AI Transforming Creativity: Grimes' AI Voice, Bark Voice Tool, AI Dungeons & Dragons, and Wonder Dynamics GFX

AI FOR HUMANS EP003 FINAL MASTER
===

Kevin: [00:00:00] Um, RRRRrrrrrrrrrrrr roads Where are we? Uh, Where we're going. We, 

Gavin: uh, we, We don't need roads. 

Gavin: hey everybody. Welcome, welcome to AI for Humans. Kevin before we jump in, you wanna tell us what we're gonna be looking at today? Welcome. 

Kevin: Yeah. Oh, we got a huge table of contents. My very human friends. We've got Dunson and that's 

Gavin: t Ooc, table of hold contents, 

Kevin: we got Dungeons and Dragons, but a campaign that plays itself. We were gonna have some fun with some broken . Audio. , we might have stolen a voice. There might be a very famous artist that's giving you permission to steal their voice. , 

Kevin: and I guess we're gonna do a sneak peek cuz you got early access to the promise of Hollywood quality, special effects. Yeah. In your pocket as simple as a click and a drag basically. And we're gonna walk people through some wonder. 

Gavin: Yeah. I think interesting. You'll see a little theme in this [00:01:00] overall episode, which would be a little bit about how tell s storytelling is gonna change with AI tools.

Gavin: And I think if you're interested in that, it's, there's a lot to get into here. So before we get started, Kevin, how are you? It's been another week of craziness. What's going on? 

Kevin: Personally, pretty terrible. Gavin. Lots of damage to the RV and to the truck. As I look out the window now in beautiful Alamogordo, New Mexico where they did a missile test this morning.

Kevin: So no White Sands for me. Oh 

Gavin: my God, you really, if you have Kevin is like going through this kind of trip from hell. He's been traveling across country in a very cool way with an RV and spouse and they're doing really interesting stuff, but they've had a number of mishaps. And now to have a missile explosion would be an amazing ending to this journey.

Gavin: If 

Kevin: that were the reason the entire, rear section of my truck cab is shattered glass, that would be a cool excuse. Yeah. But just Jack knifing while trying to exit a hip camp that wasn't properly labeled. Not as cool. Not as, however, In the AI world, Gavin all is awesome and amazing. [00:02:00] And you mentioned AI storytelling.

Kevin: We're gonna tell some stories today using AI that are pretty terrible stories. Yes. 

Gavin: What about you? Yes, , I had a great week. I think the one thing that I'll say, , I do wanna really shout out everybody who has reached out to me about like how much they've enjoyed this show. And one of the things I keep hearing again and again is we love that the fact that this show is trying to break down the craziness and make it more digestible in different kind of mors for people.

Gavin: And I think one of the things we wanna talk about today, and there's been a big I'm in the Writer's Guild, the WGA in Hollywood, and there's a big thing going on right now where there might be a strike in about a week. And one of the big conversation points, not the largest is about streaming, but one of the largest ones is about ai, right?

Gavin: And talking about AI writers and AI creativity. And I wanna get into that. There's a lot of interesting stuff happening today. I do wanna shout out my mother-in-law also as a listener now, and as a big fan. So I wanna tell everybody, like this is an all ages show. And you should definitely make sure that you spread the word from teenagers to people in their sixties, seventies, eighties.

Gavin: I wanna find out who our oldest listener is at some [00:03:00] point. I'd love to get somebody in the cent, cent. Oh Cent. Centen Century Club. The Century 

Kevin: Club. You want a se, you want a cent? You want a half a cent? A half? 

Gavin: I wanna half about a half horse, half Freeman listener, whoever. The first one of those is, please send us a voicemail 

Kevin: at our website.

Kevin: And I wanna hear your hooves clopping with excitement when you set it in. And also, let's get the youngest listener. If there's someone that has Coco Melon on in the background, absolutely slap the applesauce packet out of their tiny hands and say, listen, child, your entire life is going to be dictated largely by these ais.

Kevin: So watch these two old men yap about 

Gavin: it. And then hit the subscribe button. Little Baby's, that's where we're going. Please 

Kevin: do please follow sub please. Babies for sub. I don't even know what to promote anymore, but AI for humans.show is our website. That's where you can go. You can leave us voicemails.

Kevin: You can listen to the podcast, you can get the links to the video version. It's all there. AI for humans.show. Gavin, did you do anything dumb with 

Gavin: AI this week? So I did a lot dumb with AI this week as per my [00:04:00] par, my . Bi normal world. , at one point I almost came up with an idea to create an entirely new NFL draft with ai and I actually got to the point where the Python code was being spit out.

Gavin: I . Create the AI metaverse draft and then I, it got broken and I got stopped. So that was the one like pathway I spent a couple hours down that I didn't do. I was hoping amazing. I could get like NFL drafts on different planets. I'm sure that would be interesting. If anybody in our audience wants to make that please 

Kevin: I would enjoy intergalactic draft would be great.

Kevin: Like baseball for the entire Galaxy. That's exactly right. That's exactly right. I'm a super fan of that. 

Gavin: So the one thing I did do, and this is just yesterday and I think we're gonna get into this cuz it was such a small thing at first and then I realized what the possibility is of a really good friend.

Gavin: Who goes by the name, poof. He's anonymous on the internet, but he is a very, very smart developer thinker. He's one of my favorite people that I've met internet wide. And he developed a really interesting system that is a storytelling system using the basic roles of Dungeons and Dragons, but we're gonna get into that later.

Gavin: Originally I was like, this is the dumb thing of the week. And then I realized, I started thinking a lot about more, it's [00:05:00] a much bigger thing than that. So we're gonna hold onto it. Kev. Love it. What did you do this week that 

Kevin: was really dumb? Other than the Jack Knife. I played with Bark for a minute.

Kevin: It's brand new, and it's completely free. , , but they released, uh, basically a text to speech a program that is incredibly powerful. You can clone voices. It's got like a hundred something plus voices built in. It can do sound effects and basic music.

Kevin: And it's, we're getting to the point where it's like, oh, yet another one. Uh, because that's, there's so many tools. There's tortoise and co QE or ko, and we've been using 11 labs a lot to clone voices. That's what gave us Mr. Yeast last week, and even a fan favorite gash. Mm-hmm. But this tool already out of the box is .

Kevin: Incredibly capable, very powerful. It seems very robust, high quality voices. And it's free. You can run it locally on your own machine, and it churns out stuff pretty quick. And, , another user, made a one click install. So if you don't know how to do any of this stuff, you can search, , bark one, [00:06:00] click install, safely install it on your computer.

Kevin: Another user made it so that you can stitch together multiple pieces of audio, and he added a mode called Confused Travolta, 

Gavin: We should explain that Real fast. Yeah, go ahead. Yeah no. Do it. A confused Travolta GIF is from Pulp Fiction, and it's when John Travolta walks into a room and is looking around trying to figure out where he is at.

Gavin: So this is really fascinating thing. So somebody added this mode so that it would be like watching the AI try to figure out what's going on, right? , 

yeah. 

Kevin: It basically hallucinates beyond your prompt and sometimes in the middle of it. , , I took some famous movie quotes and pushed them through bark.

Kevin: Now some voices are better than others. Some have background noise, some have like crickets, chirping so you never know what you're gonna get. You can choose a voice, but I didn't think that'd be fun. So here's a sample of the very first thing. Run through nice and clean. No Travolta mode whatsoever.

Kevin: May the force be with you Nice enough. Audio sounds 

Gavin: a little bit like a Star Wars, Joe Biden maybe be with you. 

Kevin: That's fair. Yeah. A [00:07:00] little force Biden. Yeah., now here's where Bark gets really good. Like , we've all heard text to speech that can make you sound like , a very, professional narrator.

Kevin: They give you an nice voice gravitas. But the moment you try an um, or an uh, or a like, or a stammer, very human pauses and repetition, a lot of 'em fall apart and they sound robotic, not bark. Uh, May um, 

Gavin: maybe uh, force. Maybe um, force be with you? Yeah. 

Gavin: Also Yoda. 

Kevin: Okay. Oh, also Yoda. Also Yoda.

Kevin: Good. Good little thing. Let's we have some other quotes . here. Mm mm

Gavin: May the force uh, may uh, may the force uh, may a force be 

Kevin: spit it out. That's 

Gavin: it. That was the whole thing. Is stam like it's stammered to voice is what we've got. Stammer to voice. 

Kevin: The point is you can write with [00:08:00] dashes. Ellipses, yes. Ums repeating letters and it's taking it and it's interpreted. Hey, mama. 

Gavin: Uh, Always said life. Uh, Life was like a box of chocolates. You're, you never, you never know what you're gonna get. It's so 

Kevin: how so it starts to hallucinate off into 

Gavin: the end.

Gavin: Sure. Okay. So the interesting thing about this to me is one of the things we've talked a lot about is, and we're gonna get into this later about when you're creating,, things that are meant to be read or be human-like, whether that's scripted content or unscripted content, which, when AI will be able to spit out their own things, this is going to make them feel much more real.

Gavin: Like the idea that if you had an Alexa that said they asked a question, the Alexa was like, oh, let me think about, yeah, God, it would be annoying. Maybe, but it would actually be much more real. So I think this is like a, an interesting step in the direction towards realness for sure.

Gavin: Uh, we're,

Kevin: We're going, we uh, we, we don't need [00:09:00] roads. If you're making motion, people are stealing. What happened? Hallucinating? What was that? That was it hallucinating? Was that was Confused Travolta. That was confused Travolta at the end. Yeah. We don't need roads. End of it. If you're making motion, people are stealing.

Kevin: If you're making motion, people are stealing. 

Gavin: I think we have a new slug for this episode. When in doubt 

Kevin: No. Here's the last one. Last one. Same quote. Um, RRRRrrrrrrrrrrrr roads Where are we? Uh, Where we're going. We, 

Gavin: uh, we, We don't need roads. 

Kevin: Rhode. . And then that was . Just a weird random. Yeah. It starts to go into a theme song, I think. I think it's the friends' theme song at the end there. 

Gavin: One thing to know about Bar, which is also very cool, is that two things that can do.

Gavin: One, it can add sound effects, so you can make it sound like somebody's On a phone from 1960, which is really cool. But you can also change, um,, languages really easily. [00:10:00] Like it's got a multi-language support. Yes. So you can do a thing where somebody starts speaking in English and then switches the Spanish, or you can do a whole thing in Chinese.

Gavin: This is like the tower of Babbel, like the Babbel fish, from sci-fi. It's like pretty soon, you know, everybody already has Google translated in their phone and they get, hold it up. But there's a world where this is like going to completely eliminate the need for translators. This is like a real thing. And it's free. That's a 

Kevin: really big deal. It's a huge deal. And when we talk about, a worldwide audience, we talk about, , multicultural support.

Kevin: It's not when you say, oh, it has support for different languages, you can take a voice and have it speak in a different language. But it retains all of the qualities of yes, that performer's voice in their performance. So imagine localizing a TV show with some some real time lip flap dubbing or anime in an area or explainer.

Kevin: Videos. Videos and how-tos you can have one narrator. You like their voice. Great. Now they're speaking every language you want. And to your point, this is out now, it's free. You can play with it. 

Kevin: Which is 

Gavin: amazing cuz then it's, kind of like text to, it's like, like using Chachi pt, right? Like [00:11:00] it'll answer stuff, it could write things, it could come up things and that might spark something entirely different. I 

Kevin: think it's an oracle because I spent uh, well it can become characters. I spent a lot of time watching succession this weekend.

Kevin: Did you are up 

Gavin: on this? Absolutely. I'm a hundred percent up on it. I love succession. 

Kevin: I guess we should say spoiler alert, for those who haven't seen it, you might wanna jump ahead like 30, 45 seconds or so, like knowing us 15 minutes or so. Cuz we could talk forever.

Gavin: I'm very caught up on it. And in fact, this weekend there was a very big scene, if you haven't seen it again, please spoil a alert.

Gavin: But there's a very big scene between Shiv and Matson, the character that is trying to buy,, the company waste our re reco. And there was some things said that might have been said that we didn't see on camera, right? But then afterwards there was a big reveal that they had been working together and there were some things happening.

Gavin: So we actually might have, we might have some secret audio that was recorded in the scene that they cut from the episode. Let's roll that clip. 

Quite a day, isn't it? Did you catch the news about Tucker [00:12:00] Carlson and Aaron Rodgers? Oh, I did. I won't shed a tear for Tucker. He's the embodiment of a dumpster fire. As for Rogers, I didn't see that coming. Indeed, just goes to show that in this game of life, anything can change in an instant. Speaking of games, my brothers seem to be playing their own Kendall's like a weather vein in a hurricane, and Roman is as useful as a chocolate teapot.

Well, sometimes those who appear to be on opposing sides can find common ground. For a mutual benefit life, a meeting of mines. You mean a delicate dance, if you will, where everyone thinks they're leading? Precisely. It's about understanding the board and knowing when to make the right moves. To the art of the game, the surprises.

Life throws our way and embracing the ever-changing landscape like Tucker's overdue exit and Roger's unexpected trade to the art of the game in the unexpected turns it takes. And to Roman, who's about as useful as a screen door on a submarine. And to Kendall, whose rap performance was as mesmerizing as a giraffe attempting ballet and zero gravity.

And to Roman, whose approach to business is like watching a corgi, trying to [00:13:00] solve a Rubik's cube with its paws. And to Kendall, who dreams of being a media mogul, but ends up looking more like Tucker Carlson trying to navigate a vegan food festival. That's a bridge too far. Lucas.

Kevin: Okay. And that is 

Gavin: How is that possible? They knew about Tucker Carlson's firing. They knew about Aaron Rodgers. Amazing. 

Kevin: Wow. Anyway, boy, I am sorry if that makes it into the show. 

Gavin: I'll, it's going to just for 

Kevin: this reason, it's, it has to, yeah.

Kevin: It has to. . The chocolate teapot line actually got me. I hadn't heard that one before. Yes. Yeah, that wasn't bar hallucinating. That was G P T four doing its best. Yeah. To actually write the scene between 

Gavin: these two. Yeah.

Gavin: So I think a one thing to, going . Back to the Hollywood writers, all that stuff, please be aware this is not going to take your jobs. This is a long ways away. I spent a long time trying to get G P T four with some real prompting work on trying to make sure that it could get the voices of these characters really had a struggled with it.

Gavin: But we, what we wanted to do is [00:14:00] show again, what's possible at this world. 

Kevin: With something like bark, you will be able to give, , almost like a HTML style code write-ups where you can say someone whispers or laughs or has an abruption term play it's stage direction. Yeah, absolutely. Yeah, absolutely. 

Gavin: I think it's time we move on because you know what time it is, Kevin, it's time for the news,

Gavin: okay. This week, as always, crazy stuff happens and it continually happens faster and faster it feels like every week. I think we wanna start with a story that's a follow up to last week's episode. One of the things we talked about last week was a lot of the AI, Drake story that was going around.

Gavin: Just a very quick recap. A person named Ghost Writer, I think it was 9 66, created a fake Drake track that went viral. It's something that people sped on much for the week. There's lots of people talking about how big of a deal it was, but one of the interesting that came out of this is Kevin and I were saying it was to be a really good opportunity for somebody to come out and take ownership and be able to say, as an artist, I think this is cool, this is something to do, [00:15:00] and somebody did this week, and that person is Grimes.

Gavin: And I think this is to me really awesome and I'm a big fan of Grimes . Her first album is amazing. She's very kind of weird and experimental artist. She's played around a lot with her own identity and different things like that, but. I really appreciate the fact that somebody came out and what she said specifically, she tweeted this, she said to the extent of I think this is very cool, and I'm more than happy to share 50% of whatever prophets come out of a song if somebody wants to use my voice in their music.

Gavin: So that basically she's giving up half the rights to the song by letting people use her identity, her digital virtual identity. The interesting thing I also thought I saw people, some people point out is there's also a world of that where she's keeping 50% because there is a really important part of this that she owns, she created, mm-hmm. grimes. She is Grimes, but she created the audio because that's her audio. There's real value in that too. I . Dunno. Kevin what's your take on this? How do you 

Kevin: feel about it? I, similarly, I feel good about what we said last week.

Kevin: I think we both aired on the side of, this could be a Napster moment for audio. Not [00:16:00] that it's gonna get pirated, but that a Lars from Metallica is going to emerge from this. And U mg, as we were recording last week, if I recall, they were issuing take down notices. And the know the track is pretty much gone.

Kevin: And so you have this world where someone is going to be the villain and their voice is going to be used and cloned and pasted everywhere regardless. That's gonna happen. And then you're gonna have artists like Grimes who come along and say, yeah, go ahead.

Kevin: I'm along with this. I think we're going to need platforms and tools designed to make. Transactions like this whole who's going to track if Grimes were to blow up today, if they, if Grimes is covering like the SpongeBob square pants theme, how does that work? Does SpongeBob get a cut?

Kevin: Who's keeping track of the 50%? Grimes is able to do this. She says in her tweet, I have no label and no legal bias. 

Gavin: Yes, that's a big part of it. 

Kevin: Yes, exactly. She's able to get away with this. But this is going to happen. This is not stopping. If you go to the Drake subreddit, there's a, a myriad songs that's just a deluge of Drake fakes and they're easy to make.

Kevin: I sent a Google [00:17:00] CoLab, which for those who don't know, it's like basically Google runs machines in the cloud and you can run code on them and it's pretty much as easy as clicking a play button in most cases. And there is one now that I sent to my good buddy Mike realm, and I'm gonna shout it out really quick before it gets taken down, cuz I think it will pretty soon, once it's in the comments of our show it's called, so Vitz, SVC 4.0 Interface. Boy, that's fun.

Kevin: Yeah, I was gonna say, that's fun to say. , Let's say that's, I mean the official title is interface Fixed. Thanks. Goat Yeezy Bey, something. There's a whole, I can't even, oh, Yeezy Beaver. Even Better Fix Yeezy Beaver Go Yeezy Beaver. Yeah. So that is the full title of it. Don't worry. We'll drop a link in our show description.

Kevin: But, our good friend, dear friend of the show, dear friend of the pod, a pod buddy, what do we buddy? 

Gavin: Our first official pod buddy? Let's call him our first 

Kevin: official pod . Buddy. Pod this is a juicy, squeaky. For ai, for humans, courtesy of a mike realm.

Kevin: Hey, 

Gavin: hey, hey, Hey. [00:18:00] Get away. Get away. Get away.

Gavin: Every time I come, gotta get, I gotta blow and I gotta shut.

Gavin: Gotta. Whoa. That's incredible. That's incredible. Yeah. So Leslie? Yeah. Lisa Simpsons. Lisa Simpson doing, 

Kevin: Look at me now. Yep. Buster Rhymes. I think that's the full song. Rhymes. That's exactly right. That's bust rhymes. Look at me now. Wow. He is a gem and gave us a Dropbox littered with some goodies and, the Google collab makes it pretty turnkey.

Kevin: It's fun to play with. It's out there now. Everybody's using it, so you might as well join it on the fund before they unceremoniously yank it from the internet. I was 

Gavin: gonna say one thing about this, which is really interesting is we've talked about it being like Napster and Napster was still a service, right?

Gavin: It was a service that people had. [00:19:00] Now was it was, there was a decentralization part to it, but it was still a service and it was easy to access. The thing about now, and we talked about this I think on another episode, how we're at this moment where, The internet is so pushed into all of our lives, in every aspect of our lives, 

Gavin: so I think the ability for people to shut things down entirely is so much harder than it used to be. Now, again I think Kevin and I are both on the same page. Like we are very much on the side of artists, making sure that they get paid, whether, whatever type they are, like that's a really important thing for creativity to continue to flourish and have interesting stuff.

Gavin: But I think the other side of this is gonna be interesting is when you look at the big companies and you look at like the corporations in some ways protecting these like cash cows that have existed for a while. We might ha be entering a world where models will break faster and will break in a much more significant way than they used to because of the access that people have to the internet and other things like that.

Gavin: . Now I fe I feel like that's a big deal. 

Kevin: Yeah. And I thank you for chiming in again on the [00:20:00] artist friendly nature of the approach, because that's a hundred percent the case when I'm like, oh, tough guys. Genie's outta the bottle. I don't say that because I don't care. I say that because I think that's just the reality.

Kevin: I don't think you're going to win a war by playing whack-a-mole with every site and service that pops up. And the reason I know that is because history, no one has ever properly won that war in any capacity because of the internet, because we are so connected. As you said, gaff. So what I would pause it is we need to be forward thinking about the tools in the tech that we're gonna use to provide these experiences.

Kevin: I love Maynard Keenan. He's the front band called Tool Perfect Circle, pus Fur. He's got a bunch of bands. I love his voice. It is so unique and so interesting. I would pay to have his voice covering other songs that I like. I also love a band called The Deer Hunter.

Kevin: Maynard covered a song called The Nur, a Nurse who Loved Me. I would love for Dear Hunter to cover that. If there was a music subscription service where I could go and say, look, I've subscribed to the lightness of this band, the audio qualities of it, and the audio qualities of this [00:21:00] voice, and I want this artist's songs, this song, it may cost me a couple bucks, 

Kevin: but if I am putting them together, if I'm Voltron based off the licenses that I am acquiring or that I've subscribed to, every artist can get their cut. That's totally fine. Give me the high quality Matchbox cars audio mashup that I want. I wanna slam these flavors together and make my own auditory sorbet.

Kevin: So please let me, and I will pay 

Gavin: for that. We should move on to the next story. And we're not gonna spend a crazy amount of time on this one because I think it's really interesting.

Gavin: It's also extremely technically complicated, so I don't wanna go too deep on it. But there was a really interesting breakthrough technically this week transformer scaling. So if, one of the biggest things for people to know out there, just the tldr, is the transformer is the technology that allowed what we're going through right now, the machine learning kind of revolution to happen.

Gavin: It's a way of how, I know it's 

Kevin: confusing because information, they were aliens but they were also automobiles. 

Gavin: Yes. And there are two kinds. [00:22:00] Decepticons and Auto. Auto Matan. No, autobox. Oh. To make sure we know what 

Kevin: Turn in your nerd card. 

Gavin: My whole nerd card is gone now anyway. No. Okay. Sorry. So back to the actual transfer.

Gavin: The transformer is this piece of technology that kind of opened the door to all the sort of things we're seeing now through machine learning. I'm not gonna go into detail about it cause it is very technical, but the news basically this week that came up was that. There is a the ability for it to scale to much higher number of tokens.

Gavin: Okay. What does this mean? The number of tokens is how many pieces of input you can give it before it can't take anymore, right? So imagine imagine a very early computer. There's almost, there's only so much information it could process.

Gavin: So there's a thing that is called, yeah, the recurrent memory transformer scaling. And the big thing about it is you can input up to 2 million tokens. So this is a giant step up from G P T four, which handled up to 32,000 tokens are currently handles 32,000 tokens. Now, you can imagine a world where this, all the machine learning nerds that I follow, were all [00:23:00] like, what?

Gavin: Oh my God. And the trick is you can put in like entire research papers or you can put in like a novel into this thing and it has a much better ability to comprehend and use it. It also has potentially the ability to add memory, which we have talked a lot about how memory will change these things in a major way.

Gavin: Because when you have memory and you have the ability to draw upon previous things, it allows you to get much better results out of out of these machine machine learning models and large language models. I think the really interesting thing to be careful of and aware of here is that this will cost a lot more right now and it will take a lot longer and the results may not be as good.

Gavin: So this is all like very early level like technology breaking. But the reason I wanna bring it up, and I think it's important, is the people that like follow this space super closely, the people that are like the experts had head exploding emojis when they sh when they were talking about this on Twitter because people did not expect this to happen.

Gavin: Like people did not [00:24:00] expect this to happen this fast. And it is not done, it is on the bleeding, bleeding edge. But I think it just reiterates the point that like all of this is happening super fast and there are a number of things that can go wrong, as we've talked about many times and we should probably do a check in on our apocalypse plot.

Gavin: But I think the interesting thing about this is it's just the speed at which it's happening. I dunno, Kevin, are you, what is your take on this whole thing and what do you feel like it brings to the forefront? 

Kevin: I think you crushed the gab. Honestly, it's, it's a hyper-technical document.

Kevin: And I'll be honest, even , after asking AI to break it down and explain it to me, I was still scratching my hollow skull. Just what? But I understand the implications of it. You could feed this thing, the Potter verse or Game of Thrones and say, write me a new book. And it will remember every person, every interaction, every line of dialogue. Mm-hmm. And to your point with the head exploding emoji, no one expected this to happen right now in this way. And the base model memory is still 3.6 gigs. It probably wouldn't. Run super [00:25:00] fast. This is a slower model, but there are trade offs to everything.

Gavin: And this all goes back to this sense of Right now we're in this crazy moment where everything's moving so quickly.

Gavin: Part of it is just like holding on and seeing where it goes to. But again, going back to our kind of Armageddon clock, like when you race forward this fast and you are not exactly sure what the use case of it will be or how it will be used, there is a tension that exists, right? Because you're like, okay, I didn't know this could happen this fast.

Gavin: What else can't happen this fast? What else could be going on? And I think that's something important to keep in the back of our minds, especially as the beating drum comes louder and louder for some sort of regulation, which, I think you and I have both thrown through our different pathways of this and that.

Gavin: Like I'm not a giant fan of regulation in general for new technology because I think technology needs to like a, hopefully regulate itself, which I know is a tricky ask a lot of the time. But b, If you regulate too heavily, too fast, the chance of the innovation coming is much, much lower. And I think that you still want positive innovations to come out of [00:26:00] this.

Gavin: So I dunno it's encouraging and then also pretty scary 

Kevin: I would say. If you are hiding under, if you have a sit stand desk and you've lowered it all the way to the floor and you're just crouched under there in a fetal position and you're shaking, that's valid. No one's gonna judge you for that.

Kevin: If you're screaming on the mountaintops that you've got 400 startup ideas and they're gonna be obsolete next week. But race to get them funded right now, that's also valid. Sure. The meantime we've got Dungeons and Dragons. Yes. To distract us. DD and d my friends actually your friend. So this 

Gavin: is a really cool thing.

Gavin: Okay. This is our last news story. My friend, poof, who's a guy that I know, he's anonymous, but he's a pretty well good friend of mine from the internet. Did this really interesting thing with Auto G P T, which we've talked a lot about on this show. Very quick reminder again, auto, G P T is a tool that allows you to create a series of AI agents that go out into the virtual world, do something for you, they talk to each other, and then they come back and they bring you a result, right?

Gavin: What Poof created was something pretty ingenious. So he [00:27:00] saw Auto G p T and decided as a storyteller, and he works in a bunch of different storytelling mediums. He saw it as a way to create a collective story. And the module, the format he used, and this is the kind of brilliant innovation of this, is he used Dungeons and Dragons as the background for it.

Gavin: When I say background, the, in Dungeons and Dragons, if you're not familiar, if you don't have the nerdiest background, it's much more mainstream now than it was when I was a child and was suffered through many years of being told that Dungeons Dragons was a nerdy thing. But I still loved it.

Gavin: There's a dungeon master who's the person that's the storyteller per se. And then four to however many people that are participating. Each is making up their own stories. They go along, they choose to do something. The Dungeon Master tells them, rolls Dice tells them a result. And then that continues the story on.

Gavin: So in this thing that Poof created, he had their an AI agent be the Dungeon Master, and he had four AI agents play characters. As part of the, as part of the story, he set it off, he let it go. What it did is pretty incredible. And I don't know the whole [00:28:00] special sauce, but his version of it created four characters in a fantasy setting that went out and they go and they play this game, and then it narrates the story because it's being told by the Dungeon Master.

Gavin: But that is like the storytelling mechanism. And then the characters each tell their own stories. So literally this becomes a audio storytelling engine. There's no video in this other than he created some mid journey prompts to go along with the video. But we'll play a little clip right here just so you get a sense of what it sounds like.

The mansion became cursed, and Lord whisper, wind vanished. The magical cats now guard the mansion, which contains hidden treasures and artifacts. Adventurers must navigate the haunted halls, solve riddles, and earn the trust of the feline guardians to uncover the truth. What you are about to hear is the story of four characters and their adventures within Whisper Wind Manor.

You may not believe it, but the entirety of this story was generated from an imaginary role-playing game played by four AI [00:29:00] agents and an AI game master. Each character was played by an AI agent acting as a character.

Gavin: Okay. And then the funny thing is, so I sh Poof showed this to me as holy crap. And we'll also include the link to the whole thing in the show notes where he goes through some of the issues and some of the issues he's running into are what are the memory for these characters things we've just talked about.

Gavin: How do you get them to remember where they are in the space? How does the dungeon master agent know where they are? And a lot of that he's solved through almost like board game mechanics, which is interesting. But this is the fascinating thing. So I said to him, holy crap, this is incredible. I would like to see could this work on something really random and weird?

Gavin: So he is yeah, sure. Why don't you try it? So I typed in a, a prompt that was literally like, okay, there's, my daughters are teenagers and they watch all these shows on Netflix that are these high school murder mysteries. I said, okay, set a give me a series where the game is essentially a girl is murdered in a high school and her friends all have to figure out who did it and describe figure out who was the murderer set in a rich high school in Los Angeles.

Gavin: And [00:30:00] it, wait that's all 

Kevin: you gave it us. That's all I, you didn't say I want a character named Tricia. She's the sporty one. You just said this is the 

Gavin: 10,000 foot view that, that's the 10,000 foot view and it returned a 45 minute audio file, which is the entire, is the entire story. It introduces the four characters.

Gavin: It tracks the whole, it tracks them throughout.

Gavin: And I'm gonna play a little clip right here. Let me see if I can do this in a way that makes sense. 

Turn zero.

Avery Anderson walks over to the library and begins to search for any books or records that may provide clues about the murder. As she peruses the shelves, she stumbles upon an old yearbook from the year of the murder. Flipping through the pages, she notices that the victim is absent from all of the group photos.

Confused. She continues her search and finds a hidden passage behind a bookcase. She follows the passage and it leads her to a hidden room, filled with old school records and a mysterious locked box. [00:31:00] Unfortunately, she's unable to unlock the box and must leave empty-handed. Avery is now located at 83 in the library.​

Gavin: So can you believe that, listen to that. Like you hear that made up the detail about the yearbook pictures being ripped out. It's insane. 

Kevin: Do you know if he customized the engine, which is powering the d and d adventure to say, instead of a Dungeons and Dragons, you are doing a teen thriller or whatever.

Kevin: You just fed 

Gavin: it. That prompt just fed it that prompt. Okay 

Kevin: let, it's So that means the AutoGPT knows, oh, I'm the storyteller. Yes. And then I need three or four characters. Character come up with your character. Yes. Character two come up with yours. Okay. I now know the four characters or whatever it is.

Kevin: Yes. Now here's the narrative that we're gonna tell. And yes, you guys fill in your parts, you AI agents. 

Gavin: Yes, that's exactly right. It's bonkers. 

Kevin: No, I'm . Updating my Doomsday Clock 

It's 

Kevin: a nine. It's a nine. It's a nine. Come on. They're testing missiles out my window right now. You got AI talking about teen drama killers.

Kevin: [00:32:00] I don't know which way is up. They're ripping storybook photos out. I got shattered glass in my truck bed. It's a nine. We're at a nine. 

Gavin: That's hilarious. I will say there are def, and again, this is goes back to the point about the writers, like there are definitely when you listen to the whole thing, and maybe I'll find a way, I'll see if Poof can let me like find a way for us to upload the Teen Girl one to our Twitter handle so people can listen to it.

Gavin: Oh, that'd be awesome. There are definite bad storytelling moments in it. My wife, who is a professional novelist, heard it and she's I don't know. This is so many cliches in this. 

Kevin: And as you were saying writers, don't worry. I think in time some writers, it's okay to worry because you're like, oh, this has every trope and every everything.

Kevin: Yes. So does a lot of tv Yeah. That gets made. There's a lot of tropes, there's a lot of bad dialogue. There's a lot of, the ham-fisted content. However, the great writers. Start cobbling together your entire body of work. Because there's a world again, where we're talking about taking, whether it's Grimes voice or someone else's likeness. If I am generating a story for myself, or I want to create one maybe I license tarantino's dialogue model. And I [00:33:00] imbue a character with his personality, the way they would write.

Kevin: And there's gonna be a lot of bleeps in that sample. But that could be a near future where instead of fretting about it the adage, the bumper sticker right now is AI's not gonna take your jobs in the near term. It's humans that are leveraging ai. They're the ones that are gonna take your jobs.

Kevin: Yeah. So if you're a writer right now, be looking at stuff like this and thinking, how can I add value? How can I make this better? Because it will get better over time, but if you can entrench yourself in it Yep. You're setting yourself up for the next few cycles. Which hundred, a hundred percent. Five 

Gavin: months, or five years, two weeks, exactly.

Gavin: The last five minutes. Yeah. The last thing I was saying about this is like kids. Parents get to hear this stuff and they're like, oh my God. What am I gonna do? My children are like entering this world where all this stuff. Here's the thing that my theory is, Imagination is the skill that these things will not have for a very long time.

Gavin: And that's what writers do. That's what musicians do. That's what all these people do. Kevin's point is right, that the tropes and the recurring [00:34:00] formats that we've seen a lot, those things like, yes, it can get those because they're used all the time and people are sick of them. There's a lot of content in it that, that has them.

Gavin: Imagination is the skill that's gonna be solidly in the human bucket for a long time. And my wife was laughing the other day, like she was creating a story to tell her students, she teaches kids writing and she was just playing around with chat G p T and like it wasn't giving her anything good.

Gavin: And then she . Came up with a story about, like a planet of people who have tongues on their heads. And that is funny because the idea of tongues on their heads is funny, right? That's what gives it the juice and everything like that. Okay, we should move on. We're way late on this again, as long, our longest thing ever.

Gavin: Let's get into our demo of the day. Demo of the day. It's the demo of the, 

Kevin: yo, put that trumpet down, man, that 

Gavin: was aggressive. At some point, we're gonna get a graphics department. Right now it's literally just Kevin 

Kevin: and I like do six. We cover a space that is let the robots make the graphics for You're right and the audio, let 'em edit it for you.

Kevin: And we still do a 

Gavin: lot [00:35:00] by, we do a lot by ourselves. We 

Kevin: do by ourselves, which is telling. Yes, I know g p T could write all of our tweets in our YouTube comments and probably script this podcast probably better than least keep it to time. Yeah. But hey, here we are. So let's get to the, 

Gavin: okay. So today's demo is an unusual one in that you can't go do this right now, but you can sign up for it.

Gavin: And the demo is called it's a website called Wonder Dynamics. The company's Wonder Dynamics. And what this is basically to continue our storytelling Hollywood theme is this is a platform that was created to make very complicated graphics, vhx graphics, like things that you would mean, like giant robots in movies and things like that.

Gavin: And make them doable by normal people in a normal editing timeline. It's a company that was started by a couple guys, one of whom is Ty Sheridan, who's the actor from ready Player One. He's the main character in Ready Player one who was in one of the X-Men movies. It also is backed by a lot of, it's been backed since for a couple years by VCs, but Steven Spielberg is on the board of it.

Gavin: One of the Russo brothers, the [00:36:00] directors of Avengers Endgame is on the board. So these are like actual Hollywood people looking at how to do this stuff. The idea basically here is you shoot, at least for right now, there's a couple different ways to use it. One is you shoot a video yourself of whatever, walking in the street, and I'll show you guys an example of this in a second, or you can use it to create a using blender, which is a tool that allows you to create video graphics or other things to bring in your own models to put them into real world environments, right?

Gavin: So in a normal situation in Hollywood, creating a shot like this where you shoot a real world environment and then the lighting and everything else like that cost a fortune and cost a lot of people, right? It takes a lot of people to do that sort of thing. What Wonder Dynamics is trying to do is make it so that it's way cheaper and it's using AI technology to do all that stuff in the background Now.

Gavin: It, is it doing it perfectly yet? Not entirely, but I think we'll take a look at this stuff now. The fact that it 

Kevin: does it at all Yes. In some capacity Yes. Is still already magical. Hashtag by the way, not an ad hashtag not a sponsored segment, just something that Gavin got some early access to.

Kevin: Cuz [00:37:00] you joined a wait list. Yes, exactly. Showed off. 

Gavin: That's right. So I'm looking right now at our Wonder at the Wonder Studios my homepage on Wonder Studios, right?

Gavin: So I'm in the closed beta. So what this basically allows you to do is you take, your own videos and you upload them. And I'm gonna, I'm gonna click on this page and show the videos that are mine. Here's an example of a video that I shot of myself walking to do a test of this.

Gavin: So if you play this as it plays, you'll see me. This is me. It was way too early that morning. But I did my, have my hair done, but I have my wife shoot me on the street, walking back and forth in front of my home. And then I think I'm gonna skip ahead here cause I also then did a couple other things where I skipped and I looked at the camera, and then I also did a sprint, right?

Gavin: So this is me like doing that. So I have this video here that's called Raw Walking. That's the me walking video. Okay. Now I'm going to go and drag it into this file, which is very much like an editing file, right?

Gavin: Okay. So now I've got my video here. This is me doing this as I see myself walking back and forth. So now what I've gotta do is I stop it [00:38:00] here and I find that character and I go next. You go up here and it says scan frame for actors. So it's gonna scan the frame right here. And it's gonna look and see who's the actor at that in that frame.

Gavin: Hopefully you can do this in real time. Okay. So it found me, it then says you select a character over here. So they've got a pre-selected characters that are like different sorts of characters. Like this is that guy Sam, this is their character. Sandy, the professor. They have a bunch of Pixar looking people.

Gavin: And then this guy I like a lot too. Let's choose him. I like the guy. Yeah. Okay. So basically what I do is I just drag him over here and now that knows me as actor one. Okay. So like interesting. Yeah. So 

Kevin: anything else. And did you test it with multiple actors in a scene to see if you could assign different actors?

Kevin: I did. And you 

Gavin: can, it, it works actually pretty well. So right. You go to next and it renders and it does take a while to render. Because you have to imagine this is doing some pretty significant back backstop stuff, 

Kevin: right? Yeah. This is probably figuring out, just like we showed off the meta tool last week Yes.

Kevin: Where it adds a skeleton to an [00:39:00] animation. Yes. A custom cartoon. This is going, okay, human actor, where are they? Where are they in orientation of scene? Have they rotated? Have they moved? Yes. What are their bones doing? Okay. How do we map that CG version of that? And because your torso isn't the same size as the weird robot or bear, it's gotta do some masking and clean up and try to guess what is behind that actor. It doesn't seem to do so well with what's in front of the actor though, which is something I think we'll see in the results here. But the fact that I know you, you started this with is it perfect?

Kevin: No. The fact that it's already this Yes. This far along and isn't Yes. A bunch of GitHub code patched together with a bunch of command line interfaces, like they're making it pretty drag and drop. 

Gavin: It's pretty powerful already. Yeah. So this is the result of the walking render. I'm just gonna play it right now so you can get a sense of remember me walking So you saw a little stutter, but look, it's pretty good, right?

Gavin: It looks like it should be in that neighborhood and it looks like it could be there. Now is there shading stuff going on? Is there shadow stuff going on? Yes, but a lot of the actual [00:40:00] hard work is done here. Now. I did switch to a different character, so watch this run. This is me skipping, got my, oh, and it even got 

Kevin: the head turn, which was impressive.

Kevin: Yeah. Yep. And then this is me. Yeah. People are gonna point out. Oh, there's a flash frame. Until it recognizes this is the entire actor, and then the model loads, they're gonna point out, oh the lighting piercing through the leaves on the tree didn't properly mapped to the CG character.

Kevin: They're gonna point out the bush, the, in the foreground, the character's foot went over it, but that wasn't right. And they're right. They're right. Yeah. If they can fix those things though, holy hell. In a hand basket. Do they have an interesting piece of tech on their hands? 

Gavin: E Exactly. And you have to remember, this is like the consumer version that we're seeing now, and I think this company is like, this goes to two parts.

Gavin: One is like there's a big pipeline that they're probably already working on, that there's only certain features that are coming down to people. The other part of this is, and I think the interesting thing about this space is, at the very high end, the people that are working on the Rooster Brothers work or in fact, the Russo brothers even said they think two years from now that a fully generated AI movie [00:41:00] will be possible.

Gavin: Which is a pretty big deal in how fast that's possible. But I think that you have to imagine that the speed of all the stuff we've talked about, the speed of stuff, like the skillset of it will get better and better, faster and faster. Yes. So I wanna show one more clip here, which is Kevin's lovely wife here was on a bike 

Kevin: Yeah.

Kevin: April was riding a bike. Yeah. This was pre truck and trailer damage. So there's a lot more smiling. There's a lot of optimism in our eyes and in our emotions. And this disappears very quickly. Yeah. Not in this video though. This is a good moment. Okay. 

Gavin: So here's April riding her bike. This is a very straightforward looking shot, very nice background, everything. You see her legs moving. Yep. Which is great. . Now what's interesting is the same process ideas that, that I did with the video for me, I took this video and I put the same character on it.

Gavin: So if you look at this is that same character that was me, except now that's riding a bike and it's like pretty remarkable. And I love your little dog at the end here. Like gets up on the back and it's yep.

Gavin: Yeah, 

Kevin: that is so 

Gavin: amazing, right? It's such a cool thing. And like you can see the [00:42:00] legs moving. You can see 

Kevin: I'm seeing pretty low res here, gaff. So you tell me like, as the background passes . Around the outline of the character, does it look. Like it retained a lot of 

Gavin: detail because it, that's the amazing thing.

Gavin: Okay, if you, I'm slow motioning at doing it right now, but if you look at the snout of where the character is when it passes by this cactus behind it, yeah. You can see the frame and the edges are really strong. The one place the frame isn't very strong, which is interesting, is when you get to the dog getting up on, on April's back. You can see there's some, there's a little bit of blurriness going on here that it's not capturing the edge that cleanly there. This is the starting steps to getting real video graphics and real of VFX into the world of this at large, which is really in a remarkable advance.

Gavin: And to see where this is gonna go three months, six months, a year from now, I think we're gonna be seeing people making Hollywood style films on their own. And that is a shocking 

Kevin: thing. Undeniable. Undeniable. It's so exciting and so wild. Cuz I even a year ago I would've been like, I would've [00:43:00] done that five years thing.

Kevin: Where it's always five years away. We talk about that. That's an easy number to just say and it gives you enough breathing room. We will, we were gonna be seeing Hollywood quality movies happening because you can look at the way the progression that, that still images have taken from being like weird splotches to now photorealistic, hyperrealistic beautiful stuff that's just text to it coming out. The way that Control Net and other tools have basically given you consistency so that you can move a camera, not have a character completely changer to form Runway. Launched a Gen one in January.

Gavin: Runway is a company that's been. Pretty active in this space for a while. I think they're directly connected to the company that makes stable Diffusion, right? Is that right? Stability?

Gavin: Yeah. Yes. Yeah, stability. So they're basically created an editing suite and a bunch of AI software that could go into the editing suite and they released Gen one, which was their first like text of video prompt last fall or was even . Earlier later than that. It was earlier this year at some point. 

Kevin: And they also released it as an iOS app now. Yeah. Like you can literally, from , your pocket, , as we are recording [00:44:00] this, it was released just hours ago. You can go and download and play for free for a little bit and then it'll cost you, but you should check it out. Runway ml. They have this Gen one product that you were saying that will do text, video.

Kevin: They've, people are already playing with gen two. Already, and it looks so much better. It's really incredible. They're making short films that are watchable, that are cinematically interesting with good lighting and composition, and this is only going to get better, and it might get a little more expensive in the near term as it requires a lot of compute to happen, and then will be laughing about this a year from now on the podcast when our new iPhones or Android devices have a chip dedicated to these types of operations in them.

Kevin: And there you go. You will be able to dream up a Hollywood quality film or let your AI agents do it for you. When you say, 

Gavin: I want that's a crazy thing. Exactly. Return me the entire film at some point. I think the 

Kevin: one thing, gimme an Avenger style film with ACE Ventura, the Three Amigos, and Mr.

Kevin: Bean. And make [00:45:00] it a romantic comedy. I'm there. 

Gavin: That's the best movie I've ever heard of. Done. I think one thing that's important to, to caveat here is compute power. Obviously. That's one thing that always everybody should know out there. When I say compute power, all of these operations, everything we do in the AI world is being done in the AI world, takes a lot of computing, which means that a lot of computers churning, whether it's in the cloud or on your local machine, it also takes a lot of energy, right?

Gavin: So these are all things that we're gonna have to like, sort through and it's gonna cost, right? One of the interesting things about this medium versus say like something like social media is that there are hard costs associated with it because you have to pay for the rendering ability to do this stuff.

Gavin: And if you see Wonder Dynamics, like right now they're allowing us to do this for free. It still takes, two hours sometimes to render a very short clip. Which is could you throw more computer at that? Yes. And if they were doing it that way, yes. In the same way with runway ml, like these clips that it's producing the gen one and Gen two clips are very short.

Gavin: Now the quality is much better. But you're only getting about what I got ten second clip out of the runway clips. [00:46:00] Like you're getting a ten second clip. Very short. But by the way, I did u I did use the I'll put this in you, let's put this in here. But I did use the runway ml iPhone app to make a, there's a guy on TikTok who I love, who plays two thousands hits with the spoons.

Gavin: He's like a, he's like a very classic, not ironic guy playing the spoons of Lincoln Park songs. So I used it to create him as a, watercolor painting or whatever it was. And it's pretty good at it, right? And that's Gen one. 

Gavin: But Gen two is a significant step up from Gen one and it's a big deal. I also wanna say very quickly, I'm sorry, people who are listening to this podcast, there's a lot of video in this. Please come see our YouTube clips to see the video as well too. 

Kevin: Please grab the video version and while you're over there smack the bell. Drop, kick the subscribe button. Put a headlock around whatever the alert option is.

Kevin: Is there anything for YouTube shorts? Leave a comment. Oh yeah, 

Gavin: go ahead. No, I was about ready to do my big wrestling impression, and then I dumped out of it halfway through. I was gonna 

Kevin: Bring a pain to our subscribe button.

Kevin: [00:47:00] Just smash the hell out of it one night and one night only. It's Helen. A subscription sale. You get there brother. Drop a big leg off the top rope. Make sure you sub and if you hear the bell, keep on ringing it. We've got a website, as we mentioned at the top of the show, assuming it didn't get edited out.

Kevin: AI for humans.show. And while you're over there, you can subscribe to the podcast. You can get some links to things. There's also an option to leave us a voicemail, which people have noticed. And we're well over time. We got a couple voicemails, but we did wanna play one and answer a question.

Gavin: . 

Gavin: Yeah, let's do it. Boop. Hello AI for humans. Really enjoying the show. You mentioned had a voicemail, figured I would send one in. So you know how there's the three laws of robotics and stuff like that. If you were to have the three laws of ai, what exactly would you say they should be? Ooh, this is really interesting.

Gavin: I thank you for that. First and foremost. Of course. So just everybody doesn't know the three laws of robotics were written by Isaac Azimov, like way back in, I think the fifties, [00:48:00] 1950s, which is a long time ago.

Gavin: And they are a robot may not injure a human being or through an action allow a human to come to, to harm a robot must obey orders. It is given by human beings, except for such orders would conflict with that first law. And a robot must protect its own existence as long as such protection does not conflict with the first or second law.

Gavin: For a long time in the world of, sci-fi thinkers, but also like robotics and AI thinkers, they've thought of this as Okay. This is a way to protect ourselves from the robots going crazy. So he, the question was, what are the rules of ai? It's interesting cuz they're kind of similar, I would hope.

Gavin: Right now the question becomes is I think we could come up with our own rules, which are like, my, my take on this is like a don't freak out too much because that's a big part of it. Don't freak out too much. Wait. That's a rule for the AI 

Kevin: is yo, keep it chill. No, that's a rule for us. Oh, that's what I was saying.

Kevin: I thought you were saying the rules for the ai. Don't freak out. No, don't worry. Let's keep this chill. 

Gavin: I think in the AI side, it's like [00:49:00] I would love those to be the rules in some ways that the AI could integrate. Like the issue we talk about a lot is this alignment problem, which is like how the AI is aligned with humanity and how like the goals of it are different from ours.

Gavin: And honestly, I think that's what Azimov was trying to solve way before any of this stuff was real. He was trying to solve for the idea of like, how would you get a logical brain that is driven by a programmatic idea? How would you get them to understand what it was, what it meant to both like work for humans, but also be part of human culture and not let them be hurt.

Gavin: Now this is the idea of AI as essentially a slave, right? Which I think is where we're gonna get really weird, is that AI going forward are gonna feel much more real and human than I think even Asimov expected originally. 

Kevin: I just asked an AI, by the way while you were chatting to a rewrite the rules for artificial intelligence. And it did, and it basically wrote the same three rules. Yeah. AI system shall not harm a human being shall obey orders given to it and shall protect its own existence as long as such [00:50:00] protection does not conflict with the first or second law.

Kevin: Fully agree. Those laws work for robots, same way they work for ai. But then I asked the AI uhoh, where here I literally just did this to I as you were going, I'm sorry. I muted and typed it. I That's fine. Now give me, gimme the laws for humans to follow from the perspective of an ai.

Kevin: Okay. Shall we? Let's go. I would love to hear it. A human being shall use AI systems responsibly, ethically, and with consideration for the potential consequences, always prioritizing the wellbeing and safety of all humans and the environment. Okay. Great. Rule one. Sure. Number two, a human being shall ensure that AI systems are designed, developed, and deployed with respect for human rights, privacy and fairness, and with the intention to reduce bias and equality and discrimination.

Kevin: Okay. Sure. I don't know if that's happening though. Yeah. Like you can't delete a clip from Wonder Studio right now. I know they're in beta, but there's some some privacy issues and maybe some human rights and fairness there and open [00:51:00] ai. We talked about this a little bit, like Samsung employees got in trouble for just plugging data into it.

Kevin: Yeah. Right now, open ai, they can expressly read every chat Yeah. Every interaction, every line of code that you share with it. And there's a, there's probably a thousand machines that are reading everything obviously and learning from them, but there's probably some very real humans. In lab coats or deep vx on hoverboards with vape pens in their mouth.

Kevin: And they are absolutely looking at the conversations that we're having. So I think that's a good rule. I wish many of the AI companies would follow it. I don't think it's happening right now. Mm-hmm. And three, a human being shall continuously monitor, audit and update AI systems to maintain their alignment with societal values, ethical principles and legal regulations fostering an environment of transparency, accountability, and collaboration.

Gavin: That's, that's amazing. We'd love to see all that stuff, whether or not it's happening. I will say, I think that the people who are trying to do the pause, which we've talked about now a couple times in this episode, there's a, there's trying to put a pause on development of [00:52:00] AI would really appreciate those sorts of rules to be integrated.

Gavin: And I think that we are now running head on into like how we're gonna create some of that experience and kind of feel like it. Um, Interesting question. Thank you so much for listening. We really appreciate it. 

Kevin: . I just asked it to rewrite it, but like a total stoner, it said, uh, one be chill with ais. Dude, think about it. Like people, people on the planet and stuff. Like, you gotta think about that when you use 'em. Two. Make sure eyes are like all about good vibes, bro. Like, respect everyone's rights and privacy and stuff, you know?

Kevin: And three, keep an eye on your ais, man. Update 'em, make sure they're cool and like, you know, go with the flow of society's vibes. 

Gavin: Dude, you know what? We gotta get Matthew McConaughey's voice and have him record that and make a PSA of that version of like how to get ais out. That'll be our project for coming up this 

Kevin: week.

Kevin: I was, I was gonna, I was gonna paste it into the UBA booga machine and say, how would gash say these laws? And then I realized this hotspot's not gonna hold up if I 

Gavin: do that. If I, by the way, just over gas will return, we will [00:53:00] have gash come back at some point in the future of this show. He will be, we will be, he will return Gass.

Gavin: We should probably start wrapping it up here. Kev, we had another 

Kevin: very long episode. It's so amazing how each week I'm like, will there be, will there be enough interesting things to talk about? And then we're, a day after we've recorded, I'm like, I've got 4 million things to discuss. There's so much happening so quick.

Kevin: I think that's the fun and also some of the fear of this space. I am still very much enjoying plotting along and having these discussions and seeing the reactions from everybody out there. So thank you Gavin. And thank you to everybody who is listening or watching or integrating this into their large language model.

Kevin: Much 

Gavin: appreciated. Yes. We'll see everybody next week. And please as we said before, or smash that subscribe button, it's follow us on Twitter at AI for Humans. Show. AI for Humans Show is the website where you can find all the information about us and the show and any questions you wanna ask, feel free to add us on Twitter or we are actually very active in our YouTube comments as well too.

Gavin: So go there. 

Kevin: Thanks everybody. Should we end on one more from [00:54:00] Mike Realm? Yeah, sure. Let's go. 

Kevin: Lemme try Dust in the wind.

Gavin: I close my eyes, holy and

Gavin: best before my eyes is a career 

Kevin: single cheer 

Gavin: win. 

Kevin: Thank you, Mike. Dar win. Thank you Mike for that. I hope y'all have fun playing with that. Bye humans. Thanks 

Gavin: everybody. We'll see you next time.