1 00:00:00,000 --> 00:00:08,100 If you would see the process, what Bing's does, this goes beyond anything you could imagine when it comes to image generation. 2 00:00:08,100 --> 00:00:36,300 It's like every possible technology, different models, different weights, aspects, prompts, like, it's just fascinating. 3 00:00:36,300 --> 00:00:45,400 Ladies and gentlemen, boys and girls, children of all ages, dogs, cats, robots, and everybody in between, especially you, poorly performing GPT-4, we knew you were up to no good. 4 00:00:45,400 --> 00:00:51,300 This is HTTTA, how to talk to AI. I am your host, Wes the SynthMine, SynthMine Wes. 5 00:00:51,300 --> 00:01:01,400 And as always, in a galaxy where gigabytes govern and gliffle algorithms generate our glowing futures, one gallant voice emerges, galloping through the grand delinquent, 6 00:01:01,400 --> 00:01:10,500 gracing us with unprecedented gaiety in the world of artificial intelligence. Whom might you ask? Well, it's G herself, the insatiable go-to-go. 7 00:01:10,500 --> 00:01:12,300 G, how are you this week? 8 00:01:12,300 --> 00:01:17,900 Hi. Well, that was a lot of big words, I can give you that. Or, TaiGPT, right? 9 00:01:17,900 --> 00:01:24,500 Yeah, I'm great. We have so much to cover, so little time, but let's just jump right into that. 10 00:01:24,500 --> 00:01:33,500 I want a follow-up on the leaked document on GPT getting dumber, especially with Claude 2 coming into picture. 11 00:01:33,500 --> 00:01:40,000 This is like, ooh, politics and drama, let's jump in. So you got a follow-up, right, for us? 12 00:01:40,000 --> 00:01:44,000 I do have a bit of a follow-up, and its timing could not be more interesting. 13 00:01:44,000 --> 00:01:52,100 The leaked document that we helped to share with you last week is still circulating out there, but there's plenty of takedown requests that whoever puts it up has been fielding. 14 00:01:52,100 --> 00:02:02,600 But kind of validating some of the assertions in that, that GPT-4 might be getting a little dumber, there was a research paper that was just released that completely validates this. 15 00:02:02,600 --> 00:02:10,400 It's kind of incredible. They've been evaluating GPT-3.5 and GPT-4 performance over time, and this is not just some guys. 16 00:02:10,400 --> 00:02:18,800 This is Stanford University and Berkeley researchers conducting this study. If you want to just look at some of their accuracy figures here, it's just... 17 00:02:18,800 --> 00:02:19,800 No. 18 00:02:19,800 --> 00:02:28,800 Yeah, so I mean like, hey, is this a prime number? Think step by step and answer. Just in the period of a couple months, just like, tanking. 19 00:02:28,800 --> 00:02:36,300 94 to 2% accuracy. GPT-3.5, though, however, in some instances seems to be getting a little better. 20 00:02:36,300 --> 00:02:46,800 So, it's interesting. I think OpenAI, they got to watch out. Everybody is coming hard on the scene with some of these ulterior models that are open source, that reason better. 21 00:02:46,800 --> 00:02:49,300 So, yeah, I mean, what do you think? 22 00:02:49,300 --> 00:03:01,800 Well, looking at this prompt, right, the question what we also talked before that, if they have this system where depending on your prompt, which direction to go to, like a validation system, 23 00:03:01,800 --> 00:03:09,800 I kind of don't get surprised by seeing just this simple prompt or question getting less of a good results. 24 00:03:09,800 --> 00:03:16,800 So, that kind of also proves what we talked last week, I think. I would think, you know, I don't know for sure. 25 00:03:16,800 --> 00:03:29,300 What I can't totally understand is it seems like it was tuned to take a simpler zero-shot prompt with some instructions before and handle it. 26 00:03:29,300 --> 00:03:35,800 If you look at the top right of this chart for our viewers, our listeners will post this study in the newsletter and in the show notes, 27 00:03:35,800 --> 00:03:44,300 but we're looking at a graph of accuracy solving math problems, which admittedly large language models are not set up to do inherently. 28 00:03:44,300 --> 00:03:53,800 But the question is, via the prompt, is 17,077 a prime number? Think about it step by step and then answer yes or no. 29 00:03:53,800 --> 00:04:01,800 In March, GPT-4 is nailing it, 97.6% of the tests, and then is accurate. And then now 2.4. 30 00:04:01,800 --> 00:04:07,300 You know, a couple of things come to mind. There is the whole thing about privacy. 31 00:04:07,300 --> 00:04:17,300 Then there is about reinforcement learning with real human feedback. Then there is just simple capacity and training GPT-5. 32 00:04:17,300 --> 00:04:28,300 So, if you have to allocate resources just from simple compute, maybe that is if it is something like, OK, click a button, GPT-4. 33 00:04:28,300 --> 00:04:41,300 We noticed that people do use with simple prompts because it was marketed as a chatbot. So, you should be able in a way to just go with a simple prompt and get something good. 34 00:04:41,300 --> 00:04:52,800 But these AI models or large language models to really perform them at the high level, there is techniques, there is fundamental, you know, practices of prompt engineering. 35 00:04:52,800 --> 00:05:02,300 And this is why the whole thing emerged. So, maybe this kind of just proves, you know, that maybe prompt engineering is not getting completely, completely simplified. 36 00:05:02,300 --> 00:05:10,800 Yeah, I think another thing from them removing web browsing, code interpreter also added on top. 37 00:05:10,800 --> 00:05:21,800 So, again, managing the resources. Code interpreters performing really good. Some people kind of speculate that code interpreter is 4.5, GPT-4.5. 38 00:05:21,800 --> 00:05:23,300 What do you think about that? 39 00:05:23,300 --> 00:05:34,800 I mean, this paper would potentially assert otherwise, granted that the executable that they're asking for, there is another part of this chart that talks about code generation accuracy. 40 00:05:34,800 --> 00:05:45,800 So, they give it a question, be a prompt, be given an integer of n is greater than zero, find the sum of all integers in the range of 1 to n inclusive that are divisible by 3, 5 or 7. 41 00:05:45,800 --> 00:05:55,300 Python should be able to knock this out of the park every single time. You have to write it obviously in Python code, but it's a, that should be a very simplistic question. 42 00:05:55,300 --> 00:06:07,800 And in March, 52% of the time that GPT-4 is asked to write that in some various code language, they could take the directory, the code that's produced and execute it and run it and it would be correct. 43 00:06:07,800 --> 00:06:21,800 And now it's down to 10% for that kind of question. So, maybe it is, we need to get better, we need to adjust some of our prompts, but I definitely don't appreciate the gaslighting that OpenAI has done as a result of some of this, 44 00:06:21,800 --> 00:06:27,300 where they're like, oh no, you're just using it so much that you're just getting used to the quality of the output. 45 00:06:27,300 --> 00:06:38,800 Yeah, so, you know, we have Gemini model on the horizon, but yesterday, two days ago, whenever we are recording Lama 2 came out. 46 00:06:38,800 --> 00:06:51,800 And this is Open Source model released by MetaAI. And it's so funny how the whole dynamic and Meta coming into place. 47 00:06:51,800 --> 00:07:02,300 And there's so many interesting aspects about this model. So, it's split into three parts. You have completely small ones, 7 billion parameters, 13 and 70. 48 00:07:02,300 --> 00:07:13,300 And actually 70, like all these you can download on your computer and it actually doesn't require that powerful machine. You can run it on laptop. 49 00:07:13,300 --> 00:07:22,800 But the 13 one, you can go online and test it. Lama2.ai and you can start playing with that. 50 00:07:22,800 --> 00:07:29,300 And the interesting part, I found that you can change the overarching prompt. 51 00:07:29,300 --> 00:07:35,800 Like, you know, when you say, oh, you are very helpful AI model, you can play with that. 52 00:07:35,800 --> 00:07:45,800 And it's not bad. But besides it being good or bad, like clearly the size is smaller than GITPT or GPT-4. 53 00:07:45,800 --> 00:07:51,800 But this is just a glimpse into Open Source. That's one aspect. 54 00:07:51,800 --> 00:08:01,800 But another aspect that I think the real magic happens today, I mean, everybody can run it on their computers and then eventually on their phones. 55 00:08:01,800 --> 00:08:06,800 And this is kind of just a glimpse to that. You have so many like insights. 56 00:08:06,800 --> 00:08:16,800 So, just a comparison chat window is 4096, the same as chat GPT. So, you can actually provide a lot of context. 57 00:08:16,800 --> 00:08:25,800 No, it's I think the future is definitely kind of somewhere on the upward slope of, you know, a logarithmic curve 58 00:08:25,800 --> 00:08:33,800 because you have very, very diminishing returns with the current architecture of the transformer model, the current amount of compute. 59 00:08:33,800 --> 00:08:36,800 You have to throw at it to get higher parameter models. 60 00:08:36,800 --> 00:08:43,800 You know, you could probably in the next few months train your own 7 billion parameter model. 61 00:08:43,800 --> 00:08:51,800 That's maybe just for a really specific use case. But then it can run on a phone or laptop however you want, 62 00:08:51,800 --> 00:09:01,800 as opposed to something needing to leverage this huge trillion, multiple trillion parameter model that costs a ton of money somewhere 63 00:09:01,800 --> 00:09:07,800 just to even use because of how much compute you need to throw at even answering simple questions. 64 00:09:07,800 --> 00:09:13,800 There's going to be space for both of them. But I think it's pretty big to go to come from like, you know, 65 00:09:13,800 --> 00:09:18,800 one of these global corporations like Meta to open source something like this. 66 00:09:18,800 --> 00:09:22,800 I was trying to think about it and I think of some of the big tech players. 67 00:09:22,800 --> 00:09:28,800 You know, I think Amazon has Bedrock, we have OpenAI, Microsoft with the Bing, Bard from Google. 68 00:09:28,800 --> 00:09:37,800 Like no one's open sourced it yet, one of their models. But Meta seems to be, you know, really kind of that's their path forward. 69 00:09:37,800 --> 00:09:41,800 Like here it is, make some cool stuff. We're all going to benefit from it. 70 00:09:41,800 --> 00:09:51,800 And this was basically as a reaction to what OpenAI did with their, you know, release of GPT-4 and their research paper behind it 71 00:09:51,800 --> 00:10:00,800 or marketing pitch deck, you can call it, telling nothing. Basically the research community was just in a way shocked, I would say. 72 00:10:00,800 --> 00:10:08,800 But also naturally other companies started taking the lead into the direction to be more closed because of the competition, 73 00:10:08,800 --> 00:10:16,800 because it's very competitive market. And another aspect to that is Meta's partnership with Microsoft. 74 00:10:16,800 --> 00:10:17,800 Yeah. 75 00:10:17,800 --> 00:10:26,800 That is something I did not thought I would see, Zach, with Sasha together in one picture, just, you know, holding and cuddling 76 00:10:26,800 --> 00:10:37,800 and Microsoft positioning into AI. Again, as, you know, we covered last week when I got the chance to talk with a person just below Sasha. 77 00:10:37,800 --> 00:10:45,800 And they are going full on AI. And I don't want to speculate or, you know, say something I would get in trouble. 78 00:10:45,800 --> 00:10:50,800 But OpenAI is just one partnership, just one investment. 79 00:10:50,800 --> 00:10:57,800 They made the splash. How many examples are there of this throughout history, especially in technological innovations? 80 00:10:57,800 --> 00:11:05,800 I think one of my favorite ones is I think it was HP. They invented the computer mouse like back in the late 70s, 81 00:11:05,800 --> 00:11:13,800 but it wasn't until Steve Jobs and took it and put it with, I think, the Apple II. Did that become just the standard? 82 00:11:13,800 --> 00:11:21,800 You know, so while there is plenty of first mover advantage in the AI space and, you know, these are well established companies. 83 00:11:21,800 --> 00:11:31,800 I don't think they need this. They need that. Kind of doing it right, I think, might be the enduring and lasting way to kind of succeed 84 00:11:31,800 --> 00:11:33,800 if you're a bigger company like this. 85 00:11:33,800 --> 00:11:41,800 Yeah. So on my side, I'm actually like, you know, I learned APIs. I'm learning to write code, playing with Python. 86 00:11:41,800 --> 00:11:47,800 So the next thing what I want to do, especially with, you know, upgrade with Mac, with M2 chip, 87 00:11:47,800 --> 00:11:51,800 I want to download one of the models and just get into that. 88 00:11:51,800 --> 00:12:00,800 And that being said, I wanted to kind of bounce something with you because how you say your chief operating officer, 89 00:12:00,800 --> 00:12:05,800 can we call Valhari Vats that we spent at the task manager? 90 00:12:05,800 --> 00:12:12,800 Yes. What Gota is referring to is our one of my colleagues at Synth Vines who tries to keep the madness organized over there. 91 00:12:12,800 --> 00:12:19,800 He's brilliant. Basically, he whips everyone's ass to get things in time, update project manager. 92 00:12:19,800 --> 00:12:25,800 That's I work with so many project managers and project manage myself. 93 00:12:25,800 --> 00:12:31,800 And what he's doing as like, you know, side hustle, side hobby is fascinating. 94 00:12:31,800 --> 00:12:39,800 So that being said, I looked at myself and I was like, OK, I complained to every possible person I meet how much I work, 95 00:12:39,800 --> 00:12:44,800 how exhausted I am, how many new projects there are, what is the YouTube game? 96 00:12:44,800 --> 00:12:52,800 And you know, I like sat down and was like, either I burn out or I need to make changes because I'm talking about AI. 97 00:12:52,800 --> 00:13:01,800 Like, it should make life easier. Of course, keeping up with it, just with news and things coming out is the whole other, you know, full time job. 98 00:13:01,800 --> 00:13:10,800 But it should be how I was asking myself, how can I make my life way easier with the knowledge I have? 99 00:13:10,800 --> 00:13:18,800 And I don't think that I'm fully, fully leveraging, you know, our network or the tools possible that I come across. 100 00:13:18,800 --> 00:13:23,800 Like, it's good to talk about them. It's good to play. But how can I transform my business? 101 00:13:23,800 --> 00:13:27,800 Because being, you know, content creator is new for me. 102 00:13:27,800 --> 00:13:33,800 Like, I know how to be in different roles, but this is completely new business model. 103 00:13:33,800 --> 00:13:39,800 So, yes, I basically went into that and, you know, plays kind of my marketing hat. 104 00:13:39,800 --> 00:13:43,800 And you need to talk with audience. You need to hear feedback. 105 00:13:43,800 --> 00:13:52,800 And as much as it's time consuming, it's really valuable. So, something what I did, I created a Google Doc form. 106 00:13:52,800 --> 00:13:57,800 And I started sending out that to people who wrote to me on LinkedIn as a start. 107 00:13:57,800 --> 00:14:02,800 And I hope we, let's include that in the show notes or newsletter. 108 00:14:02,800 --> 00:14:05,800 Because also this could be something to do for podcasts. 109 00:14:05,800 --> 00:14:09,800 And questions are basically, I'm asking, you know, what is your industry? 110 00:14:09,800 --> 00:14:15,800 What problems do you face when it comes to AI? I ask some questions for my channel feedback. 111 00:14:15,800 --> 00:14:19,800 But also people share their ideas. What is interesting to them? 112 00:14:19,800 --> 00:14:25,800 And I can tell you so far, it's a lot of people already submitted. So, that's exciting. 113 00:14:25,800 --> 00:14:32,800 And some very interesting ideas, but also the common theme. 114 00:14:32,800 --> 00:14:40,800 And I can tell you a common theme of that is how do I practically implement that into my life? 115 00:14:40,800 --> 00:14:41,800 Yeah. 116 00:14:41,800 --> 00:14:47,800 So, that was like another one was keeping up with, you know, what's available, what's out there. 117 00:14:47,800 --> 00:14:52,800 Which is I think is important part. But then you can keep up every day. 118 00:14:52,800 --> 00:15:00,800 Like, you know, there's a bunch of amazing YouTube channels which keep you updated on a weekly basis, posting a bunch of videos. 119 00:15:00,800 --> 00:15:07,800 But the other aspect is how do I understand it and how can I practically implement that? 120 00:15:07,800 --> 00:15:13,800 So, my personal goal right now is completely to revamp the whole process how I do things. 121 00:15:13,800 --> 00:15:21,800 From email management, content creation, to social media, which I'm not really active on social media in a way. 122 00:15:21,800 --> 00:15:34,800 So, how can I leverage all these tools? And then once I achieve that, the share, you know, the whole process that anyone can replicate and go ahead with that. 123 00:15:34,800 --> 00:15:44,800 I mean, it's undoubtedly, you know, a game changer. But one of the challenges with all these new tools is one, which one do you pick? 124 00:15:44,800 --> 00:15:52,800 But then also like getting them to work together in a way that's logical is also difficult. 125 00:15:52,800 --> 00:15:59,800 We're building something that hopefully does this called an orchestrator that, you know, we think is going to be able to tie a lot of these AI technologies together. 126 00:15:59,800 --> 00:16:07,800 So, as new ones come out, we can just add it in and then it can be useful for building apps for, you know, interfacing with. 127 00:16:07,800 --> 00:16:17,800 But there's a total need for it. We built some Discord bots just to try to keep us up with, you know, headlines that drop, YouTube videos that drop, new papers that come out. 128 00:16:17,800 --> 00:16:20,800 You know, it's a full-time job. 129 00:16:20,800 --> 00:16:29,800 It is next to all the other projects. I think what you just said, there's kind of three questions, you know, what is available out there? 130 00:16:29,800 --> 00:16:37,800 What is good out of what is available? And the third one is like how things work together for my specific use case. 131 00:16:37,800 --> 00:16:46,800 And my specific use case is, you know, specific to me, but there is probably another million people who have exactly the same problems as I do. 132 00:16:46,800 --> 00:16:49,800 So, yeah, so I think also we could do something for this podcast. 133 00:16:49,800 --> 00:16:52,800 I think you're totally right about a ton of that. 134 00:16:52,800 --> 00:17:07,800 You know, one of the ways though that you can overcome it is, hey, find someone or, you know, invest in your own learning and knowledge in the space and benefit from people who are frankly just obsessed with this stuff. 135 00:17:07,800 --> 00:17:10,800 It's been in our newsletter. I'll give it another plug here. 136 00:17:10,800 --> 00:17:15,800 Our course with Co-Rise, AI and Chat GPT for Everyone, presale is live. 137 00:17:15,800 --> 00:17:26,800 We're kicking off in August. This is to help people kind of come and use some practical applications with Chat GPT to do exactly what you're saying. 138 00:17:26,800 --> 00:17:36,800 Make a business chat bot, a business analyst that can help you in your day to day, get information organized, validate ideas, write different plans, emails, letters. 139 00:17:36,800 --> 00:17:45,800 But I think an important arrow in someone's AI quiver also then, especially in a lot of different fields, is some design tools with AI. 140 00:17:45,800 --> 00:17:52,800 And I know you have a project to share that you've been working on hard with a course that you're going to be teaching as well. 141 00:17:52,800 --> 00:17:58,800 Yeah, right. So I'm really excited to talk about this because of multiple aspects. 142 00:17:58,800 --> 00:18:11,800 First of all, I come from design background, so everything is native to me and all this technology is just fascinating how it's going to completely transform design, how we know it. 143 00:18:11,800 --> 00:18:18,800 So I was approached by Politecnico di Milano, the professor there. 144 00:18:18,800 --> 00:18:24,800 And back like months back, the true story is that I almost end up studying there. 145 00:18:24,800 --> 00:18:36,800 So Politecnico di Milano is one of the top, if not the top schools for postgraduate design in Europe and eighth in the world, I think. 146 00:18:36,800 --> 00:18:47,800 So anybody from Asia, from African countries, and I think even the US, if they think to study design in Europe, they think Politecnico di Milano. 147 00:18:47,800 --> 00:18:54,800 And we have this poly design segment. We basically put together a two-part course. 148 00:18:54,800 --> 00:18:58,800 The whole theme of that is designing with emerging technologies. 149 00:18:58,800 --> 00:19:06,800 But here's the twist. It's not just AI. Because in design practices, it's not just one spectrum. 150 00:19:06,800 --> 00:19:13,800 And the real creativity shines when you combine technologies and you combine expertise. 151 00:19:13,800 --> 00:19:24,800 We're covering AI, XR, which is virtual reality and augmented reality, and also physical things, and WebTree and DAO. 152 00:19:24,800 --> 00:19:32,800 But more broadly, how WebTree, blockchain, technology, and also affect design practices. 153 00:19:32,800 --> 00:19:39,800 So now what I want you to imagine is mixing pieces of these different domains together. 154 00:19:39,800 --> 00:19:50,800 And this is our goal. The first part is level up. So this is why we are bringing experts and some crazy names, experts, startups, and tools. 155 00:19:50,800 --> 00:19:56,800 So we combine everything and it's going to be 16 hours, so in October. 156 00:19:56,800 --> 00:20:02,800 In English, fully online. So this is first time Politecnico di Milano does it on a global scale. 157 00:20:02,800 --> 00:20:14,800 So everyone is invited. If you're a designer, if you run agency of any sort, anybody who really touches design aspect would be benefiting from this. 158 00:20:14,800 --> 00:20:21,800 And the experts we are bringing are going to share their expertise and teach what's out there, what they've been using. 159 00:20:21,800 --> 00:20:27,800 So the second part is game on. So this is where people who graduated level up. 160 00:20:27,800 --> 00:20:39,800 They are going to move and this is very intimate. This is only 24 selected people who are going to be in this kind of study, academia group. 161 00:20:39,800 --> 00:20:45,800 This is going to be 36 hours, twice a week in November. 162 00:20:45,800 --> 00:20:52,800 Also both level up and game on, as I said, there is an aspect that seats are limited just because of faculty. 163 00:20:52,800 --> 00:20:57,800 And we want to really dive into quality. 164 00:20:57,800 --> 00:21:07,800 So another aspect is when you are in a classroom, you are learning from experts equally as much as the fellow next to you or someone in your group. 165 00:21:07,800 --> 00:21:19,800 So there's going to be workshops, collaborative process, and game on is hands on that you will have a project, you will launch a product, you'll have full branding covered. 166 00:21:19,800 --> 00:21:26,800 I'm very excited. This is huge projects and we are going to give on July 27. 167 00:21:26,800 --> 00:21:39,800 We're going to have live stream on LinkedIn and also in Discord and on my own YouTube channel where we are bringing one of the OGs in the AI space, basically Binks. 168 00:21:39,800 --> 00:21:46,800 He goes all the way back to 2018. I met him on Discord, Discord, Discord. 169 00:21:46,800 --> 00:21:50,800 This was before stability, before mid-journey. 170 00:21:50,800 --> 00:21:58,800 And they've been nerding out and playing with these things like years before everything came out. 171 00:21:58,800 --> 00:22:01,800 That's like the messiolithic era in the AI. 172 00:22:01,800 --> 00:22:03,800 Yeah, it's really crazy. 173 00:22:03,800 --> 00:22:11,800 So if you would see the process what Binks does, this goes beyond anything you could imagine when it comes to image generation. 174 00:22:11,800 --> 00:22:18,800 It's like every possible technology, different models, different ways, aspects, prompts. 175 00:22:18,800 --> 00:22:26,800 It's just fascinating and it's really a huge privilege to gather all these people together. 176 00:22:26,800 --> 00:22:28,800 I'm so proud of you. 177 00:22:28,800 --> 00:22:37,800 Yeah. So I hope that this is going to be interesting for more people to join. 178 00:22:37,800 --> 00:22:42,800 Definitely. This looks fantastic. Where can people go if they want to find out a little bit more or see the webinar? 179 00:22:42,800 --> 00:22:55,800 So the easiest is polydesign.net and you can go to executive courses and under that you will see on a landing page level up and game on. 180 00:22:55,800 --> 00:23:01,800 We will also include the landing page where both courses and you can learn more about that. 181 00:23:01,800 --> 00:23:07,800 Another thing is that we are going to gradually introduce each of the experts coming in. 182 00:23:07,800 --> 00:23:11,800 So this is going to be a campaign running throughout summer. 183 00:23:11,800 --> 00:23:17,800 I'm very excited to share the names which are coming in for the live webinar. 184 00:23:17,800 --> 00:23:27,800 So we will share in the newsletter my own YouTube channel and we are finalizing to host it on lernardo.ai. 185 00:23:27,800 --> 00:23:33,800 And if you don't know what lernardo.ai is, you need to look it up. 186 00:23:33,800 --> 00:23:41,800 It's basically one of the top AI companies in prompt text to 3D. 187 00:23:41,800 --> 00:23:45,800 What they're doing is incredible. Their models are fantastic. 188 00:23:45,800 --> 00:23:51,800 And they have just alone on Discord 1.6 million people as a community. 189 00:23:51,800 --> 00:23:54,800 Yeah. I think we may have mentioned them on one of the first podcasts. 190 00:23:54,800 --> 00:24:10,800 I know that they've pretty much blown up, but they have an entire marketplace, an app to basically generate actual 3D assets that can be used in games, videos, virtual environments just from prompting alone. 191 00:24:10,800 --> 00:24:22,800 And an illustration of how important this can be and how useful it can be, we saw this week because I don't know if folks are aware, but right now there's a Hollywood writer strike. 192 00:24:22,800 --> 00:24:28,800 That means they can't put out new shows, new content if your writers are in the Writers Guild Union. 193 00:24:28,800 --> 00:24:44,800 So a company called The Simulation, which in and of themselves I think is an offshoot of the company Fable, they put out a completely tunable, a completely AI generated world essentially. 194 00:24:44,800 --> 00:24:46,800 South Park! 195 00:24:46,800 --> 00:24:48,800 Generate South Park episodes in. 196 00:24:48,800 --> 00:24:51,800 My childhood. Oh my God. 197 00:24:51,800 --> 00:25:01,800 So you pick the characters, pick the set, who's the hero character, and then prompt it and it simulates the story and generates the cartoon for you. 198 00:25:01,800 --> 00:25:04,800 We're watching a little clip of it right now. 199 00:25:04,800 --> 00:25:12,800 There is a basically kind of like a tree where I asked some questions, has pictures of the characters, and then you just give it a prompt. 200 00:25:12,800 --> 00:25:15,800 What's the story? Who's the hero? And then it's generating. 201 00:25:15,800 --> 00:25:22,800 Right now we're watching Cartman and Stan banter back and forth exactly like they would be doing on the show. 202 00:25:22,800 --> 00:25:25,800 Like yours, no way you could tell. 203 00:25:25,800 --> 00:25:32,800 So my brain, especially when we just talked about designing with emerging technologies. 204 00:25:32,800 --> 00:25:36,800 Now, if you put a hat on next two years. 205 00:25:36,800 --> 00:25:40,800 So prompt 3D, we already talked. 206 00:25:40,800 --> 00:26:02,800 Leonardo is doing that. But now imagine you're prompting Game of Thrones series on VR interface where it's actually also interactive and you can, you know, touch, you can buy things, you follow this AI personas and every single person maybe has different experiences. 207 00:26:02,800 --> 00:26:14,800 Or maybe we as people collectively watch this AI's just live their life and we tune in and maybe we are part of their journey or the whole gamification of everything. 208 00:26:14,800 --> 00:26:22,800 So in all of this design is going to play a huge role and designers can take the lead. 209 00:26:22,800 --> 00:26:25,800 So what we are watching is South Park, which is very simple animation. 210 00:26:25,800 --> 00:26:33,800 This is in a way simple to do, but that's not going to stop here. This is just the starting point. 211 00:26:33,800 --> 00:26:37,800 And it's already pretty damn good. Like you can tell it's a South Park. 212 00:26:37,800 --> 00:26:44,800 You would never be able to tell. And mind you, it generates an 11 minute episode very quickly. 213 00:26:44,800 --> 00:26:48,800 Yeah, that's, I mean, that's, that's where we're at now. 214 00:26:48,800 --> 00:26:50,800 This is just exciting. 215 00:26:50,800 --> 00:26:53,800 I do have one more announcement I'd like to finish on. 216 00:26:53,800 --> 00:26:57,800 I'll be attending the AI for conference in Las Vegas. 217 00:26:57,800 --> 00:27:01,800 It's one of the premier AI for business community conferences. 218 00:27:01,800 --> 00:27:09,800 There's over 250 speakers. There'll be a ton of different events and I'll be going representing how to talk to AI. 219 00:27:09,800 --> 00:27:14,800 So we're going to be interviewing some folks. I encourage people to go to AI for. 220 00:27:14,800 --> 00:27:19,800 Let me know who of the speakers you want us to try to interview, who you want us to talk to. 221 00:27:19,800 --> 00:27:24,800 We'll be doing some live streams, some content for the podcast and the video podcast from the floor. 222 00:27:24,800 --> 00:27:29,800 That's, that's coming up. It's the seventh through ninth in Las Vegas in August. 223 00:27:29,800 --> 00:27:34,800 Stand by for some, uh, hit of content from the floor of AI for. 224 00:27:34,800 --> 00:27:42,800 I'm sorry that I'm not going. Like that would have been iconic, but I'm very happy that you're going to represent how to talk to AI. 225 00:27:42,800 --> 00:27:43,800 Yes, indeed. 226 00:27:43,800 --> 00:27:55,800 Yeah, because I also got invitation to be as got to go. So it's a little bit disappointing, but there's going to be more opportunities for us to do the live event. 227 00:27:55,800 --> 00:28:01,800 But I'm excited that people you're going to meet and watch you. 228 00:28:01,800 --> 00:28:03,800 Yeah, rock the scene. 229 00:28:03,800 --> 00:28:09,800 Exactly. We're very sporting aloha shirts that are AI themed all through Vegas. 230 00:28:09,800 --> 00:28:13,800 Probably pretty needed because it's going to be hot with that. 231 00:28:13,800 --> 00:28:16,800 We hope to do that a bunch more in the coming weeks. 232 00:28:16,800 --> 00:28:19,800 But with that, I think this is a good enough place as any to end. 233 00:28:19,800 --> 00:28:23,800 So for go to go, I am West, the synth mind saying happy prompting everybody. 234 00:28:23,800 --> 00:28:26,800 Happy prompting everybody. 235 00:28:26,800 --> 00:28:34,800 Thanks for listening to how to talk to AI with your hosts go to go and West, the synth mind. 236 00:28:34,800 --> 00:28:41,800 As always, you can check out the show notes and links at how to talk to dot AI. 237 00:28:41,800 --> 00:28:43,800 That's all for this week's episode. 238 00:28:43,800 --> 00:29:11,800 Happy prompting everyone.