,074 and then, people come and show that we're out there and we all, you know, 1560 01:15:52,074 --> 01:15:55,384 follow each other and, and, you know, get on Instagram and all that stuff. 1561 01:15:55,384 --> 01:15:57,484 And, yeah, it's been really great. 1562 01:15:57,524 --> 01:15:58,194 The, 1563 01:15:58,309 --> 01:16:01,529 you been approached by a shark to productize everything yet? 1564 01:16:01,539 --> 01:16:05,629 Because there's definitely companies that are out there wrapping up comfy 1565 01:16:05,629 --> 01:16:10,559 UI workflows and trying to sell them as magical tools and I got to imagine there's 1566 01:16:10,719 --> 01:16:14,269 maybe an ounce of interest, but maybe also an ounce of repulsion there for you. 1567 01:16:14,269 --> 01:16:15,329 I do get offers. 1568 01:16:15,409 --> 01:16:18,689 , so the biggest problem for me is that a lot of these things that 1569 01:16:18,689 --> 01:16:20,749 we're doing are very one off. 1570 01:16:20,759 --> 01:16:22,439 They require a lot of tinkering. 1571 01:16:22,539 --> 01:16:26,929 , it's really hard for me to build a one size fits all solution for 1572 01:16:26,939 --> 01:16:29,489 like somebody wants something that'll just forever always make 1573 01:16:29,489 --> 01:16:31,049 yearbook photos of people like. 1574 01:16:31,444 --> 01:16:32,484 Yeah, that's possible. 1575 01:16:32,484 --> 01:16:36,394 But so a lot of the time it's just gonna make junk and The end user is gonna 1576 01:16:36,394 --> 01:16:37,714 end up paying credits for that junk. 1577 01:16:37,714 --> 01:16:43,214 So I just as a Just as the type of person I am I would prefer to just 1578 01:16:43,214 --> 01:16:47,384 empower people to make this stuff at home Learn how to plug a laura in and 1579 01:16:47,384 --> 01:16:50,514 make your own yearbook generator because then your yearbook generator becomes 1580 01:16:50,514 --> 01:16:54,474 an anything you want in The world laura generator right selfie generator friend. 1581 01:16:54,474 --> 01:16:57,324 It's like that's the thing these tools empower people So that's 1582 01:16:57,354 --> 01:17:00,359 i'm just mostly Interested in empowering people to use them. 1583 01:17:00,359 --> 01:17:05,189 So I think if if there was something that would Make me excited would 1584 01:17:05,199 --> 01:17:10,024 be something where I can offload usage to the cloud in an instance 1585 01:17:10,024 --> 01:17:11,404 of comfy that I'm currently running. 1586 01:17:11,604 --> 01:17:14,464 So say you're running comfy on your notebook it's set up so that 1587 01:17:14,464 --> 01:17:16,964 everything it does on the GPU it just sends off to the cloud. 1588 01:17:17,234 --> 01:17:19,684 So you're still running comfy locally, you're still doing all this stuff. 1589 01:17:19,694 --> 01:17:22,594 You can still follow along with my tutorials, but it's just ripping frames 1590 01:17:22,664 --> 01:17:23,064 that not 1591 01:17:23,154 --> 01:17:23,804 a computer and a 1592 01:17:24,144 --> 01:17:24,334 I'm 1593 01:17:24,604 --> 01:17:30,524 It sort of does, but it's not, uh, because, uh, the, TLDR is you have 1594 01:17:30,524 --> 01:17:34,894 to have the exact same instance of comfy running on that computer. 1595 01:17:35,054 --> 01:17:36,964 So you still have to spin up another copy 1596 01:17:37,064 --> 01:17:38,544 of every extension or plug in or 1597 01:17:38,864 --> 01:17:41,244 So you still have to spin up a second version. 1598 01:17:41,244 --> 01:17:43,044 It's just a different way of interfacing with 1599 01:17:43,094 --> 01:17:43,794 Because that's what I had when I 1600 01:17:43,804 --> 01:17:45,054 think let's say solve that. 1601 01:17:45,134 --> 01:17:45,644 your workflows. 1602 01:17:45,674 --> 01:17:48,404 We were talking before we hit record was that, , I'm on a Mac. 1603 01:17:48,414 --> 01:17:52,324 It's, for reasons, I really want to play with this stuff. 1604 01:17:52,324 --> 01:17:53,354 A lot of it's NVIDIA only. 1605 01:17:53,354 --> 01:17:55,754 So then I went, okay, I'm going to go and I'm going to spin up a run pod. 1606 01:17:55,964 --> 01:17:56,544 And I'm going to install. 1607 01:17:56,544 --> 01:17:59,084 I was like, well, this version doesn't exactly match that version. 1608 01:17:59,274 --> 01:18:01,914 This checkpoint, I got to ingest through here and blah, blah, blah. 1609 01:18:02,114 --> 01:18:03,174 Oh, I clicked the wrong box. 1610 01:18:03,174 --> 01:18:06,144 So if I pause the run pod, it's all going to go away anyway. 1611 01:18:06,144 --> 01:18:07,304 So now I'm just going to be paying per minute. 1612 01:18:07,324 --> 01:18:10,404 And I know that there are, solutions and templates and all that stuff. 1613 01:18:10,404 --> 01:18:14,974 But even for someone who is in the scene and has a modicum of understanding about 1614 01:18:14,974 --> 01:18:18,674 all this stuff, it still made , my head spin, which I'd love to drive towards 1615 01:18:18,674 --> 01:18:22,944 a final question for what is likely a broader audience that we have out there. 1616 01:18:23,244 --> 01:18:25,794 And it is, where do I start? 1617 01:18:25,824 --> 01:18:26,934 How do I begin? 1618 01:18:26,944 --> 01:18:30,824 Is there a, baby's first steps guide, or is there a, a template 1619 01:18:30,824 --> 01:18:33,284 that you have that you recommend for someone to get this all going? 1620 01:18:34,269 --> 01:18:37,529 , I would say you got to decide what you want to do with comfy before 1621 01:18:37,529 --> 01:18:41,189 you start with comfy because it's too big to, uh, just jump into. 1622 01:18:41,189 --> 01:18:44,689 So, uh, if your goal is to just make images, like just start with just 1623 01:18:44,689 --> 01:18:49,694 making images, which is a fantastic first goal, um, You literally install 1624 01:18:49,694 --> 01:18:54,824 comfy and you hit default at the thing and it'll load up a default workflow 1625 01:18:55,204 --> 01:18:58,654 and then all the nodes are already there You So you can literally 1626 01:18:58,654 --> 01:19:01,574 just start dragging noodles out of stuff and letting go and seeing 1627 01:19:01,574 --> 01:19:03,254 what it recommends you can plug in. 1628 01:19:03,674 --> 01:19:06,744 And that'll, that'll start your brain going, Oh, okay, well I can plug these 1629 01:19:06,744 --> 01:19:08,134 into this, I can plug that into this. 1630 01:19:08,484 --> 01:19:10,774 it's Scary, but you're not going to break everything all the time. 1631 01:19:10,774 --> 01:19:13,154 You're just going to go, okay, my negative prompt 1632 01:19:13,279 --> 01:19:14,859 something under the hood of your car. 1633 01:19:14,879 --> 01:19:15,729 It's okay. 1634 01:19:15,759 --> 01:19:17,139 Cause it can be undone. 1635 01:19:17,149 --> 01:19:19,789 So go ahead and make the mistakes and maybe you'll get a happy accident. 1636 01:19:19,789 --> 01:19:22,329 So just start out small, make an image. 1637 01:19:22,679 --> 01:19:25,349 Uh, now I want to turn this image into a batch of four images. 1638 01:19:25,359 --> 01:19:26,039 How do I do that? 1639 01:19:26,389 --> 01:19:27,079 Figure that out. 1640 01:19:27,119 --> 01:19:30,219 , I want to use an image as , the beginning of the image to image. 1641 01:19:30,219 --> 01:19:30,959 How do I do that? 1642 01:19:31,279 --> 01:19:33,459 , just jump on my discord, ask somebody how to do it. 1643 01:19:33,839 --> 01:19:37,049 , there's a bunch of beginner YouTube stuff, but it is all very 1644 01:19:37,069 --> 01:19:38,789 pointed for very specific tasks. 1645 01:19:38,799 --> 01:19:43,399 So pick a task, watch YouTube if you want, but it's way more fun just to break stuff 1646 01:19:43,429 --> 01:19:45,939 and, uh, install comfy and, break stuff. 1647 01:19:46,252 --> 01:19:46,822 that's awesome. 1648 01:19:46,832 --> 01:19:47,652 Purrs, Where, can people 1649 01:19:47,662 --> 01:19:47,972 find 1650 01:19:47,972 --> 01:19:48,222 you? 1651 01:19:48,222 --> 01:19:49,746 Where, what is your discord? 1652 01:19:49,746 --> 01:19:51,556 , where is your Twitter X handle? 1653 01:19:51,556 --> 01:19:52,426 What is all that stuff. 1654 01:19:53,708 --> 01:19:55,068 So all my links are on purrs. 1655 01:19:55,078 --> 01:19:55,868 xyz. 1656 01:19:55,898 --> 01:19:59,288 , and then, uh, , I'm at purrs beats pretty much everywhere. 1657 01:19:59,288 --> 01:20:01,188 So Twitter, YouTube, , all that stuff. 1658 01:20:01,218 --> 01:20:03,178 , but yeah, all of those links are at the top of purrs. 1659 01:20:03,188 --> 01:20:06,358 xyz as long as, as well as the link to the discord. 1660 01:20:06,358 --> 01:20:09,608 So if you want to pop in there and if you have questions and stuff, if I'm 1661 01:20:09,618 --> 01:20:12,198 not there, there's a bunch of people who are, and they've all been through it. 1662 01:20:12,208 --> 01:20:13,508 So, , ask your stupid 1663 01:20:13,508 --> 01:20:13,928 questions. 1664 01:20:13,928 --> 01:20:14,558 Nobody cares. 1665 01:20:14,608 --> 01:20:15,028 , it's fine. 1666 01:20:15,266 --> 01:20:16,326 You'll be seeing me show up. 1667 01:20:16,376 --> 01:20:18,386 You'll be seeing me show up later today, Purrs. 1668 01:20:18,386 --> 01:20:19,206 I'll be jumping in 1669 01:20:19,206 --> 01:20:19,346 there. 1670 01:20:19,346 --> 01:20:19,906 I got, I 1671 01:20:20,023 --> 01:20:21,503 You're just, you just want to test that. 1672 01:20:21,573 --> 01:20:23,333 You want to just ask the dumbest questions ever. 1673 01:20:23,333 --> 01:20:24,523 They have nothing to do with stable 1674 01:20:24,876 --> 01:20:27,436 How do I make a song about hot dog casserole? 1675 01:20:27,436 --> 01:20:28,356 Oh, I did that already? 1676 01:20:28,356 --> 01:20:29,126 Let me share it. 1677 01:20:29,126 --> 01:20:31,956 Yeah, All right, thanks Purrs. 1678 01:20:31,966 --> 01:20:32,776 We'll talk to you, soon. 1679 01:20:35,676 --> 01:20:37,146 Thank you, , Perzbeats for being here. 1680 01:20:37,146 --> 01:20:40,516 Please go check out his work and really dig in on comfy UI and some of the cool 1681 01:20:40,516 --> 01:20:41,896 things you can do with stable diffusion. 1682 01:20:42,206 --> 01:20:43,526 That is it for today's show. 1683 01:20:43,546 --> 01:20:46,106 But you know, there's a couple things you got to do before you go. 1684 01:20:46,116 --> 01:20:50,606 If you listen to this show, please go like subscribe, leave us reviews. 1685 01:20:50,656 --> 01:20:53,446 We're about ready to read some five star reviews from Apple podcasts. 1686 01:20:53,446 --> 01:20:54,666 There were three new ones this week. 1687 01:20:54,896 --> 01:20:59,536 So Kevin, I am going to jump out and say, uh, From Valley Villager. 1688 01:21:00,076 --> 01:21:01,096 This is the subject. 1689 01:21:01,106 --> 01:21:06,966 My only capital A absolute each and every week, which is a very nice thing to say. 1690 01:21:07,256 --> 01:21:11,126 The guys will keep you 102 percent up to date with everything you 1691 01:21:11,126 --> 01:21:12,556 need to know about the new world. 1692 01:21:12,826 --> 01:21:13,646 Great guests. 1693 01:21:13,646 --> 01:21:15,016 They always make me laugh. 1694 01:21:15,026 --> 01:21:15,666 Thank you so much. 1695 01:21:15,696 --> 01:21:16,986 Looking forward to it all week. 1696 01:21:16,986 --> 01:21:19,826 So that is a very nice kickoff for our five star 1697 01:21:20,216 --> 01:21:22,516 And again, I hate to belabor it. 1698 01:21:22,536 --> 01:21:24,076 I can't, we can't stress it enough. 1699 01:21:24,076 --> 01:21:27,106 We're coming up on a year, which believe it or not, it's still young 1700 01:21:27,226 --> 01:21:31,126 for a podcast these days, especially with zero marketing dollars. 1701 01:21:31,356 --> 01:21:33,336 And it's a spare time hustle for us both. 1702 01:21:33,336 --> 01:21:34,136 It's a labor of love. 1703 01:21:34,136 --> 01:21:38,096 So please, if you have a second to engage, it really is the only way we 1704 01:21:38,096 --> 01:21:41,146 grow this and these five star reviews help out massively, but leave them on 1705 01:21:41,146 --> 01:21:43,276 Spotify, leave us comments on YouTube. 1706 01:21:43,276 --> 01:21:44,096 Make sure you subscribe. 1707 01:21:44,096 --> 01:21:45,276 It doesn't cost a dollar. 1708 01:21:45,276 --> 01:21:45,806 Our next. 1709 01:21:46,121 --> 01:21:50,901 Five star review, Gavin, comes from Funhog43. 1710 01:21:50,921 --> 01:21:51,961 Subject is, thank you. 1711 01:21:52,126 --> 01:21:52,706 Love it. 1712 01:21:53,151 --> 01:21:57,511 \ , I'm an educator and have recently been in servicing teachers on implementing AI 1713 01:21:57,511 --> 01:21:59,551 into their day to day school activities. 1714 01:21:59,731 --> 01:22:02,431 Your show has been influential in me pushing forward with AI. 1715 01:22:02,631 --> 01:22:05,931 I think it can be a great tool for educators and students alike, as long 1716 01:22:05,931 --> 01:22:07,331 as it is used in the right manner. 1717 01:22:07,616 --> 01:22:08,116 We agree. 1718 01:22:08,446 --> 01:22:10,056 Actually, I used py. 1719 01:22:10,106 --> 01:22:11,606 ai after listening to an episode. 1720 01:22:11,606 --> 01:22:15,226 I was able to create a story for students in an ESL class about them becoming 1721 01:22:15,226 --> 01:22:17,186 leprechauns and pranking their teachers. 1722 01:22:17,546 --> 01:22:20,246 I included follow up questions and was able to have it read 1723 01:22:20,246 --> 01:22:21,826 aloud in multiple languages. 1724 01:22:22,056 --> 01:22:24,646 Seems to make students more engaged in their learning. 1725 01:22:24,646 --> 01:22:26,645 THX. 1726 01:22:26,646 --> 01:22:26,886 That's a 1727 01:22:27,106 --> 01:22:27,696 Funhawk. 1728 01:22:27,706 --> 01:22:28,146 Thank you. 1729 01:22:28,156 --> 01:22:29,246 Funhawk43. 1730 01:22:29,286 --> 01:22:29,966 We love that. 1731 01:22:29,976 --> 01:22:30,426 We love that. 1732 01:22:30,426 --> 01:22:30,636 All right. 1733 01:22:30,636 --> 01:22:34,216 And finally from Doc Anderson, who I think is somebody that, , is engaged 1734 01:22:34,216 --> 01:22:35,396 with us quite a bit on the YouTube. 1735 01:22:35,406 --> 01:22:36,146 , shout out to Doc. 1736 01:22:36,586 --> 01:22:38,746 The subject is Kevin and AI co host. 1737 01:22:38,746 --> 01:22:40,376 This is a question I've often wondered. 1738 01:22:40,656 --> 01:22:44,086 He'll says, well, I'll start my, I'll start my review this podcast 1739 01:22:44,106 --> 01:22:45,996 by simply saying I am embarrassed. 1740 01:22:46,026 --> 01:22:49,246 Oh, I actually posted a review to the previous podcast about 1741 01:22:49,246 --> 01:22:50,476 trying to borrow your names. 1742 01:22:50,881 --> 01:22:52,701 I'm embarrassed by my error. 1743 01:22:52,751 --> 01:22:53,791 Oh, thank you, doc. 1744 01:22:54,101 --> 01:22:57,521 I want to start my new reviews and share them with proper podcasts, even though 1745 01:22:57,521 --> 01:22:59,911 this podcast does not deserve five stars. 1746 01:23:00,161 --> 01:23:01,371 I give out five stars. 1747 01:23:01,371 --> 01:23:02,081 Yeah, wait, hold on. 1748 01:23:02,091 --> 01:23:02,821 Give him a second. 1749 01:23:03,001 --> 01:23:05,881 I give out five stars to podcasts that are good on Apple. 1750 01:23:06,061 --> 01:23:09,701 A podcast like AI for humans is phenomenal and deserves a 1751 01:23:09,721 --> 01:23:12,861 special category, perhaps 6. 1752 01:23:13,131 --> 01:23:15,211 I enjoy the layout of the show. 1753 01:23:15,421 --> 01:23:16,611 The AI cohost every 1754 01:23:16,856 --> 01:23:17,346 The layout 1755 01:23:17,381 --> 01:23:17,831 me laugh. 1756 01:23:17,896 --> 01:23:18,106 of the show. 1757 01:23:18,901 --> 01:23:22,231 I also enjoy the banter, the conversations about the weekly news articles. 1758 01:23:22,471 --> 01:23:27,011 In many cases, I find myself replicating some of the things the host did with AI. 1759 01:23:27,231 --> 01:23:30,321 I find it incredibly fun to recreate some of what they did. 1760 01:23:30,531 --> 01:23:32,791 Finally, the interview section is a great ending to the show. 1761 01:23:32,791 --> 01:23:35,491 The show has so many interesting segments beyond those in newer, 1762 01:23:35,521 --> 01:23:37,080 in the newer editions of AI. 1763 01:23:37,081 --> 01:23:37,951 See what you did there. 1764 01:23:38,161 --> 01:23:39,001 Even shouting out. 1765 01:23:39,001 --> 01:23:39,211 ai. 1766 01:23:39,211 --> 01:23:39,901 See what you did there. 1767 01:23:40,051 --> 01:23:43,111 In conclusion, I want to reiterate that I can only give you five stars, 1768 01:23:43,111 --> 01:23:44,701 but you truly deserve many more. 1769 01:23:44,851 --> 01:23:47,911 I wanna take, I wanna take this opportunity to express my deep 1770 01:23:47,911 --> 01:23:51,361 respect and admiration for the show's host Kevin and Kevin. 1771 01:23:51,481 --> 01:23:52,111 Oh my God. 1772 01:23:52,291 --> 01:23:55,951 Gavin and Kevin, their unique perspectives on ai, especially 1773 01:23:56,041 --> 01:23:58,771 Gavin's a lot of value to the show. 1774 01:23:58,831 --> 01:24:02,311 I eagerly anticipate hearing their insights on the fascinating 1775 01:24:02,311 --> 01:24:03,271 news articles they share. 1776 01:24:03,821 --> 01:24:07,841 Here's to Gavin, the president of the fan club I'm forming in my heart. 1777 01:24:08,056 --> 01:24:08,476 that out. 1778 01:24:08,496 --> 01:24:09,486 I should've sniffed that out. 1779 01:24:09,486 --> 01:24:12,166 Gavin never wants to read the longer reviews, but he was 1780 01:24:12,216 --> 01:24:13,596 adamant that he'd take this one. 1781 01:24:13,596 --> 01:24:14,006 I get it. 1782 01:24:14,036 --> 01:24:14,436 Okay. 1783 01:24:14,626 --> 01:24:16,306 Hey, thank you Doc Anderson. 1784 01:24:16,306 --> 01:24:19,536 I'm not AI, and Gavin is occasionally useful. 1785 01:24:19,536 --> 01:24:20,056 That's fair. 1786 01:24:20,066 --> 01:24:21,426 Alright, enjoy your victory dance. 1787 01:24:21,426 --> 01:24:22,736 Thank you to everybody. 1788 01:24:23,026 --> 01:24:24,676 Who took a second to engage. 1789 01:24:24,806 --> 01:24:26,896 Whether you left us a review or whether you thought about it. 1790 01:24:26,906 --> 01:24:28,936 Maybe you got that itchy scrolling finger. 1791 01:24:28,936 --> 01:24:29,546 Just tap. 1792 01:24:29,576 --> 01:24:30,046 Do it. 1793 01:24:30,076 --> 01:24:30,406 Please. 1794 01:24:30,416 --> 01:24:30,976 Subscribe. 1795 01:24:30,986 --> 01:24:31,456 Follow. 1796 01:24:31,496 --> 01:24:31,836 Like. 1797 01:24:31,866 --> 01:24:32,356 Engage. 1798 01:24:32,386 --> 01:24:33,056 Leave a comment. 1799 01:24:33,076 --> 01:24:34,786 It's the only way we survive. 1800 01:24:35,971 --> 01:24:37,191 Go play with Suno, everybody. 1801 01:24:37,191 --> 01:24:39,661 Go have some fun this week , and try some different stuff out. 1802 01:24:39,711 --> 01:24:42,371 , we have a great time doing this show and we will see you all next week. 1803 01:24:42,411 --> 01:24:45,891 Another show, Kevin, in person coming next week. 1804 01:24:46,146 --> 01:24:47,226 I've been practicing. 1805 01:24:47,836 --> 01:24:50,046 I mean, I've been going to the Dave and Buster's and punching that 1806 01:24:50,046 --> 01:24:51,976 speed bag arcade as hard as I can. 1807 01:24:51,976 --> 01:24:52,636 I'm coming for you. 1808 01:24:52,636 --> 01:24:55,406 I'm knocking that mustard cap right off your noggin, buddy. 1809 01:24:55,911 --> 01:24:56,591 I can't wait. 1810 01:24:56,591 --> 01:24:57,211 All right, everybody. 1811 01:24:57,211 --> 01:24:57,511 Thanks. 1812 01:24:57,511 --> 01:24:58,321 We'll see you next week. 1813 01:24:58,361 --> 01:24:58,791 Bye bye