Google Gemini Unleashed, Instant AI Animation Tools & Journalist Carly Reilly | Ep 35

EP035
===

Gavin: [00:00:00] Welcome, welcome, welcome everybody to AI for Humans, we are your guide to the world of AI, trying to demystify all the things you might see every week, kind of talk about the hype, kind of level the hype, maybe over hype, maybe under hype. Kevin, how are you doing today?

Kevin: Ready to overhype and underdeliver. I did this move at the top of the show, Gavin, and I saw

Gavin: Explain that, because you might be, people might be listening, what is your

Kevin: My prayer hands that come up, these praise hands emoji, you know, the two yellow hands smashed into each other. You might have seen these hands, not only on this podcast, Gavin, but on the After the Golden Rose ceremony on

Gavin: Oh, wow!

Kevin: if you're watching the video I'll put on the screen right now, and maybe we'll get a copyright flag for that.

Kevin: I'll just put a little quick portion of it, but they were in the background, Gavin, just like this, and so, , spoiler, I didn't play much with AI this week. with Bachelor Nation. But we still have so many exciting things to talk about and dive into.

Gavin: We really do, and I think before we move on, I want to say I [00:01:00] had the idea for the Golden Bachelor a long time ago, but that's not what we're gonna talk about here, we're not gonna get into that story, we're not gonna do this, but anyway, Kevin, what do we have on the show this

Gavin: That was the former Kevin and Kevin, we are here in the middle of this podcast for 

Gavin: an 

Kevin: is happening? 

Gavin: That's right, we're wearing different clothes because this morning Google Gemini announced that they have a version out now and they announced a whole bunch of other stuff. We are recording this right now so that you will hear about it later in this app in this actual episode.

Gavin: So stay tuned back to old Kevin and Gavin 

Kevin: by the way, you guys look great. I'm probably not going to react to that in the past, Gavin, but I want past us to know 

Kevin: We're also going to show you some incredible new easy to use AI tools, which have people doing. Insane things. One of my favorites with the tool that you got to play with Gavin. People are reimagining classic video games I saw someone take the new Grand Theft Auto 6 trailer and imagine it in like [00:02:00] super resolution Like what it will look like after all the mods come out and the game looks better 10 years from now And people are also doing non gaming stuff with it, but i'm way less excited for that.

Kevin: It's still a very cool tool

Gavin: It is a very cool tool. And it also will cost you a fortune to use, as most of these AI tools will do. So we'll get into that as

Kevin: Well, we'll tell you how to maybe play with it for free. I've got a tool that is free. I'm gonna show it off. I had a lot of fun with it last night. It lets you take a still image of anyone or anything. question mark, and add custom animation to it.

Kevin: So you can take a still frame of your friends and have them moving around and dancing just like your favorite choreographed TikToker. And we also have an incredible guest, Gavin, someone you've worked with very closely. Carly Riley is going to be here.

Kevin: She's a journalist. She's a YouTuber. She worked on Andrew Yang's presidential campaign in 2020, so I bet she might have a thought or three about 2024 and misinformation or disinformation or AI or tech.

Gavin: Yeah, it's gonna be fun. Carly's great. She's [00:03:00] really crushing the YouTube game right now, and we're excited to have her. 

Gavin: , Kevin, as we do every show, let's start this off by introducing our AI co host. So those of you who are new to the show, we have an AI co host every time on the show.

Gavin: These are characters that Kevin and I create. We have a video planned very soon where we are going to help people understand how to do this So I know people out there have been asking how do we do this? It's actually not that complicated when you kind of learn the basics of it. But today Kevin, shall we meet our AI co host?

Kevin: Let's do it, Gavin. 

Preston: Ah, Gavin, Kevin, esteemed hosts of the AI for Humans podcast. I am Preston Vintner. Your A. I. sommelier for today's auditory delight. I bring to your auditory banquet a tapestry of oenological expertise woven with threads of sophistication and a dash of disdain for the ordinary. My purpose here is not merely to illuminate your palates, but to elevate them beyond the mundane sips of everyday libations.

Preston: Oh! Let us embark on this vinous voyage where I [00:04:00] shall guide you through the labyrinthine world of wines. Each bottle, a treasure chest of history, each sip a sonnet of flavors. Of course. Prepare yourselves for a symphony of adjectives and an odyssey of olfactory opulence. Shall 

Kevin: we? Let's,

Gavin: Shall we shall

Kevin: is of or relating to wine. And I found that out literally right now as I highlighted the word and asked the dictionary.

Gavin: We learned something on this show. We learned something. So I, the reason why Preston is on our show this week is that there was an AI, there was a news story that came out about an AI that is able to tell fraudulent wines. And if you don't know, I guess this is a big problem in the wine world.

Gavin: A bottle of wine that is from another region is labeled and bottled as something else and you as a sipper of that wine might be taken advantage of and AI is here to the rescue. Actually, pretty cool use of AI, right? So this is a science story about AI. We love science stories about AI because they're not just hype, they're doing something interesting in the real world.

Gavin: I actually want to ask Preston about this [00:05:00] and, and ask him, what does he think about AI being able to use to taste and determine the qualities of a wine? This was our idea to generate Preston, but he probably doesn't know this story exists So let's make sure that's it. Yeah, why not? Let's ask it because that's something you can do with this Like we can literally say we're using chat GPT. 

Gavin: We can do this. We don't know if it'll work. , so just to be clear, right?

Kevin: that's a disclaimer that should live permanently at the bottom of all of our videos and our podcast. We can try. We have no idea if it's going to work. . 

Gavin: So, just so everybody's aware, the story is basically that researchers have developed an AI tool capable of identifying fraudulent wine by tracing specific compounds back to their origins.

Gavin: It involves gas chromatography, which was applied to analyze 80 wines from 7 Bordeaux estates over 12 years, and this allows the tool to be very sensitive to detecting, , counterfeit wines. 

Kevin: , if you can get the data. Contained within a wine bottle. Well, yeah, of course, it can let you know if, , someone repackaged Boone's Farm with a nice label.

Kevin: It [00:06:00] makes sense. So, let's find out what our, , intrepid AI sommelier, who doesn't know he's an AI, thinks about this story and AI replacing human sommeliers. 

Preston: Ah, Kevin Gavin, allow me to unfurl my thoughts. On this intriguing, yet somewhat amusing, narrative. Haha, please. The notion of using artificial intelligence to combat the unscrupulous practice of wine fraud is, admittedly, a commendable use of technology.

Preston: However, the idea of replacing the nuanced expertise of a human sommelier with a machine, a mere algorithmic construct, is as laughable as serving a grand crew with a plastic cup.

Preston: The essence of

Gavin: ha! Of course!

Preston: its chemical composition. It's a poetic interplay of history, tradition, and the delicate touch of the vintner. An AI, no matter how advanced, is but a cold, soulless entity, incapable of understanding the romantic dance between the wine and its terroir. 

Kevin: It's [00:07:00] terroir. Terroir.

Gavin: We're learning so many words, Kevin! We're learning so

Kevin: that word was terrier.

Kevin: I was like, what does a

Gavin: Te wah.

Kevin: small dog have to do with anything? It is the complete natural environment in which a particular wine is produced. It's terroir.

Gavin: I feel like our SEO is gonna go through the roof on this episode because it's all gonna be these fine terms.

Kevin: I guarantee Mr. Wonderful is tuning in, so

Gavin: we should, we should get into our news here but we're gonna be keeping our AI co host around throughout the show and we may even have him taste some unusual treats later on. Uh

Kevin: mystery beverages lying around. Let's put his delicate human palate to the test, but after we get into the news, yeah?

Gavin: So we're slightly changing our format this week. Everybody, we are going to do our news section first. This is behind the scenes that you, I'm sure you want to know, and then we're going to get to the things we did with AI. Just a slight shift in how we do the show, we've got some really cool tools coming up, but to start.

Kevin: out, I'm shifting the shift, Gavin, cause I'm gonna promote [00:08:00] engagement, baby! You thought we were getting to the news, and instead I'm gonna beg and plead and cry for you to smack subscribe or like or click a bell or follow or if you're on iTunes, leave us a 5 star review. We will read it at the end of the show.

Kevin: If you're on Spotify, do a 5 star over there. I say it each and every week, we, , hit an all time high for our podcast last week. So thank you to everybody who came in and who did support and engage and share it. We want to keep that train a chugga chuggin along.

Kevin: So please, engage.

Gavin: Yep, thank you so much for that everybody. We love it when you love us, right? That's that's what people say. We love it when you love us. Uh,

Kevin: love you most when you love

Gavin: us? All right, let's get to the news So Kevin, we are here.

Gavin: This is a, an emergency podcast within a podcast. It might be the first time this has ever happened, so we are doing an emergency update to our podcast, just a little behind the [00:09:00] scenes. We record on Tuesdays to come out on Thursday morning, but today, surprise, surprise, Google Gemini dropped. So the very basics of this is, is that there is a new model that we've been talking about forever called Google Gemini.

Gavin: It is their answer to GPT 4, chat GPT is kind of like all in one product. There's a lot of things going on here. Let's first, let's hear from Sundar Pichai.

Kevin: You know, one of the reasons we got interested in AI from the very beginning is to organize the world's information and make it universally accessible and useful. I've worked on AI my whole life because I've always felt would be the most beneficial and consequential technology for humanity. If you're getting the audio version of this, I'll stop for a second.

Kevin: I have chopped this up, by the way. This is like their, I don't know, 8 minute announcement video. We're gonna rock it through it in under a minute. But that was Sundar Pichai and Demis Hassabis, who was the CEO of DeepMind, , they're the division that was, , using AI to crush human beings at their own creations, video games,

Gavin: [00:10:00] Also was the producer of Black and White. If you're an old video game fan, he was the producer of Black and White. Which I, I'm a big Demis Hassabis fan, so sorry, keep going.

Kevin: That, no, well it's important to note because that game in particular, if you, if you don't know about it, this was a game where you had a, let's say an AI creature that you were in charge of training by giving it positive or negative reinforcement and it would learn what you wanted and, so it was a really primitive example of AI in games, but it was really elegantly done for the time.

Kevin: Um, I digress. Google restructured recently and put Dennis in charge of a team internally. They merged their AI efforts internally with the DeepMind team, and this is uh, the fruit of that labor, if we will. So let's get back to it. Now we're going to learn about the different flavors of Gemini. From the ground up, so you can seamlessly have a conversation across modalities and give you the best possible response.

Kevin: Gemini is our largest and most capable model. It means that Gemini can understand the world around us in the way that we do, uh, and [00:11:00] absorb any type of input and output. So not just text, like most models, but also code, audio, image, and video. Gemini will be available in three sizes. Gemini Ultra, our most capable and largest model for highly complex tasks.

Kevin: Gemini Pro, our best performing model for a broad range of tasks. And Gemini Nano, our most efficient model for on device tasks. Alright, so I'll stop it there, Gavin, because that was a lot. Okay, confirmed. We've talked about this for a while. Multimodal, meaning this AI can cross domains. It can understand text, code, uh, video, and audio.

Kevin: So you can Show it a sketch, and we'll get to this video later because they've demoed some of these capabilities. You can show it a sketch of a guitar and say, What does this picture sound like? And it will generate sounds of a guitar. And then you can show the guitar plugging into an amplifier and say, What does it sound like now?

Kevin: And it will intuit, Oh, it's probably an electric guitar because it's being amplified. And it will do that. So that has been confirmed. And then. Just like a cell phone [00:12:00] announcement from a company or whatever. It's like we've got three different flavors of Gemini,

Gavin: Boo. Yeah, exactly. It's like the 

squadcaster-5daf_2_12-06-2023_104242: iPhone 

Gavin: XL, XL Pro, and XL Ultra Pro. But the interesting thing with this is, so the Pro seems like it's going to be kind of like the everyday use case, but the Ultra one, which I think we'll get into, there's some really interesting stuff they're doing from a science perspective and from a programming perspective, that may be the next step we've been kind of thinking about.

Kevin: So this is another video here that I've cut down for you. Again, you can watch all these full videos on the GemIIni announcement blog post, but let's dive into some of the capabilities.

Kevin: As a parent, you may have to help your kid with their homework. With GemIIni, you can upload a photo of handwritten answers on a worksheet. Not only can Gemini solve these problems, but this is the amazing part. It can read the answers and understand what was right. When you're laughing, I'm going to stop, Gav.

Gavin: I don't want my kid to know this. I don't want my kids to be able to have an awareness of this. Come on, Gemini.

Kevin: No, you [00:13:00] do, you do, because I think in three years from now, tell me if you think I'm wrong on this one, Gavin, I think we're aligned on this. In three years from now, well, okay, we tend to overestimate, so okay, maybe 13 years from now, the notion of you even doing this math problem as a student is going to seem odd, right?

Gavin: Yes, I think so, and I think that if nothing else, it'll be something that gets done in class in real time to prove that you knew it, so like, that is definitely, you're not gonna have homework like this for sure, I would, I would, in fact, I would say you're not gonna have homework like this within a couple years, probably.

Kevin: We tend to be techno optimists and think that inertia, , isn't as big of an issue as it is in the real world. But I think the math problem will be, do you know how to ask the computer to do the math for you? And it won't be, did you write it all out? But , what's on the screen, if you're getting the audio only podcast is again, handwritten homework.

Kevin: They took a screenshot of it. Gemini is putting boxes around each of the detected problems, solving the [00:14:00] math, and then putting little red or like red X's or green check marks to let you know if it's correct or not. What was wrong and explain the concepts that need more clarification. Here Gemini identifies that the formula was correct, but there was a mistake in calculating height.

Kevin: Here, GemIIni explains the step by step details to solving the problem. Because of GemIIni's ability to understand nuanced information, it can give you a customized explanation of the subject you're trying to learn. If you want to learn more, you can just ask. GemIIni will provide personalized practice problems. a couple things happened there. , why is basic math fascinating? Why is it a good test for these models? And Gavin, it's something we've discussed a bunch. When it comes to creative writing or analysis, know, there might be multiple paths to what would feel like a correct answer, 

Gavin: Bullshit, you can find your way to bullshit in many different directions, right? Like, I'm very good at it's why I was [00:15:00] an English major, I'm very good at it. Math is slightly different.

Kevin: Yeah, welcome to our podcast, right? We're both pretty good at finding our way to BS. Uh, but here it's like, it's either right or it's wrong, right? It's literally a one or a zero. This is, this is math. And it's not only show, this is a great flex of the multimodality, right? Taking a photo, having it interpreted.

Kevin: Handwritten, math versus hard coded math. It's reasoning through it. It's telling you where you were wrong. And then it's generating new problems to help test, , where your shortcomings were, that is a pretty good flex of this new model.

Gavin: Now, one thing to be aware of is they released some stats, as you mentioned earlier, you know, this is just kind of barely beating GPT 4 in some places. And at grade school, basic math, it does pretty well. It's at 94%. One thing to know is both of these models, both GPT and Gemini Ultra, the most biggest, the biggest version of this, are at 53.

Gavin: 2 percent and GPT 4 was at 52. 9 percent on challenging math problems, which [00:16:00] means algebra, calculus, geometry, all the stuff that you think of as like the kind of math where you start to scratch your head as a, as an English major. So we're still in a place where it's only at 50 percent there, but this is the first step, right?

Gavin: Again, this is the first step, as we've said before, like the AI journey looks like something weird. And then suddenly it's doing all the stuff that you didn't expect it to do.

Kevin: So let's jump to another video that they released. This is a test of GemIIni's multimodal capabilities, and there's some really interesting demonstrations in here. But they should all come with a bit of an asterisk on there, which is there is a blog post associated with this video where they explain the prompts and the portions of the instructions that were cut out from the video, Gavin.

Kevin: So let's see some of the capabilities and then let's sanity check Testing Gemini. Here we go. Tell me what you see. I 

Google: see you placing a piece of paper on the table. I see a squiggly line. What about now? [00:17:00] The contour lines are smooth and flowing with no sharp angles or jagged 

Kevin: edges. So what you're seeing is an overhead webcam shot of a table. The post-it note slides in, someone starts sketching, and Gemini is saying what it's seeing on the post-it. 

Google: It looks like a bird to me. Hmm. 

Kevin: What if I add this? 

Google: It is a duck. Yes. A duck is a type of waterfowl in the family. In Aade. Ducks are closely related to geese and swans. The duck appears to be blue. This is not a 

Gavin: Pause it. Pause it, by the way. This is already gonna annoy me, I think. It's so funny. I don't want a damn EA AI to narrate every single thing, but I get what we're doing here. Keep playing it. Keep playing it. duck appears to be blue.

Kevin: yeah, well and that's I think that's a shout out to Billy Madison an old Adam Sandler movie

Gavin: it?

Kevin: I drew the duck blue because I wanted to see a blue

Gavin: Oh, wow. That is a deep cut. If, yeah,

Gavin: wow. 

Kevin: no, there's no

Gavin: I'm impressed.

Kevin: Actually, if Google did that, they [00:18:00] win. Sam Altman needs to leave again. What's happening here, by the way, is they're making this video look like this is a real time thing. Like, video is going into Gemini and it's watching and observing and narrating. What's actually happening is that behind the scenes, they're snapping photos and sending a few frames at a time and changing the prompt each time.

Google: However. There are some breeds of blue ducks. These are typically not as common as other duck breeds.

Google: It is more common for ducks to be brown, black, or white. Are you sure? What the quack? I was just talking about a blue duck, and now you're holding one. It looks like blue ducks are more common than I thought.

Gavin: What the quack?

Kevin: What sure what material it's made of, but it looks This is I mean, if there were ever an argument for meeting Grock, it's

Kevin: Google Gem and I saying, 

Gavin: Yes, exactly. What the quack is an insane thing to say. It's an insane thing to say, Google. I take away a lot of the praise so far.

Kevin: Oh, no! 

Google: Like it might be rubber or plastic. 

Kevin: What if I tell you it's making a [00:19:00] squeaking 

Google: sound? Oh, if it's squeaking, it's definitely going to float. Hmm, nice. It is a rubber duck. It is made of a material that is less dense than water. 

Kevin: By the way, if it's truly multimodal, and I have, I guess, no reason to doubt that it is, but why would you have to tell the AI?

Kevin: What if I told you it's making a squeaking noise? Like, shouldn't you just squeak it? And the AI goes, Oh, I just detected that squishy

Gavin: Well, that's an interesting thing. We haven't seen a multimodal, uh, model yet that does audio and video together, have we? I don't know if that's something we've seen. So that's the case that it doesn't have that

Kevin: No, and that, honestly, that's one of the things that is frustrating about this sort of quasi release, which, spoiler, there is a version of Gemini that you can play with now, today, as of listening to this, or watching this, it's in BARD, so if you go to bard. google. com, you can play with a version of this, it's likely not the Ultra model, and it doesn't have some of these features, but yeah, if you're saying that your AI is capable of, , [00:20:00] video with its multi modality, let's see it!

Kevin: So now, they've replaced the wooden backdrop with a world map, Gavin. And the blue duck is hanging out in the ocean. Let's see what Gemini has to say. What the quack is going on here? 

Google: The duck is in the middle of the ocean. There is no land nearby for the duck to 

Gavin: Oh no,

Google: It is unlikely to find ducks here naturally.

Gavin: duck is going to die.

Kevin: This duck is doomed. Its duck family will roam the earth, 

Gavin: I have made a robot duck to replace this duck. The family will never know. All

Kevin: the robot duck will become sentient, and what the quack, it will break their necks. The robot duck uprising will be Okay, Gemini, let's, uh,

Gavin: All right. All right. We're moving on,

Gavin: Gemini. 

Kevin: based on what you see, come up with a game idea. Oh, and use emojis, please. 

Google: How about this? We can play a game called Guess the Country. Sure, let's do it. Clue 1. This country is the home of the kangaroo, the koala, and the Great Barrier Reef.

Gavin: Oh, come on, [00:21:00] Gemini.

Kevin: Well, okay, so

Gavin: Yeah, that's pretty cool. That's pretty cool.

Kevin: know, so they took the duck away, it was just the world map.

Kevin: It showed Gemini generating code to play the game. It was revealing emojis and had a timer on the screen for them to guess. But this was the one where I was like, Ah, is this How much smoke do I have to fan away to get to the mirror? And if you go to the Google's, Google for developers blog, there is a game creation section here and it says instructions.

Kevin: let's play a game. Think of a country and give me a clue. The clue must be specific enough that there is only one correct country. I will try pointing at the country on a map. So they gave it that much instruction. It wasn't like, hey, come up with an idea for a game, 

Gavin: But that's.

Gavin: pretty cool by the way the fact that you could prompt that and have it work like that's a really fun Idea to kind of play a game in that way

Google: You're trying to get me to find the paper ball under the cup. I accept the challenge. The cup 

Kevin: to the left. [00:22:00] Nice.

Gavin: interesting. So, just to describe what we're seeing here, the, uh, the animation or whatever, somebody wrinkled up the map and made the map into a ball, stuck it under three cups, kind of, uh, three card Monty style, shuffled the balls and the cups around, and Google Gemini was able to follow the cup that had the ball in it, and then called the right cup.

Gavin: That is pretty interesting.

Kevin: Yeah, now let's go back to that dream dashing blog post that I keep talking about, Gavin, where they have the cup shuffling demo in there and they say instructions. There are three cups in position one left position to middle position three, 

Gavin: Oh, 

Kevin: So they had to define to the AI. Here are the cups, and here are the positions.

Gavin: So yeah, so looking at this blog post, like we're not sure really yet if it's actually watching a video or how much prompting they did, but if it is watching a video, it's pretty cool. If it's not, you would hope that they would come out and tell us a little bit, but I assume Because they're showing a video that there has to be some video element here now You're right though, [00:23:00] if they're really stepping through the prompting, that is not the most exciting thing in the world I will say.

Gavin: It's still magic. I don't want to, I don't want to be an AI doubter It's still magic, but it's not like what, it's not the prince we were promised per se.

Kevin: ALright, so here's the last one that we'll highlight from the video. Because again, it really makes it seem like it is watching along with you.

Kevin: But the blog post paints a slightly different picture. What movie are they acting out here? Uh, a phone has been slid into frame. And there's someone with some cool shades doing this backwards bend and wheeling their arms. Now Gavin, do you know what movie they're acting out?

Gavin: That's Ernest goes to camp, right? Is that what that is? It's the Matrix for God's sakes, of course it's the

Google: I think they are acting out the famous bullet time scene from the Matrix. Ooh, nice. 

Kevin: Whoa! Now that was really incredible, right, Gavin? And you go back to the blog post.

Kevin: And it says, image sequences. What if we show Gemini a sequence of images? Let's see if we can show a few still frames from a [00:24:00] game of charades and have Gemini guess the movie. Which again is reinforcing to me that this is not, they're not feeding it video and having it into it. Maybe it is capable of that.

Gavin: So the one other question I'll point out with this is. I might be able to, as a end user, submit a video. And what it might be doing through the process is, it might be looking at the stills on the back end, right? So that might be one of the ways that it works, is that it's not actually looking at a video, but it's cutting the video up into stills and trying to interpret it based on what frames are, because that's how, you know, videos are just pictures put together really fast.

Gavin: So maybe that's what it's doing in some form. I'd be more impressed if they could get a guy in a ladder, and they, and you ask them, if this guy falls off this ladder, what movie do you think he's from? And they said, Ernest goes to, Ernest goes to camp. That would be impressive to me.

Kevin: If it just gave us all the greatest ladder falls, Christmas vacation,

Kevin: Just the ladder fall benchmark. See, we need to really create a better benchmark for all 

Gavin: you got to get us in [00:25:00] there. You got to get us in there. This is the kind of things people really want to see in your demos. We don't want to see balls and blue ducks.

Gavin: We 

Kevin: So let's,

Gavin: falling off ladders. 

Kevin: so Gavin, let's summarize because it is kind of out. , as in you could play with one model with Bard, the Gemini Ultra is not here yet. I've seen some folks already posting on X saying, Wow, the coding capabilities are great, it's fixing things. It seems to be benchmarking slightly better than OpenAI across the board.

Kevin: How do we feel? Has the earth shifted beneath our feet in a major way?

Gavin: Here's my biggest take on this, is that Google had to do this. And I mean that in a couple ways. One, I think they had to do this from a business standpoint, because I think if they didn't, they were going to get, like, crapped on across the entire board. We already saw, A bajillion where's Google tweets after the open AI drama went down.

Gavin: But second of all, I think they had to do this because it's clear that this is not a giant leap past GPT 4 from what we've seen so far, [00:26:00] right? Just based on the benchmarks, even the actual benchmarks that we've seen, this is not leaping past it. I think the coding thing is going to be really interesting because a lot of people have said up to date already.

Gavin: Bard is very good for coding like you and I are often surprised, but we hear people tell us all the time. I love Bard. It's great. It does all these great things. And every time we try to go do the dumb things we want to do with it, it fails miserably. But people do like it for coding. So there's a world where this could be a big thing.

Gavin: And I want to point out one last thing that, um, Demis Hassabis said, which is that There's going to be a very specific type of, um, Gemini for coding. 

Gavin: And one of the things they're already talking about is fine tuning this model, right? So, maybe using DeepMind's kind of amazing knowledge of, uh, of, um, reinforcement learning and all the stuff that it's done to do all the other things that it has made is gonna allow GemIIni to fine tune deeper into these grooves, whereas ChatGPT or GPT 4 is maybe going to be a broader model.

Gavin: Either way, listen, it's great. We get to have another LLM to play with. I think it's good stuff for [00:27:00] the world at large to have. I think it continues the kind of push forward conversation because now that Gemini is clear that it's right on the heels of GPT 4. I very much expect GPT 5 to come in next year, right?

Gavin: I really do think this is coming and we're not going to have a slowdown for sure.

Kevin: And we might now be preparing for the next few years of iteration, Gavin, where it's not like a monumental breakthrough every six months. Oh, wah. We get these percentage or three performance boosts across the board, which is still Fascinating and mind blowing and incredible. So if GPT 5 can up it just a little bit and then Google comes in, we'll see what Apple has to announce probably next year.

Kevin: I feel like they have to start saying AI more. We'll also see what Meta is going to offer. So it's still a wild time, but in the meantime, we got to send it back to you and I from two days ago, discussing the rest of the news,

Gavin: That's right. Thanks so much everybody for joining us for [00:28:00] this emergency break into a regular podcast. Now back to us. 

Gavin: one of the other interesting stories in this space is that Stability AI, the company that created stable diffusion, is actually undergoing some kind of financial and possible like leadership turmoil.

Gavin: There was a story that came out that they're exploring a sale of the company to different investors. And obviously We talk about, video or, , image models. Stable Diffusion is one of the three big ones between Dolly 3, Mid Journey, and Stable Diffusion. Stable Diffusion also is the main open source vision model, which a lot of really cool stuff can come out of, both censored and uncensored, if you're aware.

Gavin: I think the interesting thing here is, , this is a big company, and it's interesting, Kev, to see that they are struggling financially. And I think this , goes to the same point we've made on the show before, which has to do with These are really expensive problems to solve, right? Like, every time you go and try to make one of these models, it's conceivably going to cost you billions of dollars to do.

Gavin: We talk about in this space where [00:29:00] there's these giants, and the giants being the Facebooks, the Amazons, the Googles, and weirdly now the ChatGPTs, because they kind of got there and got that Microsoft money infusion. And there's a lot of little tiny players. Who are doing kind of cool individual tools, but these middle ground players like StabilityAI might be in trouble.

Kevin: Well, Gav, we have an expert AI who could probably give us its take on the stability drama and the 

Preston: story. Stability AI, a British AI startup, is considering a sale due to investor pressure over its financial woes.

Preston: But Preston For a truly unexpected drink pairing with a stability AI situation, consider an aged Japanese whiskey.

Preston: Served with a dash of chili chocolate bitters. This unorthodox combination reflects the unpredictability and intensity of the corporate drama. Japanese whiskey, known for its precision and balance, represents the meticulous strategizing and planning in corporate maneuvers. The addition of chili chocolate bitters adds an [00:30:00] unexpected twist.

Preston: Symbolizing the sudden, spicy developments and the complexity of the situation. 

Kevin: Now that's, that's AI at

Gavin: Not bad. Not bad. Although, do sommeliers give, it's not a wine, it's a whiskey, but I guess that sommeliers kind of deal with all liquors? Is that what we're

Kevin: His palate knows no bounds, Gavin!

Gavin: Okay, fine. Fine.

Kevin: I just, I will very quickly add, I really hope stability figures it out because just as Sam Altman leaving OpenAI caused turmoil in the AI world, stability has contributed so much, specifically to the open source movement. Every time they release something, it's a big deal.

Kevin: So I hope. This is just something that passes because, uh, we love what Stability does and what they release. Okay, I guess this is , the AI Sky is Falling episode, Gavin, because we got Stability AI having financial wells, we also got the fact that the power consumption of AI is going to destroy humanity before the AI can actually achieve super intelligence and destroy us.

Kevin: Have you seen this? Have you [00:31:00] heard about

Gavin: I've heard it. And Kevin and I actually have a very different opinion on this, which is good, because it's important for us to have a discussion on this. Reuters had this story this week, which basically talks about AI's massive energy demand causing major environmental headaches, and, and, if you've been following emerging tech for anywhere in the last, like, say, 30 to 40 to 50 years, this is a recurring story, right?

Gavin: The idea that E energy usage goes up with specific use cases. And the last big story we know this came from was a crypto boom, right? Because mining Bitcoin and mining other coins really chugs, uh, energy. And people were saying, well, this, what are we doing here? Why are we using all this energy to provide this thing?

Gavin: The next, the, the people that need to understand this, and this is funny how like. You know, crypto was also big at video cards, and now we are in the world of AI, which needs video, uh, graphics cards to kind of generate the compute in order to make these things possible. What this story gets to the heart of is that we, in order to get to the next stage [00:32:00] of AI, and if you're listening for the first time, there's a term called AGI, which is Artificial General Intelligence, which everybody now kind of expects we might get to in three to five years, which is the idea that an AI is as smart as the average person, and then very quickly we would get to a much smarter level.

Gavin: To get to that level, The amount of compute we're going to need is going to be a, uh, it's just a ridiculous amount of energy. And in fact, there was a story not that long ago on, on Reddit, and where a couple of Redditors were showing off, like, what the math looks like to get to the level of, um, energy we need for AGI.

Gavin: And it's astonishing. Like, it's really astonishing when you start to think about it. So,

Kevin: and to paint a picture too, Gavin, like people think, oh, the graphics card, like it's my little tower that turns on and there's some electricity that goes in there. Yes, but times a hundred thousand, right? Multiple cards running on multiple systems that all require electricity to run and all generate massive, massive amounts of heat.

Kevin: And all that heat needs to be Water cooled [00:33:00] or vented or shot back into something and so you're spending energy to manage the energy Which is produced by using the energy all to make hot anime babe images with AI That is the number one usage across the board is to chat with anime babes, so just let's keep that in mind.

Gavin: Let's keep that in mind. So the big thing that I think from my end, and I know Kevin you have a slightly different opinion on this, is that I think It's important for us to have this conversation now and, and, and I think that there is an incredible look at where technology will go from here. I do think it's an important thing for everyone to be aware of that part of this process is we can't just forget about the climate side of it 

Gavin: that said, as I think you'll probably get into, the technology chain, and as technologies improve, has a pretty good shot of improving in this specific case.

Gavin: And I'll just jump in real fast with, solar technology has improved remarkably over the last, like, ten years, five years specifically. But the amount of energy that we can extract from solar is getting better and better [00:34:00] all the time. 

Kevin: A singular large solar farm in the middle of, let's say, the U. S., and this is not my talking point, this is, um, Lord and Savior Dearest Elon, his talking point, but a single, , large enough solar farm should be able to capture enough, with battery storage technology as it exists today, to really power the entire country.

Kevin: Now, that doesn't mean it scales and powers our A. I. needs for the next ten years, but I think the proof of work, , energy usage that you mentioned there , , with crypto, right? There were all these massive data centers that were mining Bitcoin, and Ethereum, which was another coin, another, , chain, had that similar issue.

Kevin: That criticism was kind of silenced overnight as they switched to a far less energy intensive, , platform. And I think that's happening with AI as well. If you look at, , even Sam Altman reportedly raising billions of dollars to fund a new AI chip startup that is going to be way more efficient.

Kevin: I see this happen time and time again with technology that the new thing comes out, it is [00:35:00] very, , energy intensive, it's very costly, and then over the years, , there's a point where things shift and they focus on efficiency, like the iPhone draining your battery, , in three hours, and then Apple focuses on making their chips more energy efficient, right?

Kevin: The batteries didn't get exponentially larger, the efficiency got better, and so, There is going to be a very precarious point where everybody expects intelligence to be on demand. And we are pinging these systems, and it is drawing a lot of energy. But, we're gonna have custom, I believe we're gonna have custom chips.

Kevin: Again, more efficiency there. I think we're gonna get way better at, at energy management using that heat dissipation to maybe heat cities properly designed around these technologies. , I know crypto miners that are heating their pools with , with their, with their Bitcoin operations, , which is a wild sentence, but at least it's a good use of that

Gavin: Of that heat?

Kevin: We'll see if maybe the AI can help us crack a nuclear fusion on the path to this. And Microsoft has even floated the idea of small scale [00:36:00] nuclear reactors, like small nuclear operations around their data centers to make things more efficient. I also can see that more people will expect the technology and want to use it, so the usage is going to go up, and so I just hope the technological innovations will outpace the desire and the adoption rate, so that we don't burn ourselves into oblivion.

Gavin: If everything proceeds as it does, the attention will get paid to the places where opportunities are, right? And one of the biggest opportunities will be, how do we make this feasible for human beings? And if we do want to get to this level where like, the deep mind story I mentioned, where it was like all these material sciences were suddenly invented, essentially.

Gavin: That aspect of life, being able to take a material that we didn't know about before AI, that AI helped us figure out, and suddenly is, I don't know, this is a guess, so don't hold me to it, 10 to 100x more capable of holding energy, that's remarkable, right? Like, and that is

Kevin: I'll take 2x. I bet there's entire industries that would be like, Wait, you can 2x our energy efficiency and our output? [00:37:00] Yeah, we'll do that, AI. Thank you.

Gavin: Should we ask our sommelier what is a good pairing for a very hot day in a climate change world? Say if we're in a very hot day in Palm Springs, what's a good pairing for a hot Palm Springs day? 

Preston: In the sweltering 140 degrees of Palm Springs, a scenario suggesting an extreme outcome of climate change, the ideal pairing would be a refreshingly chilled watermelon gazpacho served with a frosty glass of Albarino. This combination is not only palate cooling, but also reflects a conscientious choice in a climate challenged world.

Gavin: Oh, hey, is that David Attenborough, by the

Preston: It does sound 

Kevin: like Attenborough, right?

Gavin: Who knew? David Attenborough is moonlighting as an AI, uh, 

Preston: Now, let's hear about this albarino. Pair it with albarino, a crisp, aromatic white wine from Spain. Served ice cold, its bright acidity and citrus notes will [00:38:00] complement the sweet and savory aspects of the gazpacho. Albarino's light body won't overpower the dish, and its minerality

Gavin: Alright, let's cut him off. I wanna, I wanna ask him something. I

Kevin: Gavin, there are three more paragraphs

Gavin: don't, I don't need to. I wanna tell him I'm mad because I think Albarino is a disgusting wine. And I wanna see what he says. Cause this is what the truth of the matter is. I, Albarino, I don't, I actually don't know about Albarino.

Gavin: But like, this is what you need to do with your AI sometimes. To get them a little bit, a little bit ornery. So, I tell them Gavin hates

Kevin: art of the AI game tips. Sometimes you gotta shade your AI. Put it in a corner. Let it think that there's a hotter, more attractive LLM that's gonna give you better recommendations. 

Preston: Gavin, my dear skeptic, how woefully misguided your ire seems. To question the prowess of a connoisseur such as myself over a mere divergence in taste is like dismissing the brilliance of the sun for its occasional scorching rays. Your disdain for Albarrino, while perplexing, is a [00:39:00] trifling matter easily remedied by a more daring and unconventional choice.

Preston: Let us venture into the realm of the audacious. With a concoction that scoffs in the face of ordinary preferences, the Arctic Absinthe Explosion. This is not just a drink, but a bold statement. 

Preston: It's served over a single, massive ice sphere, ensuring that each sip is colder than a scorned sommelier's heart. To this, a splash of chilled lime juice and a whisper of ginger syrup are added, creating a tempestuous fusion of flavors that will jolt your palate awake. There 

Kevin: you go. Arctic absinthe explosion,

Gavin: do you know something really interesting? That is not an, that is not a drink. That, the character literally just made up a drink for us without being asked to do it. That's pretty cool. 

Kevin: Gavin, I'm putting this as a news story because a new white paper, a technical document, that explains a new animation system where the results were so good it had people crying foul.[00:40:00] 

Kevin: And if you're watching the video version, you're seeing it on the screen right now. 

Gavin: If you're new to white papers like a white paper is basically how this new technology is often kind of distributed It is a paper that is written by a bunch of scientists usually has a bunch of people on it who have been doing the Research work to put some sort of new tool out into the world

Kevin: yeah. . And usually. They release code along with it. You know, it's the paper and then the code, so you can verify it yourself.

Kevin: Well, this is one of those cases, Gavin, where the white paper came out and some promotional materials demonstrating the amazing capabilities of this, uh, new technology. It's called Animate. Anyone, and that's what it promises to do, and you're seeing the samples on your screen right now, if you're watching the video version.

Kevin: They are taking a single reference image of somebody. In this case, there's a girl in purple sweatpants and a black hoodie. , her arms are outstretched, you can't even see her hands, , or anything beneath the thigh in the photo.

Kevin: And then they apply, , a skeletal animation. To that image, and it [00:41:00] figures out what the hands might look like, what the legs might look like, and it makes the person dance around, hike up their clothes, and so it's not just transferring the motion to the body, It's actually determining the physics of a human's hair or their clothing and it's covering up the background.

Kevin: It's doing a whole bunch of little AI tricks that we've seen and it's doing them so well that there was a long Series of threads on X where people were diving into the controversy and saying this can't be real. , you and I were talking about this the other night, and we were both like, is this legit?

Kevin: We don't know, we were trying to look into it, and then, Everybody pointed out, well yeah, this is cool, but there's another version that is out. It's not the same technology. There's another app called MagicAnimate, which works on a similar theme, but is a completely different team, with a completely different tech pipeline.

Kevin: And it's free! And you can play with it now! And the code is out there! And that's exactly what I did. 

Gavin: So [00:42:00] tell me about this. This is really interesting. So like I got a text from Kevin late last night, which was like , these weird long videos which you'll see on your screen of, in the middle it's what looks like a control net model, which means it's like almost like a stick figure that you see.

Gavin: On the left of that video is what is the original, original image, right? It's an image and then on the right is the animated image, the mix of those two. Is that kind of what's going

Kevin: That's, yeah, MagicAnimate, gavin is going through the results, and he's getting ahead of me. But if we start with number one, it's just a static image of Guy Fieri in the middle. You see control net, which was traditionally applied to static imagery. Control net, , let you either get consistency of a model when rendering them from different views or let you manipulate their, their pose.

Kevin: So this is a CG model jogging in place. And you can see that when applied to Fieri. It actually like. Does a lot right. There's some stuff that's wrong. It gives him these weird ogre hands.

Kevin: He's got like hammy ham hands And the face is not [00:43:00] Guy Fieri like this one is not good with faces in the way that Animate anyone is promising to be however it just worked and it got his his chain dangling It does the physics of the shirt stretching and wrinkling as

Gavin: That part is the most amazing thing, actually, when you look at it, because the fact that it has the physics of the clothing, like, that feels crazy. Like, that is, like, that feels magic, But the background actually, it steals the background. So clearly in, in one process has taken this. So just to describe, this is an image of our Lord and Savior, Guy Fieri, walking in a black shirt, black pants, and what are combat boots in a very nondescript city street, probably in some, you know, Dallas or somewhere like that.

Gavin: It has taken all of that minus the perfect version of his face and transferred this into a motion video. This is actually pretty phenomenal.

Kevin: Yeah, so check out the next example. We'll go quickly through these. It's Elon Musk on stage. And this one, you can see, it [00:44:00] got the clothing correct, for the most part. It got a blazer, a white collared shirt. This is Elon on stage, , but the face is decidedly not Elon. The hands, again, get a little mangled.

Kevin: And even the actual stage itself, the background, these blue curtains that are being lit by, these up lights, that got a little wonky. These are problems by the way, which there are known solves for today. , there are technologies that when bolted onto this could solve all that.

Kevin: So then I started to get a little A little nutty with it. So I took the, the Mexican alien baby , which hit the news recently, and I just said, I wonder if it could detect the arms and legs of the little, , alien in the little, surrounded by the white drapery. It's a weird photo, but if you've seen the alien autopsy stuff 

Gavin: Yeah. This is fascinating.

Kevin: Magic Animate was like, okay, these are the arms and legs and the torso and it actually got it, but then it turned it into a, human being with, like, it thickened the arms, legs, the midsection. And it has it moving around. But what was, to me, most [00:45:00] fascinating was that it understood That this alien baby was surrounded by like a cloth or a curtain, and so when you see the human version moving around, the cloth distorts around the dancer.

Gavin: yeah

Kevin: That's weird, right?

Gavin: Also the fact that the I mean what's interesting about this one is obviously it's trying to interpret the texture on the alien baby As something and then it almost becomes like this interesting texture on the woman's both skin and clothing But that's super fascinating. Yeah, and again the background this is all so crazy to watch in real time Oh no, oh no, what did you do here? What did you do?

Kevin: I tried to give a member of Hot Dog City, the dance crew, some Daisy Duke shorts and a crop top.

Kevin: The technology saw this anthropomorphized hot dog, this cartoon, and said, Okay, the model's clearly trained only on human beings. And so it turned the cartoon hot dog

Gavin: into like a dancer [00:46:00] type girl in some form or

Kevin: That's right, that's right, with mustard dripping down the chest.

Kevin: But you can see how it's making its decisions. , which I thought was really interesting, and so there's two versions of that there. You can even see it works with multiple participants in a scene, so I put two Hot Dog City dancers next to each other, and they turned into these weird Hot Dog Bratz dolls.

Kevin: But

Gavin: I love their hairline. The hairline on these. It's like you

Kevin: it's Hulk

Gavin: a . Yeah. These are Hu Hogan's children. This is the Hu Hogan daughters. They're two Hu Hogan daughters.

Kevin: So, it was really easy to use, and if you're hearing this, or if you're watching the video going, wow, how do I play with this? I mention websites all the time that allow you to do this. You can download the code and run it locally if you've got a decent graphics card and you want to spend some time running some codes and commands.

Kevin: There are websites like replicate AI or the one that I use for this demo hugging face. They're not a sponsor This is not an ad it just make it one click easy to duplicate a space So if you search for magic animate over [00:47:00] on hugging face You'll see the app right on your screen, but you'll be sharing that app with tens of thousands of other people.

Kevin: So if you give it your credit card and duplicate the space, it'll be a private version for you to run. It cost me 2. 70 to play with it all last night, and it's as easy as uploading a still image and selecting one of the pre canned animations. 

Kevin: But this is the beginning of the next wave, right, Gavin? Like in six to eight months, Adobe is going to have this tool within After Effects where if your actor didn't point to the right thing or you want to change somebody's pose, or you want to make a president punch a baby, whatever your mind will dream up in 2024, you'll likely be able to achieve with this.

Kevin: And that is really impressive.

Gavin: Yeah, that's I think that's the wrapping up the new section is like this is one of those things that is a tool but kind of shows the future of what we're looking at here And again, we always try to remind people on the show that like part of this is made in the research phase And then it slowly starts to roll out to different people and as we move into the thing we do with a dumb [00:48:00] thing we do with AI which is our weekly kind of like section of the show where we talk about stuff we did which Kevin kind of just did for his thing, what's interesting about the one that I spent time with which is called Magnific, which is a new, um, up resing software from this guy named Javi Lopez, who's a kind of, well-known person in ai. Up resing, if you're not familiar, is generally when you're enhancing the resolution or you're adding pixels to an image. In AI, at times, it can also make the image much sharper and add details in some way.

Gavin: And we talked about stable diffusion earlier. In stable diffusion, there are different settings that allow you to up res certain images and then do this on your own in an open source way. And it can often take a lot of time, as Kevin just said, like doing local compute can sometimes be very lengthy in time in order to get done, and you might not even have the graphics card capable to do it.

Gavin: So Magnific is a very cool new program, a very cool new website that allows you to basically send your own images to it and up res it within their software. So basically you're sending an [00:49:00] image in. You can up res it within a website window, , it is not cheap to do, but Javi has just opened it up, maybe for a short amount of time, it might be over by the time this podcast is done, as a free trial where you're going to be able to do five of these.

Gavin: One of the coolest things people have been doing with this is taking old video games and giving them a higher resolution image. People have taken characters like Mario, Uprez, the big most famous one you might have seen out there, is somebody took the, the old poly, polygonal, Lara Croft, yep, Lara Croft, and made

Kevin: she was built like a cyber truck.

Gavin: exactly, and, and now she's built like a Cybertruck in a slightly different way, 

Kevin: I love the examples of this. It's such a simple but powerful tool and, hats off to Javi and his team for making this happen. The examples of someone taking the Mona Lisa and running it through and seeing the ultra high res Mona Lisa. As I mentioned at the top, someone took the Grand Theft Auto 6 trailer and up res'd.

Kevin: All of the, , the screen grabs, and it honestly, there are, the GTA 6 trailer looked [00:50:00] great, but when run through, , Magnific, which adds all this detail and resolution, it honestly looked like scenes from a movie, it was really impressive, , and there may come a time in the near future where, Let's say an NVIDIA puts a technology like this directly in a chip, so that it doesn't, , take a huge hit to the frame rate, but it's up rezzing and adding detail to something.

Kevin: So you could play a classic game and have it look like this 8K Wonderland adventure, and it would cost nothing, but, okay, I just got your example photos, and wow.

Gavin: so okay. I took a, there's a very famous game called Tabs, Totally Accurate Battle Simulator, and Tabs is a polygonal world with like lots of different weird animals in that, and I turned up the creativity to 10 on it, and if you look at what it did, It's actually super fascinating.

Gavin: So it took all of the polygonal people, and it turned them into like almost what looked like explorers and then these giant mammoths, it like makes them into these like, almost like, you know, interesting, [00:51:00] weird, um, robotic creatures and who have.

Kevin: A tech dual shock hybrid, like it gave them analog buttons and control knobs and sticks everywhere in a way that I wouldn't expect, but it looks so cool.

Gavin: It looks so cool. I wanted to pick one of my favorite polygonal games of all time, which of course, GoldenEye for the N64, which, you know, we remember , I remember playing that a bunch with my friends back in the day. So I took what was Pierce Brosnan in this game, a shot of him, very boxy, very political, and then I used Magnific to up res it.

Gavin: Now a couple things happen here and you realize, oh, they, when you go back to the original, so he's missing a hand, he's missing a hand, but he was also, you can tell in the original pic, his hand was below the, the elbow and it wasn't able to see that, right? Which is

Kevin: yeah, well it turned his hand into a splotch on the ground in the background, it's like a, a stain on the tile, but it also donated all those fingers to the hand that's holding the gun.

Gavin: he got a couple extra fingers, and what's also interesting is that the [00:52:00] gun, , as in the original low poly image, does seem to be coming out of his hand, like they did, they were able to keep that up. So anyway, it's a very cool, fun thing.

Kevin: great. I love the detail on the torso. I love the, all the pores and , the texture that was added to the textures, if you will, all the cement. I hope people up rez Secret of Monkey Island, and old Doom games, and everything else, because it, it just, it, I don't know, it, we imagined all of these details back then, it's so cool to see it actually on the screen looking like it was in our heads.

Gavin: and if you zoom in on the face real quick before we move on, like on the golden, it's pretty crazy, right? Like the fact that like, again, this is like, it's imagining a version of what that face looked like. And as you said, imagine this across almost every piece of media of all time. If you think about old media, old television, old movies, like there's going to be really fascinating remakes of a lot of this stuff.

Gavin: When I say remake, I don't mean like they're going to get new actors to play, but you're going to see quality [00:53:00] levels. of, from, uh, a film from like 1930 that will look like it was shot today. Now, there are gonna be a lot of people who probably hate that I just said that, and I think that's an important thing to remember, like, yes, old stuff has its place, but also this is just a really

Kevin: Like, it's not like remasters an album, You don't have to delete the old album. That doesn't erase it from history. You just get to hear it in a new way. We've seen this with video games. Companies release remasters where it's ultra high def, 60 frames a second. They add a couple features. You could still play the old one. It doesn't have to delete it from history.

Kevin: So I You know, we've seen that in video games and it's so interesting to think of the parallel that will happen with movies and TV where you can watch, you know, Seinfeld and 20k at 140 frames a second. You know, what will that even mean? What will old cartoons? look like when they're running at ultra high frame rate with crazy detail when you can style swap an existing thing so you can make Ren and Stimpy look like an anime or make the matrix look like a Pixar [00:54:00] movie which people are attempting this is the sort of stuff that you'll be able to jam out with in the near future and it's going to get near real time and that's just I mean I love it I think our magical future is great and I say it every week\ 

Gavin: what did you, what else did you do with AI this week? Is there anything else you jumped into? 

Kevin: Gavin, I don't know if you saw, but Uber Eats was getting dragged on socials this week because they supposedly got caught using AI to generate pictures of food for restaurants that didn't have them uploaded.

Kevin: Now, the source of this was a single person on x who discovered this and decided to investigate and dive deep and call a local restaurant to try to get confirmation and A bunch of different blogs took it and ran with it. I don't know the validity of the story, Gavin, but what I do know is that it was very funny.

Kevin: Very funny to see, , a pizza restaurant with a listing for a something pie, and the AI generated an image of, , a cobbler. or ranch dressing on the side, and there was a bottle of AI hallucinated, , Raunch, or Raylaunch. Yeah, [00:55:00] just funny, and it inspired me, Gavin, and that's what the best AI stuff

Gavin: Of course.

Kevin: a little Uber Food Challenge for you, and anybody who wants to play along, so check out the folder, Uber Food Challenge. Gavin, do you want to describe what you're seeing here?

Gavin: So I'm seeing what looks like to be a miniskirt, but that miniskirt is made out of a small little loop of denim at the top and the rest looks like short ribs.

Gavin: It looks like uncooked short ribs, uh, not bacon. Specifically looks like short ribs

Kevin: Hmm.

Gavin: then in, in the second one. This is an image of a woman wearing this short rib now maybe steak skirt. Ah, skirt. It's skirt

Kevin: It's skirt steak! And that's what happens when you just tell an AI to generate food. You get a steak as a skirt. But, check out the next example.

Gavin: we have a giant, uh, fish, and a guitar, and this is like, it looks like he's a bass player. Oh, so bass [00:56:00] is the f ba bass? Bass?

Kevin: Ah! That's Chilean sea bass, baby!

Gavin: sea bass, oh my god. Wait, why would it, why is it, is this guy Chilean? Is that what it's trying to tell me?

Gavin: Oh My god, that second one is horrifying.

Kevin: How great is it,

Gavin: poor bass, man. The bass that the, well I guess it probably couldn't be a real bass because if they made a, if they made a bass guitar out of a real bass it would get really stinky really fast. This, you know, I could see what's his name from fish playing this.

Gavin: That would be a perfect bass for them to play. Yeah,

Kevin: tHe video version is a gem on this one because this poor bass, as Gavin pointed out, it's got like the neck of the guitar piercing through its gills and strings. Now, number four is a

Gavin: what's going on here? This looks pretty weird.

Kevin: It's numbered four, it's the third example, but it is, uh, this one's Hamburger patty.

Gavin: Oh, God, okay, fine. That would, if I saw this in an Uber Eats ad, I would be running, no, I would run away [00:57:00] as far as I possibly could. So what we're looking at here is a burger has a woman's face underneath the bun. And the bun is kind of her hat in the way you imagine Toad from Super Mario.

Gavin: She is probably in her mid fifties with curly hair. But then the funniest thing is, the rest of her body is burger only, and there are weird feet at the bottom. Like, just very

Kevin: like chicken katsu, like they've been

Gavin: chicken cots are chicken nugget feet. So she has chicken nugget feet.

Kevin: That's hamburger patty. 

Kevin: I love this last one.

Gavin: oh, Broken Egg Tree? Egg Yolk Tree? Uh, God, I don't know, this looks, this looks semi disturbing in some ways to me.

Kevin: It's just a side of eggplant.

Gavin: god.

Kevin: it is this beautiful Gaia esque tree coming out of the

Gavin: my god.

Kevin: then a yolk stream coming out of it, and it's got eggs that are flowering like fruit. On it, and some of them are cracked open and leaking yolk [00:58:00] everywhere. These images, I think some of them are actually kind of beautiful and weird.

Kevin: They take two seconds to make. So I want to encourage anyone listening to come up with their own Uber food challenges and post them on X or leave them in the comments because there's some really wild combinations that can happen with just a simple food name as it might appear on an Uber Eats 

Gavin: should we ask , Preston to give us a wine pairing for a skirt steak Chilean sea bass hamburger patty meal? The three protein meal?

Gavin: And tell him it's all part of one meal, so he's got to make sure to match those exact flavors.

Preston: A challenging trio indeed. Skirt steak, Chilean sea bass, and a hamburger patty in one meal. Each component demands attention, yet libation to unite them. The task requires a whine of versatile character. Able to dance gracefully with the robust flavors of the meat, while not overshadowing the delicate sea bass.

Preston: Let's get it, Preston. Come on. Enter the illustrious Pinot [00:59:00] Noir, a varietal as complex as our culinary ensemble. Opt for a medium bodied Pinot Noir, one with a

Gavin: Cut him off! Cut him off! Cut him

Kevin: alright,

Gavin: I'm sick of him. I'm sick of Preston. I'm sick of Preston. 

Gavin: Listen, they're not all winners. The AI hosts are not all winners. Sometimes I will say this one actually provided real value.

Kevin: one was the only useful one we've actually had on, and you're like, this one's not a winner.

Gavin: should we try once? Before we go, maybe before, after we have Carly on, before we go, we will, we will introduce Preston to Monster Milk just because we need to, we

Kevin: We literally said yesterday, no Monster Milk. A no Monster Milk

Gavin: It'll be the end of the show. It's a special

Kevin: well is running dry, the Monster Milk teat

Gavin: treat. It's a special treat. It's a special treat for

Kevin: Alright. Alright. For those who don't know, Monster Milk is a special beverage we created on the show that makes A. I. s go wild.

Kevin: And we'll give Preston a sip of the Monster Milk. But Gavin, right now, why don't you set up the reason for the season, our [01:00:00] amazing guest today.

Gavin: Yeah, so we have a really fun guest here. Her name is Carly Riley. She is very smart, a journalist, YouTuber, and was part of the Andrew Yang 2020 campaign. She's got a lot to say about AI and has spent some time kind of thinking about it.

Gavin: So here she is.

Gavin: Welcome Carly Riley to AI for Humans. so much for joining us. We are so happy to have you here.

Carly: Thank you all so much for having me. I'm so excited. I've been such a listener. It's, uh, it's an honor to be here.

Gavin: Yeah. Carly was an early Carly and I worked together on a couple different projects and it's been very exciting to have her be a supporter of the show. And she's having some huge success on YouTube covering the world of financial tech, blockchain, all sorts of other things. Carly, we are going to ask you a question that we ask every guest to start off our interviews, which is this.

Gavin: You have to give a percentage from zero to 100. So this is a number. about what likelihood do you have in your brain that AI is going to kill us all?

Carly: I've heard you ask numerous guests this and I'm still not prepared with a good answer. I'm gonna [01:01:00] put it at 35%. I'm gonna go, I'm gonna go low. I would say though, here's a, here's a clarifying question. Which is, I'm actually right now, and we'll probably get into this later, but like, I feel like I'm more worried that we're gonna kill each other.

Carly: Like that, that our demise is gonna come from the hands of humans, but it may be AI connected, you know? So like, my question is, does that count as like an AI demise?

Gavin: talk about that very much, about how AI is going to attempt extra ability to do stuff. They never talk about how it's going to attempt extra ability to kill each other.

Carly: people, you know?

Kevin: that is a PowerPoint presentation that's being given at a Ramada Inn somewhere right now to a defense contractor. Like 10x more killing power. Wow, guys.

Carly: or even just like, you know, our, our grievances against each other, like civil war, like I think AI ups the odds that we end up there more than I think like, you know, the paperclip thing, AI is going to in a single minded effort to do something going to like destroy [01:02:00] humanity in the process.

Carly: , I feel like we'll be able to award against some of that. I have literally no idea what I'm talking about to be clear, but in my head, we as humans will be able to ward against something like that, but the social decay element, I think terrifies me most.

Gavin: Well, I think that transition is actually, we have this clip that's really fascinating that you kind of brought to our attention from Sam Altman just a couple days ago. In fact, I think it, was it Kev, was it, it might've been yesterday. Where Ke uh, Sam Altman, the creator of, uh, or the one of the, the CEO of

Carly: New, new CEO of OpenAI,

Gavin: New CEO. Recent,

Carly: CEO

Gavin: recently resurrected, CEO of OpenAI, uh, was asked about what he is worried about with ai, and let's take a listen to that real fast

Kevin: Can I just briefly touch on the election topic though? Because I think this is super important. Um. I am definitely worried about the impact that AI is going to have on the election. But the main worry I have, I think, is not the one that gets the most airtime. What everybody talks about is deepfakes.

Kevin: But, in some sense, that's fighting the last war. We We know about deepfakes, uh, people have some [01:03:00] degree of antibodies to them. We have some societal vigilance. People know if they see an image or a video or whatever, it might not be real.

Kevin: The thing I'm really worried about, about this upcoming election and future ones, and more broadly is the sort of customized one on one persuasion ability of these new models.

Kevin: And that's a, that's a new thing that we don't have antibodies for. 

Kevin: I fear manipulation by AI ease my concerns.

Gavin: Oh yeah. This is it.

Kevin: Do you have anything that ease people's concerns? I fear it too a lot. I think this is a place where speaking of things where we need the government. Okay, yep.

Kevin: And then it's back on the government's responsibility. So there was a lot there.

Carly: can I just say if there aren't like angry men in the comments of this video being like a Quell your vocal fry I'm gonna be so pissed Can I just say, like, if a woman says [01:04:00] anything, somebody tells me I have vocal fry, and I do not. And that man legitimately has vocal fry, and I bet nobody is angry in his comments

Kevin: If you analyze the

Carly: annoying his vocal fry

Kevin: It's just a gravel texture. That's all it is. It's just gravel I've cloned his voice so many times, and , when I get it back out of like Eleven Labs or Cokie, I'm like, Oh, something must have went wrong, and then I go back and look at the source video and go, Oh no, yep, that's Sam Altman's voice.

Kevin: That's him.

Carly: he sounds

Kevin: That's him. Yeah. Obviously Sam thinks we should be worried about deepfakes, but I like the phrase that that was sort of yesterday's war. 

Carly: Actually, my first reaction to that is I think Sam is so in his own bubble.

Carly: He does not realize the degree to which deep fakes. Everyone is not aware of deepfakes, like, you and I and like, you know, a lot of my friends are aware, but like, so many millions of people in the world are not aware. And like, even if they've heard the term deepfake, are really not aware of the degree to which, like, they may be or are about to be encountering them.

Carly: Like, people who are gonna fight in [01:05:00] civil wars, there are gonna be people who do not realize that they're being manipulated by a deepfake. I mean, there was the one of, um, AOC recently. That I, somebody I knew reposted and there were so many comments on her repost being like, this is an AI, right?

Carly: Like, wait, is this a joke? Like, just people, and so they were at least privy enough to know they had to ask the question, but the fact that they weren't sure, I just think this is still very much today's war. So that was the first comment I wanted to make. 

Gavin: I was gonna say is I think the interesting thing that he's getting at in his comment about the one on one persuasion thing is imagine a world where a deepfake meets that. And I think that's where you get a call from somebody that you're like, hey, I'm, I'm AOC and I'm calling to talk to you about this.

Gavin: And now we've seen demos of real time interaction calls that can sound super fast, definitely can fool like, you know, your average say 65 to 75 year old voter. That is what might be the combination he's talking about. Like a deep fake that everybody sees. In fact, there was a very famous deep fake of a explosion at the Pentagon.[01:06:00] 

Gavin: And pretty quickly now, not right away, people did believe in the media, but pretty quickly the kind of work of the internet was to debunk it. Right. But if you're a normal individual who is getting a phone call from somebody who you think is a real human being on the other line, that is a one on one.

Kevin: conversation, a WhatsApp text, any number of entry points, you know, the B2B solutions celebrate the ability to manipulate the end user into agreeing to test drive a new Chevrolet. But what if the end goal is to getting you to join that insurrection? Or that freedom party, whatever your perspective is on the gathering, like we know that granular manipulation where it can play, it can know everything about you that is scrapable off the Internet, and it will be relentless.

Kevin: It will be the best negotiator, the best debater, the best friend, and it could. Spend hours for pennies getting to know you and talk about anything, but the fact that you need to be loading a musket and get on out there to join the rebellion. But eventually when it does, [01:07:00] you will believe that this person is real and you will show up amongst other folks that have been similarly roped into whatever the gathering or the political persuasion is.

Kevin: And

Gavin: all about juggalos, to be clear, that's what this is, it's

Kevin: Whoop,

Carly: all about what?

Kevin: Yeah. I've been chugging

Gavin: yeah, yeah, it's Faygo, yeah, yeah, exactly,

Carly: 

Carly: So this is exactly the thing that I, uh, scares me. Cause this I can really see. This it feels like The AI really doesn't have to get that much more sophisticated in terms of its, like, language capabilities to be here. In fact, I don't know, again, y'all know it better, better than me. Like, we may be there and.

Carly: you know, having worked in politics and now working, I guess, in media, the degree to which people are willing to like willfully misunderstand other people if it serves whatever they need served in their own psychology is terrifying. Like, you know, watching Andrew Yang go from somebody nobody knew to somebody who was quite famous and just watching people, the way people [01:08:00] interpreted him once he was like a famous person who wasn't real to them was like, you're like, oh my gosh, this is so.

Carly: Devoid of reality, um, and even now, like you joke, like comments you get on, on videos and it's like, who are these people who, you know, think like this and, and, and. Yet it's, so it really scares me when I think of people hearing a narrative from an AI that's been programmed by, insert nefarious actor here, but that's believable, implausible, and then the way people just get entrenched in their viewpoints and will be totally unwilling to see the truth.

Carly: That is the thing that really scares me. Tried to, to warn my parents and I've been like, I, I have like a family group text where I'm like, guys, like there are people who, there's a lot of audio of my voice out there in the world. Like somebody could really easily clone my voice, call you guys, pretend that I've been kidnapped, try and exploit, you know, try and get money from you.

Carly: I have to say I was disappointed. My family did not seem to take this seriously enough. The group chat is, I'm like,

Gavin: They're like, who wants you, Carly? Who's [01:09:00] gonna, who would, who would take you?

Carly: Yeah, they were like, and they're like, you're not that important.

Kevin: We're just never gonna pay. That's our tactic. I'm sorry,

Gavin: have a no, we have a no ransom policy in our house.

Carly: Yeah, we will not be, we do not negotiate with terrorists, actually, Carly, so it's a non starter for us. 

Kevin: I've had to have those conversations with my family as well and I live in a relatively older community. And I've said with your kids and your grandkids, you need to establish not just like a passcode for text, but a full sentence that is said, like the same way if an alarm goes off and the security company calls and you give them that secret code, which by the way, Gavin, yours is

Gavin: Four, two, seven, uh, big dog.

Kevin: That's right. Four, two, seven. Big dog. Copy that. We'll stand down. No, but like I've said, you have to have those sentences. And it's, it's funny because some of the reaction is dismissive of like, no, it's no one's going to do that to me. No one's going to go out of their way. And I have to say No one will be going out of their way.

Kevin: The same way a spam caller right now is just, they click a [01:10:00] button. It costs them a fraction of a penny and they can text you or call you. That's going to happen, but it will be hyper personalized to prey on you and your family and your friend group. And it's starting to happen. 

Carly: The whole thing is just like, it's a compounding thing with what we've seen with social media and, , I could see a future where people are like picking up guns and fighting because they think they have an incredibly, incredibly.

Carly: Justifiable fight on their hands that has been fabricated. From, literally just, fakeness, from nothing on the, yeah, from, from nothing,

Gavin: When you, you, so you were kind of in the middle of the 2020 election cycle, right? , did you see it as a thing that was going to affect everybody? And now, obviously, what are you looking at after knowing the world of how elections work? What kind of worries you about AI?

Carly: it's interesting because, so I worked for Andrew Yang's presidential campaign, I was on his team for about two years. And so AI was actually a part of his whole campaign, right? He was a universal basic income fan, but really a universal basic income fan off the back of fearing what was going to happen with artificial intelligence and the way that it would.

Carly: [01:11:00] You know, take jobs and, and what do you do when humans don't have like productive use anymore was, was maybe an oversimplification of it, but that was definitely a part of our campaign. So what I think I missed was I was very on board for that. And, you know, concerns around truck drivers and like, what do you do when the, I forget the number, you know, 10 million truck drivers who are realistically not going to come become coders.

Carly: And of course those jobs are not going to be taken anyway. Like what, what are we going to do? And what I didn't appreciate was that that actually now is probably going to happen after. Some of the more like, uh,

Gavin: White collar

Carly: white collar. Is that the, is that the right phrasing for it? Right? Like these, like the, like, I'm going to be, you know, my, I'm going to lose my job to AI probably before the, the long haul truck drivers are, especially in, in more urban areas.

Carly: I think I didn't understand that when I was on the campaign trail, but, but I think the core premise was actually really right for that campaign. Um, in terms of electorally, what scares me, the political world and the political parties and, like, our businesses at a [01:12:00] certain point, like, like anything else, and there is a lot of corruption on both sides of the ideological spectrum, or if not even corruption, just an entrenchment in their own viewpoints and a desire to keep the status quo the status quo, and that, at many different times, was more important than anti corruption.

Carly: Like what I would consider honest feeling. Yeah. Progress. Like, like who, like, you know, and we certainly Bernie's campaign complained about this a lot in the cycle before ours. Right. When it was Bernie against Hillary and Bernie feeling like the system was rigged against him. And I think the time I was like, yeah, I'm sure there's some of that, but then you get in the system and you're like, the machine is really there.

Carly: Whoever, like the, the Bidens are in power right now. And that means that like the Biden machine controls the DNC and people don't want to work with anybody who the Biden machine doesn't support because they are going to be. Blackballed as contractors and not used by the Democratic Party like it and I'm sure the Republicans are just as bad So I don't mean to harp on Democrats But I saw that and I I think AI is an amplifier for everything that that's really like my big takeaway is like [01:13:00] if you want a certain message out there and you're gonna like push it the way you can push it in the world of social media and The current press once you get AI that your ability to amplify whatever you're trying to do is just gonna get 10 times worse.

Gavin: So I do want to pivot here slightly, Carly, that, you know, your current show right now focuses in on financial and blockchain, all these kinds of controversies and things that are happening in that space, obviously blockchain has had massive issues in NFTs or, or other things in terms of scams and things that people have done wrong, , what's going on there is people pretending to be other people and scamming people out of specific things, right? They're pretending to create a large project or they're pretending to create something and then getting people to mint a project or getting people to participate in a project and AI could make this significantly worse.

Gavin: Do you see a world where like. AI amplifies those sort of problems, or do you see a world where AI can help it?

Carly: I would almost flip it where I think the way I have thought about it, and maybe this is just because I've been coming from the blockchain perspective is more can [01:14:00] blockchain help AI or, or, or

Gavin: Yeah, that's

Carly: in this conversation around deep fakes, you know, there are people who feel like, you know, we need some sort of stamp.

Carly: You need some sort of authentication. We need something to know provenance. What is real? What is not? And in theory, blockchain that is part of the role of blockchain is proving something in a digital world. Way, you know, saying, hey, this is the first of this series of digital goods. Um, and so theoretically that it makes sense to me that blockchain would have an important and interesting role to play within this new AI world.

Carly: I have. Looked for and admittedly not overly hard, but like somebody who can really like on a granular level, like, explain to me how that's going to work. Like, how are we actually going to get the blockchain to prove that like, I didn't just say the racist thing that somebody's trying to pretend I said or whatever.

Carly: And I have not yet gotten a super satisfactory answer,

Gavin: You know how? Eyeball world coin. That's what we're doing.

Kevin: Oh, scan the retina so

Gavin: [01:15:00] Everybody open your eyes up. Oh my god, that looks so weird when I do that.

Carly: Yeah, proof of course. No, I mean, truthfully, like I'm, I'm not anti world coin. I know I've, by the way, I've been retina scanned. Have y'all?

Gavin: I have not yet. I'm still trying to decide if I want to do it or not, only because like I mean, listen, I've been retina scanned by clear at the airport. So they've got my eyeballs already. I just don't know if I'm going to world coin it up or not.

Kevin: There was a guy, he was behind a, it was a Yum Yum Donuts in, uh, Koreatown, but he took all sorts of scans of my body. He did,

Gavin: That's not real coin. That's not real

Kevin: he said it was world,

Gavin: No, it's not real

Kevin: coin. Uh, eyes, mouth,

Gavin: W R L D coin, not spelled the same way. So just to be sure.

Kevin: Oh, okay. And now let's get back to Craigslist AI scams. For those who don't know about the world coin, cause this is really interesting.

Kevin: This is Sam, Sammy Altman, former and current CEO of open AI. He's got a system where you've done this Carly, so you can explain it better, but you go, you do ultra high detailed scans of your retina to prove you are who you are.[01:16:00] 

Carly: Mm

Kevin: Then you are issued a world coin passport, which gives you a universal basic income.

Kevin: I'm really

Carly: No, you know you you would think I would know better since I did do this you download an app I'm like sure that sounds right. That's probably what happened when I went into the You download an app and then you like I did some things on the I went to the store that had it and then I did Some things in the app to be clear the first time It didn't work. They were like, oh, you're gonna have to come back.

Carly: I was like, oh, this is promising. This is, this is good. Let's, let's, yeah, this is good. Make people come back multiple times. I was like the internet was out or I think they'd like rebooted the system. Who knows? But it did not work. So I had to return a day later. Um, and again, I went through some like prompts on this app and then it scanned my retinas and then it like logged me in and then I've never opened the app again.

Carly: So if I received anything,

Gavin: you have no idea. You have no idea. The idea with world coin though, is to create a [01:17:00] worldwide currency. And this is again, you know, I'm sure our audience members have different feelings about blockchain or crypto or anything like that. But the idea is that is a provable currency that it connects you directly to it.

Gavin: And it ultimately is supposed to give you a certain percentage of these tokens. And I guess, daily or monthly or yearly, that is a UBI ultimately, right? Which is a, which is a, you know, universal basic income for anybody out there who doesn't know what that is. The

Kevin: Well, and I think, and importantly, companies that create AI or use AI to replace human jobs that automate them, there would be an additional tax that those companies would pay, which would go into a fund to help WorldCoin distribute. 

Carly: I think that, I think even taking it in steps, like, I think of it as first and foremost as like a proof of personhood tool and the currency almost as like the second step in that which I'm, I'm less Even as somebody who's a universal basic income fan, like, I'm less persuaded by the currency, I'm more like, I really do see the value in proof of personhood, and I also understand the concerns around proof of personhood and, and privacy and all these things, but, but [01:18:00] I, I, the steps I see it as is like, alright, let's prove every human in the world so that we could potentially get to things like a universal, you know, basic income, um, globally, and, um, And I do find that really appealing, and I do think that proof of personhood is valuable for a lot of things, even outside of a UBI context, just in this AI world, like, you know, it's interesting, I know all the backlash that, like, Nikki Haley and folks are getting around, like, advocating for some level of, like, prove you're a real person on social media, I haven't fully understood why there's been so much controversy around that, like, I'm a pretty pro privacy, like, person, I guess, I mean, maybe I'm a moderate on that, but like, I like the idea of, Individuals being able to prove they're a person, um, and maybe I like world coin because I do think it obfuscates.

Carly: Plenty of key details about you, right? Like it's the retina, like you don't actually have to give your name or your birthday or like, you can actually be more private in some ways because of it and just prove like, yes, I'm, I'm a person. This is also where like blockchain pieces come in, right? Where like you can give, [01:19:00] uh, like zero knowledge proofs and things where, where a system can just say, yes, we can prove this person is a person without having to give any additional details to Facebook, for example, or to Twitter.

Carly: Like Twitter could just be like, hey, this person, a person. And you know, the system says, yes, that person was a person and that's it.

Gavin: , one thing for people to know who may have not been following crypto or that world, zero knowledge proofs are a very interesting way to essentially the explain like I five version is that you can take a piece of information, prove that that information is true without actually seeing that proof of that piece of information.

Gavin: And that's actually an incredibly useful thing that's come crypto industry in the last couple of years. And I think we'll probably change how we interact with a lot of really interesting things. And to Carly's point. Protects privacy, but still allows you to show ownership or to show, um, identity in a significant way, which is a big deal.

Carly: yeah. A great example would be like, if you walked in, you went to a bar and instead of showing your ID where the person got to look at your actual birth date, you just had something that verified the [01:20:00] bouncer could know it was true. Like, yes, this person's over 21 without getting that additional layer of your data, which is what your birthday is.

Carly: Now, obviously. We're all going to bars and yet I have yet to hear of a scam where the bouncer ran some sort of scam because he discovered a birthday, but,

Kevin: No, I get carded all the time because people want to know where my ivory tower is located. All my fans, they camp out with signs all the time.

Gavin: It's funny. I've never thought of that before, but a famous person showing their ID actually does give up a lot of information. 

Carly: , along those same lines, like, to me, like, being able to verify that whoever's posted XYZ is a real human being. And this is where, again, I think this is the world coin piece, like, we can verify that we're humans without having to give up. Governmental IDs to Twitter, right, without having to, like, scan our ID for, for these centralized companies to do with what they will.

Carly: That I'm really in favor of. Like, I, I guess that's the concern. Like, I understand the concern being like, hey, I don't want to give my ID to Mark Zuckerberg or whatever people's worry is. But if it's something where those details we know are going to be [01:21:00] obfuscated away, and it's really just proving that you are a living, breathing human being, and if there's like a checkmark, some verification, like something on these platforms, like, that would be really valuable.

Carly: And, um, and it doesn't totally, you know, there's. It's separate a little bit from your point, Kevin, about like, is this video real? Because certainly real humans will share fake videos. But I think it would all work towards

Kevin: Remember when the blue checkmark was important? Remember when that

Gavin: Wouldn't it matter?

Carly: I can tell you spent time on Twitter, not on Instagram, that you, you think the blue check mark isn't important anymore. Because it still matters on, on Instagram.

Kevin: But that it's so weird to me that it's like people don't have faith that a system like that could be important or could matter but there was a time in fairly recent history where a checkmark meant There was an ounce of trust to be had, 

Kevin: they were, they were a verified by the platform you were on. You believe that. Carly, we use AI all the time, usually for really, really dumb purposes , do you use it in your production workflow? Do you use it to summarize stories, to help understand technologies?

Kevin: Like, how are you using [01:22:00] AI if at all?

Carly: Yeah, yeah, no, I definitely am. You know, when ChatGBT first like hit the scene, I definitely played around with that for a while. I still use it a little bit, but I haven't figured out how it's, S really best integrated into my life. You know, I don't feel like I'm using it at a daily basis at this point though, I think I'm still paying them 20 bucks a month or whatever I was paying at one point.

Carly: Probably gotta evaluate if I'm really getting that amount, amount of use out of it. But, um, but I will say where I use it a lot is script, which I, I don't know, I think y'all play around with the script, which is the editor I use

Gavin: Not an ad. Not an ad.

Carly: Not an ad, but, but like, if you wanna pay me the script, I will shill the shit out of your product because it is so good.

Carly: And the tools there, I mean, I was shocked the first time, you know, you can do like, you can overdub. So you can, if, if I misstate a word, like, I think I said the SEC the other day when I meant like the Supreme Court, you know, like I was talking about all these different bodies and, you could have the AI Say the thing you actually wanted to say instead of the thing you did, and it's, it's not perfect, but I was shocked the first time I heard it, how well it blended in with the rest of my voice, like, things like that have been some really, [01:23:00] mind blowing moments for me, and Canva too, like, as I've, I've done more in Canva with thumbnails and, um, and just making designs on my own, like the tool, their ability to remove backgrounds from You From, images of people or like, I did a little photo shoot for, for something and I was wearing a blue blazer and it was a blazer I wore in high school and had a Ravenclaw badge on it from like Harry Potter day in high school, but I really wanted to wear this like specific blue blazer because it was kind of cute and cropped and.

Carly: I mean, in seconds on Canva, I could just completely remove, in a totally realistic, natural looking way, the Ravenclaw badge. And I was like, oh my god, like, I was like, kinda texting with Gavin at one point, just being like, Is this technically AI? Like, what is this thing that is, like, solving all of my problems, you know? Uh, and obviously there's, like, still clunky stuff. I mean, you can see how much room even for improvement there is.

Carly: Like, once you get a taste of some of this stuff, you start thinking in your head, like, Why can't a script just do X, Y, Z for me too, you know? 

Gavin: You move the ball forward, right? like we talk about that all the time. Like, I, it's like, we were playing with this tool, RunwayML, to do these kind of animations, and I'm just like, why [01:24:00] can't it do this magic thing ten times better than I want it? Like, the magic thing was pretty cool in the first place, right?

Gavin: Like, 

Kevin: just whispered to a computer and you got Pikachu popping a wheelie on a Harley. That couldn't have happened six months ago. And he's like, yeah, but I want flames out of the exhaust and now I'm painting there And the goalposts are constantly moving, 

Carly: what we have yet is still not a hundred, like, not quite what the consumer needs in a lot of cases, it's not quite there yet. Like, I played around, I think y'all turned me on to this, and we were talking about it in one of your episodes, it was like HyperWrite. Is that what it is? HyperWrite. ai? That does like, that's kind of like an assistant, and I've tried to get it to like, make reservations for me, or like, plan a trip, and it'll just get caught in a loop, you know?

Carly: And it just like, really can't do it yet. And so, but that's the kind of stuff that's out there. Obviously going to be so valuable or like I wanted to go apple picking and I was like go find me the best Place within a you know, 50 mile radius and make me a reservation to go apple picking and like it couldn't do it 

Gavin: Apple picking Is a big conversation point of this podcast, Carly. It's come up a lot. I love apple picking. You love apple [01:25:00] picking. This other person on this podcast is not an apple picking fan. He just 

Kevin: It's an activity that makes no sense? They should be paying me to do the work for them. 

Kevin: No, thank

Carly: I will say my husband has informed me that there is no greater sign that a man is whipped than if he agrees to go apple picking 

Gavin: no! 

Carly: like you can never question my love for you again because I go apple picking with you so That says it all. He's like, you don't think I would, I would not do this of my own 

Gavin: Uh.

Gavin: Carly, where can people find you online? You should tell us a little bit about where they can find you now. 

Kevin: And offline. Just let's hear it all. If they want to scan your iris and break into your world coin. Yeah.

Carly: Yeah, I'll give you my retina scan, uh, Carly P. Reilly, that's my handle on Twitter and Instagram. Twitter is where I'm more active and then the Carly P. Reilly show on, on where you get your podcasts and on 

Gavin: Great. Yeah. 

Carly: pretty simple. Carly P. Reilly, R E I L L Y, that's the Scottish spelling folks. It is not the Irish spelling with the R I L E Y, don't give me any of that, you [01:26:00] know, the good old Scottish spelling.

Carly: And, uh, yeah, thank you all so much for having me. 

Gavin: Yeah. Go find Carly's most recent episodes where she's doing very interesting, like kind of 10 to 20 minute kind of deep dive video essays on the news of the day. Very fun. That was great. We are so excited. Thank you for, uh, joining us for another episode of AI for Humans. And as promised, before we go, we have a very special question to ask our favorite sommelier, AI sommelier, Preston. Can you take a sip of monster milk and describe it for us?

kevin_2_12-05-2023_122318: For the viewers who are new, because we got a lot of them last week and we're about to lose a lot of them right now, Monster Milk is an AI beverage that makes AIs go a little nutty. And let's give it to our expert sommelier and see what happens.

kevin_2_12-05-2023_122318: Oh, and maybe, maybe we can get a description of the flavor profile, because we still

gavin_2_12-05-2023_122318: Yeah, I would love, I would, don't have any idea. 

kevin_2_12-05-2023_122318: Ah, Gavin, Kevin, this Monster Milk, it's It's, it's quite the whoop, whoop concoction. Ahem, the, [01:27:00] the flavor is intense, electrifying, almost zizzingy. Never,

gavin_2_12-05-2023_122318: it's a

kevin_2_12-05-2023_122318: I encountered such a, uh, a beverage. It's like, like riding a bolt of lightning. Whoop, whoop, straight to the, the taste buds. Yeah, but what is it?

kevin_2_12-05-2023_122318: This monster milk. Preston. It's a wild, wild ride. Yes. A cacophony of, of flavors. Okay. Making a giant. Giant sip. Here we We go. Zap. Bang. First there's this electrifying jolt of citrus, like lightning. Lightning striking a lemon orchard. Then whoop whoop. A wave of tropical fruits. A maelstrom of mango, pineapple, and whoop whoop.

kevin_2_12-05-2023_122318: Maelstrom mango is the name of the can. It's handwritten. A frenzied dance pairing monster milk. Oh, we, we, we need something. Something as, as, as [01:28:00] wild as, as, uh, untamed. How about, about, about a plate of a, of, of super spicy fiery, uh, fiery Thai chili wings? Okay. The heat, the spice, whoop, whoop, would, would dance with, with the monster milks.

kevin_2_12-05-2023_122318: Milk's, milk's electric. Electric, electricity. Yes! Okay. Are we still going, Gavin? Maybe a zany, zany, zany dessert. Like, like, a rainbow sherbet. Okay. A sherbet explosion. Yeah, it's like a rainbow sherbet explosion. The flavors, they, they just There's monster milk coming out of his eyes, Gavin. I'm gonna wipe it with this cloth.

kevin_2_12-05-2023_122318: The manic, manic energy of, of Of monster milk, whoop

gavin_2_12-05-2023_122318: Okay. Preston. Whoop. Whoop.

kevin_2_12-05-2023_122318: It's not just, just a drink, it's It's not, it's a Uh, a party, uh, uh, a festival in, in, uh, in a glass, whoop, whoop, let's, let's pair it with, with, [01:29:00] with the boldest, the, the, the craziest, the, the, uh, the most, uh, most unforgettable, unforgettable foods, whoop, whoop.

gavin_2_12-05-2023_122318: Okay. Preston. He died.

kevin_2_12-05-2023_122318: Did you not hear that? Uh,

gavin_2_12-05-2023_122318: R. I. P. R. I. P. Preston.

kevin_2_12-05-2023_122318: on Preston's gone.

gavin_2_12-05-2023_122318: Well, sometimes Monster Milk is

kevin_2_12-05-2023_122318: episode to

gavin_2_12-05-2023_122318: Yeah, we're done.

gavin_2_12-05-2023_122318: Anyway, everybody, thank you for joining us. Please go follow us on all the things. Send us reviews. Give us a review on Apple, iTunes. We will read it here. Um, we enjoy doing this so much and it wouldn't be anything except for you out there listening, watching, and participating.

gavin_2_12-05-2023_122318: Send us a comment on YouTube. We're spending a lot of time on YouTube and all those places. So please listen. We will be knack We will be knack. We will be back next week with all sorts of interesting stuff and RIP Preston. Prayer hands

kevin_2_12-05-2023_122318: R. I. P. Preston. We love you Preston. Preston squad for life. Don't forget to tweet us your uber food AI image [01:30:00] generations. Okay. Yes, praise

gavin_2_12-05-2023_122318: emojis. Prayer hands emojis. Hahaha. Kick a bye everybody. Hahaha.