Transcript 0:00 Welcome back to the Middle Tech Podcast. We finally have spring here in Kentucky. 0:04 I got to go to Keeneland this weekend, which was, uh, great to get out there, see everyone, enjoy some of that nice weather, and, uh, bet on some horse races. 0:11 So I know your allergies attacked you a little bit out at Keeneland, but- Yep... welcome to spring in Kentucky. That's how it goes. Yeah, yeah. So we have a great episode today covering all things in the world of AI. 0:21 Uh, so like we've continued to do for most of this season, we've e- even introduced a series specifically about AI. 0:27 Uh, this episode is gonna be entirely about news coming up in the AI space because things are just moving so fast right now. There is new news and new developments every day, every week. 0:36 Uh, it truly is one of the fastest technological revolutions that we're gonna get to experience in our lifetime. 0:42 Uh, so we wanna dedicate an entire episode to just talking about some of the recent news, uh, in the space and dive into that. 0:49 So the stories we're gonna be discussing on this episode are Elon Musk hinting at launching an, uh, open AI rival, uh, and what Twitter, uh, has to do with that and his rebrand to x.com. 1:01 Uh, we're also gonna be talking about, um, Amazon, uh, and them getting into the generative AI space with, uh, with their cloud computing platform. 1:11 And then we're also gonna be talking about this new development called AutoGPT, which is probably one of the most exciting developments I've seen in a long time. 1:18 We spent some time talking about it this morning, so we're excited to dive into that topic as well. Before we get into those stories, please remember to subscribe to our channels, follow us on socials at Middle Tech Pod. 1:29 Uh, and before we dive in, we just wanna get a quick word from our sponsors. [whooshing sound] Middle Tech is presented by KY Innovation, the Kentucky Cabinet for Economic Development's Office of Entrepreneurship. 1:39 KY Innovation is currently accepting applications for its SBIR/STTR matching funds program. 1:46 Kentucky's program is a national trendsetter for providing critical non-dilutive capital to support the high-tech research, development, and commercialization of novel ideas within the Commonwealth. 1:56 SBIR matching funds have directly contributed to the creation of more than 700 jobs, more than 200 patents filed, and more than $160 million raised in private follow-on funding, and to dozens of companies relocating to the state. 2:10 The matching funds application window is open through April 24th. To apply and learn more about support services for companies that want to explore SBIR/STTR grants and contracts, visit kyinnovation.com/sbir. 2:25 [upbeat music] Middle Tech is sponsored by Bolt Marketing. Take your website to the next level with a website that's built to work. 2:31 At Bolt Marketing, they're revolutionizing websites for small businesses that are affordable, customizable, and hassle-free. 2:37 Whether you have a construction company, a boutique clothing store, or you own a hot yoga studio, they have options for you. 2:43 Click the link in our show notes to explore their marketing options that can transform your marketing and grow your business. 2:49 And as a personal note, Bolt Marketing built our website, and they were awesome to work with throughout the entire process. We highly recommend working with them. 2:56 [whooshing sound] All right, so first topic we're gonna discuss now is Elon Musk kinda changing the name of Twitter from Twitter to X Corporation, so 3:07 basically merged Twitter into a newly created X Corporation, and he also bought, uh, the domain x.ai and incorporated another business in Nevada. 3:17 [clears throat] So the whole move here is he is taking Twitter from really a legacy media player, uh, and social media app into, uh, the AI world. 3:28 So part of that, like Logan mentioned at the top of the episode, is he's wanting to create a competitor to OpenAI. Uh, he has been very vocal about the fact that OpenAI, uh, has rather biased results. 3:41 Uh, it is also moving faster with, uh, a lot less governance than he would like, so he's trying to get his horse in the race and see what he can build. I think, uh, Twitter is a very unique opportunity. 3:55 Uh, I personally thought that the Twitter acquisition, you know, now in hindsight, was a very intelligent move. The timing was not a coincidence that he bought it before, uh, OpenAI released their model. 4:08 Uh, that was the reason he probably, uh, was urgently buying it and paying well above market for it, was that he just wanted to get that deal done. Um, so- Did you see on Spaces real quick on that point- Yeah... 4:20 he, he did admit to that he was forced to buy it essentially by lawyers. Yeah, yeah. But that was only after he tried to decrease the price- Right... because of the bots and fake accounts. 4:30 But I think all along he was not gonna let that, uh, get out of his hands. 4:34 So I would say what I picture happening here is there's a huge, you know, movement towards these, uh, large language models, and Twitter has just an unbelievable amount of data that's kind of real time, conversational data that, you know, spans back to when they built Twitter in the very beginning days. 4:55 It's a very large data set, so I would imagine he has a ton of training data for his models, uh, on top of everything else that OpenAI was able, able to leverage. 5:04 Um, so I think it's a very unique opportunity, but I think what he's trying to move towards with this X Corporation and this xAI is he's been alluding to the app of everything- Mm-hmm... 5:14 which is essentially like a chat or a, um, WeChat competitor, which WeChat is in China, and it's one of those apps that just does everything. 5:23 You can buy goods and services on there, you can chat with your friends, it's social media. You can, um, pay for things in physical retail stores, you can hail, you know, Uber rides. 5:35 So it's kind of an app that, uh, does everything, and he wants to do the same here in America with Twitter. Um, so that could entail payments. Uh, he could use AI to curate the feed. 5:45 He could use AI to basically give you real-time data, so if you have a question about the NBA scores, you know, you can quickly go to Twitter and get those, as well as video content and commentary from your friends. 5:56 So I think there's a unique opportunity here And again, I think that it was not a coincidence he bought it before ChatGPT was released. Yeah. Yeah, it makes you wonder. I wonder how much more Elon can take on. 6:07 It seems like he's got a new project that is coming up every month at this point. Um, but I think you did a pretty good job summarizing what's going on around here. There's not a ton of movement on this yet. 6:16 It's just more so seeing the indications that this is kinda going on in the background. There's some buzz when he, uh, filed for x.ai and the X Corporation. Uh, so we'll keep on providing updates to that. 6:27 But, uh, like Evan said, I mean, I think it is timely, uh, when he did buy that, and I think it does-- There are a lot of, um, benefits to be realized by having AI, uh, an AI company and, uh, leveraging Twitter data and conversational data that is on Twitter, uh, to, to build that company. 6:45 So- Yeah. He's, he's been recruiting from OpenAI, uh, DeepMind, and some of the other competitors, uh, trying to take away talent. Mm-hmm. 6:54 So we'll see how that, uh, turns out once those people start officially announcing they've joined, you know, X Corporation. That'll be interesting. 7:03 Um, I do wanna make one point that I think, uh, needs to be talked about as it relates to media. Um, I think there's been a, a foundational shift. I've been talking to my friends about this, um, a good amount, that 7:16 because of what's happened with, with Twitter and now AI and, uh, what changes Elon's been making, a lot of people are asking, you know, what is the future of media? 7:25 You know, whether it's, uh, marking, you know, NPR as a government-funded organization, or it's the blue check marks, whatever it might be, these are all small steps towards preparing for the future of media, which is gonna be completely different than it is now. 7:42 And media is gonna move from being a service provided by journalists to being a product. Because content is no longer gonna be a bottleneck. Content's not gonna be the hard part. 7:54 The hard part is gonna be building a product around either content automatically generated or content created by journalists that's easily accessible and curated to a person's unique taste, and that just hasn't been created yet. 8:07 Uh, Twitter didn't really do a good job of that, uh, prior to Elon taking over. But now with AI, uh, media's gonna become, uh, just a product. It's not gonna be something that humans need to create anymore. 8:20 It's gonna be how do humans interact with it and get value out of it and curate it. Because again, the bottleneck is not gonna be creating blogs, videos, tweets, um, stories, whatever it might be. 8:31 I mean, AI's gonna generate all of that, and it's gonna be how you give your unique twist. 8:35 And so a product's gonna have to be the thing that allows you to give it that unique twist but also find content that's reled- rel- uh, relevant to you. 8:43 Um, so I think, you know, legacy media players, you know, like NPR, um, are gonna be exposed over the next several years because, uh, you're gonna be able to see how far off from the truth a lot of them are. 8:54 Uh, and it's just kind of a, a scary situation. I mean, the media's kind of totally up in the air on where it's gonna go from here 9:01 just because it's, it's gonna move from the power being in journalists' hands to, uh, engineers and, and product designers. Yeah. 9:08 The history books here in, like, 20 years when we look back on this time in history of kind of the erosion of confidence in media at the same time that generative AI is coming online. You know, we went through COVID. 9:19 We saw all this, I guess, propaganda is a word you could use for what was going on in, in mainstream media around, uh, COVID and, uh, our reaction to that. 9:28 And then we get past COVID, and it comes out that, you know, the, uh, maybe the, um, immunizations weren't as effective as what they were saying, and, you know, all of these other things have come to light, uh, now that we're on the other side of that. 9:41 It's eroded a lot of people's confidence in mainstream media, and then at the same time, exactly what Evan's saying, uh, we're seeing generative AI kinda drive the cost of content down to zero. 9:51 Uh, it's really going to shift things, shake things up, and, uh, lead us into a completely different environment for media going forward, in our opinion. Um, let's move on to this next story. 10:00 Give us a summary of what's going on as Amazon throws its hat in the ring with generative AI with its, uh, Bedrock platform. Yeah. So Amazon, largest provider of cloud services for developers. 10:11 Uh, a lot of applications, including, uh, my own Simba is, is built on Amazon largely. 10:16 And so a b- essentially what's happened, uh, over the last, you know, 20 years is Amazon built, you know, cloud services, which is basically they took the need to have on-prem servers, and they put them in the cloud, and then somebody can spin up an application without having to actually own and buy the servers themselves. 10:34 They can basically just rent them from Amazon. And so they've built the world's largest platform for this, which powers the internet. 10:40 And now that AI has come out, what they're doing is they're incorporating these pre-trained large language models and other types of models into their AWS services. 10:49 So AWS is not just providing, you know, servers for businesses to build on. They're providing all kinds of other developer tools from authentication, communication, uh, to, uh, authentication, payments. 11:03 You know, there's all kinds of things that Amazon is providing businesses, uh, but they're now adding the ability to incorporate AI models, uh, into a product. 11:13 And so they're not only building their own AI models, which they refer to as Titan foundational models. 11:19 Uh, they're also leveraging third-party models like Stable Diffusion, Anthropic, uh, and others that have pre-trained data sets for developers to just out of the box immediately use. So it's a really big deal. 11:30 Um, you'll probably start to see them launch a lot of different verticals that are pre-trained, uh, within their largest accounts and kind of l- use the existing, uh, Amazon AWS service, um, as a way to say, "Okay, uh, we've got this many customers in this industry. 11:47 Let's build some pre-trained models here." So it might be healthcare. Or, "We've got, uh, this many customers over here in construction. Let's build some pre-trained models in construction and architecture." 11:57 Um, so they've got kind of a lot of data to start, uh, informing which models to do next. But it's a really big deal because now any company can, uh, build within their own AWS environment now with... 12:08 these models without having to use, you know, other technology that's gonna fragment, uh, maybe their service provider on what they're using for most of their code base. Yeah. 12:16 So when I see this play, I see it as an accessibility play. 12:20 Uh, I think this makes it a lot easier to build, um, build using machine learning interfaces and, um, you know, these, like, things like large language models and generative AI. 12:29 Uh, I see this as more of a way to make that more accessible to all the different companies that use Amazon Web Services, uh, to host their, their do- their domains. 12:38 Um, what are your initial thoughts when you see them making a move like this? 12:42 Yeah, I mean, there's so many applications to leveraging these models, and I think to your point, just the easier they are to access with better documentation, right in line with the rest of the services you might be using within AWS, it's a big deal. 12:54 Um, so I think you'll start to see so many applications start to layer this in and not have to go to, uh, switch to A- to Azure, which Microsoft provides, um, just to get, you know, these models. 13:07 Now, uh, if you're on AWS or you're on Google, uh, you can have these models wherever you are. So all of the big cloud players are gonna do this. Mm-hmm. 13:15 All right, well, let's move on to this last story here and, in my opinion, one of the most significant, uh, and it's something that's really just come out here in the past, uh, week, week and a half, and that is all the buzz around AutoGPT. 13:26 Um, so AutoGPT, it's experimental open sourced, uh, Python application that's using GPT 4, using the Open AI's API, uh, to essentially create an application that can create AI agents that can talk back and forth to each other. 13:42 Um, and the reason this is so significant, the way I'm starting to comprehend this, we spent the morning kind of researching it and talking back and forth about it, is it, it brings it closer to human thought. 13:52 So it's, if you think of each response from like ChatGPT as a thought, it's basically allowing these multiple AI agents to talk back and forth to each other recursively, uh, and string together multiple thoughts to perceive things, uh, to act, uh, to improve upon itself, and essentially have almost like an internal dialogue in the same way that humans do. 14:12 I think that is a crazy advancement. I think that's a step towards AGI. Peop- I think one of the models was named Baby AGI. Um, what are your initial thoughts on this? 14:21 This c- this is one of those things that I look at, and once I started truly understanding, I was like, "No shit." Yeah. Like this is, this is pretty significant. Yeah. 14:28 It, this is one of the larger developments that's happened since ChatGPT launched. So s- I mean, an example of this would be that I, that I came up with essentially is you're about to go on a trip to Costa Rica. 14:38 Uh, you can tell AutoGPT, "Do some research for me on Costa Rica and plan a trip and build a presentation with that so I can show it to my friends." 14:47 And so normally, that would be several steps that you would have to take within ChatGPT. You'd have to say, "Build a list of all the things to do in Costa Rica," 14:57 uh, and then you'd take all of those and you'd go to a different application and you'd say, "Okay, here are the list of things that ChatGPT gave me. 15:04 Now I'm gonna put this into this presentation software, and I want this presentation software to build that presentation." Um, so those are like two different actions, but now with AutoGPT, they're the same action. 15:15 And then let's say that you like the presentation, then you can tell AutoGPT, "Okay, go book, uh, my flights, go book the hotels, make sure the hotels have five stars and cost less than $400 a night," uh, and it'll go do all of that for you. 15:29 It'll book all that travel, uh, for you, and it'll put together an itinerary for you, and then you can even have it put it into a document really well organized. 15:37 So that could be one command, and it does all of those tasks by talking to different AI agents back and forth in other APIs. Yeah. 15:46 And so really now it's you give it a goal, and it figures out how to complete that goal without you having to interfere with it, uh, multiple times to reach the goal, which is a really big deal. Yeah. 15:56 And I think to try and kinda go through what's actually going on, it's, it's using the large language model to create the prompts, and then it's talking back and forth with itself based off of those prompts to create essentially a task list. 16:08 Then it's able to act on those tasks. So let's say you give it that high-level prompt like what Evan was saying, "Book a trip to Costa Rica." 16:14 The prompt that it might give you if you're just in ChatGPT would be like, "Hey, here are the four things you need to do to book a trip to Costa Rica." 16:20 But rather than you having to respond back and, you know, k- uh, essentially, uh, prompt it to keep on going, it can talk back and forth to itself, uh, and just continue to, uh, to kinda have that train of thought and then go and connect to the internet to actually perform actions. 16:37 Which is the-- I think that was one of the things that we were talking about early on during this is once we connect this all to the internet and allow it to take actions, that's gonna be one crazy step, and now here it is. 16:47 Uh, and the craziest part is that this is literally only like a week and a half old, but you're al- already seeing some incredible use cases come out of it. 16:53 One of the ones that we were most excited about, we saw somebody that created a, an AutoGPT for doing podcast research. So they basically said, "Here's, uh, the theme of the podcast. 17:03 Go out and research recent news topics that we can talk about on this podcast, summarize them, uh, and put them into a discussion note format for us." And it literally went and did exactly that. 17:12 It found like the top five news stories, uh, research articles on them on the internet, provided a summary of them, and then put it into a format for them to record a podcast. 17:20 So obviously something we're gonna be looking into, but it really is just like kinda that, that next step towards artificial general intelligence, being able to really have like your own AI personal assistant. 17:31 And, uh, we were listening to, uh, the All-In podcast, uh, on the way over here. 17:36 They had some really interesting thoughts about what this is gonna do to the startup space, what this is gonna do to the VC space, and I thought those were some really interesting, uh, takeaways from where this technology could take us. 17:48 Because essentially what they're saying is rather than needing a bunch of developers and a bunch of capital to go and spin up the minimal viable product for your idea, we're getting very, very close to being able to just speak your idea in plain English to, uh, something like AutoGPT, and then it going and creating, uh, the first iteration of your product by just connecting together APIs and Using this large language model and prompts to be able to- Yeah... 18:14 to give you an MVP. It'll, it'll choose the database. Yeah. It'll prove where everything's hosted. 18:19 Um, it'll, you know, choose what the payment provider is, and it'll just build, you know, those aspects of the application for you and piece them all together. Uh, it's, you know, it's wild. 18:29 It's gonna affect, uh, the business world, uh, tremendously. Um, I think- Yeah... that another thing they were talking about on that episode that I think we've talked about a lot on this podcast is governance. 18:41 How are we gonna protect, you know, ourselves from this getting out of control and not being aligned with our own goals as humans? 18:47 And, uh, this is kind of [chuckles] the moment we have to look at ourselves and say, "What are we gonna do about this?" 18:53 Because once it starts taking its own actions, and we don't have to intervene with those actions, uh, you know, it could do stuff that we don't want it to. 19:00 So, uh, I think this is just kind of a, a special time, and we gotta be paying close attention, and I think the government needs to do something, uh, to govern this thing. 19:10 I'm, I'm not sure, uh, what, what happens from here. Yeah. 19:14 Well, I mean, I think a perfect example of that is somebody has already used AutoGPT to create a program called ChaosGPT, uh, and they've, they've basically made, uh, the objective to destroy humanity. 19:25 Just seeing what it would do, I guess. I don't know. That, I don't even know why you would play [chuckles] around with that, but you have to, I mean, you have to assume it's- Well, somebody has to do that. 19:31 Yeah, as soon- Like, that's a no-brainer thing to do, though. Yeah, exactly. Yeah. As soon as you put something like that out, of course. 19:36 Um, but it really does call attention to, like, we have to make sure we're building this in the right way, or a prompt like that, you know, once you get to scale of interconnectivity and, you know, enough, uh, enough of an advancement in this AI technology, it could really cause some, some real world consequences. 19:54 Uh, and I think it's only a matter of time before we keep on tinkering with it, and you get a prompt like ChaosGPT. Um, any final thoughts to add onto that, though? This is, uh, one of those stories that I think is just 20:05 gonna continue to blow our minds for probably the next, next year into the future. Um, other than that, uh, I don't think we have a tip of the week for this, this week's AI Edge. 20:15 Our tip of the week would be to research AutoGPT, uh, try to tinker around with that, and see if you can figure out some use cases to it. Um, other than that, we do not have a spotlight series episode this week. 20:26 We'll have one of those for you guys next week, so hope everyone has a great week, and we will talk to you next Monday. Yep. See ya. 20:33 [upbeat music]