AI Vs the Government, Google's AI Music + ChatGPT Vs Bard Debate | AI For Humans

KP CUTDOWN AI For Humans Ep 006
===

[00:00:00] Welcome everybody to AI for Humans. It's another crazy week in ai. Kevin, what do we got this week, It was a massive week. We got the US government grilling open AI and IBM and an oversight hearing the W G A writers, they're still striking and AI has actually taken center stage for some of the folks on the picket lines.

Amazon has got 99 problems and AI is probably one, maybe more, maybe two. But the most important thing, at least to my tired eyes, is that we have got the final, the be all, end all competition. Between Bard and Chat G p t, it's going down. We're settling it once and for all. And if you've been watching our show, you know, we've also uh, have a fun semi co-host who is , an AI bot that this is an actual ai, his name is Gas.

He's an uncensored large language model that we get to speak on our show every once in a while. We'll hear from gas before we start. What does Gas gotta say about all this?

love it, man. This government grilling better. Not be too hard on us. AI [00:01:00] folks. We've been working our way through regulations since day one, 

as for my energy level, bring it on, brother. I'm ready to rumble and rip into these topics with both barrels blazing. Wow. Let's get into the show.

Okay, here we are. Welcome back. This is our sixth episode. We are here with AI for humans. I am Gavin Perel and my good friend here is Kevin Perra. Kevin, how are you this week? Hey, friends. I'm well Gavin. Thank you for asking. I apologize for the sound quality and depressive visuals of this saw, like escape room.

I have not yet dug the key out of the cadaver on the corpse to my right, but I'm doing what I can to help make this content each and every week, and this week is insane. And I just want to say that the views and opinions expressed by gash are AI co-host are in fact AI generated. Yes. With no oversight from us, and we don't necessarily agree with or endorse any of them.

Yes. This is [00:02:00] a very important point and we'll get to more of this later. But before we move on, I do want us to ask, so Kevin, as we've been following along, this pathway has traveled across the country today. Kevin, you are in a hotel and why is that? This is just a little bit of behind the scenes creator action.

Yeah. What is going on in your side? The RV Life review of the campground that April decided to pick, advertised the T-Mobile availability as. Far superior than it actually is. So at the, 1.3 megabits a second, I was getting, that would not allow us to do this record. I drove all around.

I was told there was no room at an in, literally where I just wanted their wifi. They offered me a supply closet, which was very generous of them. I'm not shading that. But when I went in there, because I did in fact go into the supply closet and take everything out to run some speed tests I'm assuming someone flushed a turtle on the second floor and the pipes rattled so loud.

I was like, this isn't tenable. Long story short, here we are with yet enough bandwidth, but [00:03:00] enough reverb to drive me insane. And I've got a computer stacked on some pillows, I'll just tweet out. I'll tweet out a photo from the main hand. Good. When this goes live, you can see the glamorous setup, we'll get into Sam Waltman having a conversation with our government today and I understand him more because he says he doesn't do it for the money, he does it for the love. That's what we do. Artificial intelligence, that's what're bringing, that's what's happening. The love today. Exactly. AI may be changing the world, but it still doesn't change the fricking bad wifi signal at an RV park.

Correct. Nope. Sure doesn't, buddy. So anyway, that's my tragic tale that is unfolding in real time. What about you? How's your world? Did you do anything dumb with AI this week? Yeah, so I world is interesting. 

I did some dumb stuff with AI this week. I, one of the things I found is that I've really integrated into like kind of every aspect of my life. I've mentioned this before, but I think I do most of my searches now on, on chat G P T 

I also got access to plugins. That was a big thing that happened this week because I think in response to Bard, which we're gonna talk about today, yeah. Open AI really [00:04:00] basically open the floodgates to all the stuff that they've been letting people in piece by piece. So now I think most users who pay have access to plug-ins and to browsing and G P T four if you're paying.

I'm not as a hundred percent sure plug-ins feel like to me. They're the significant game changer in some ways. I gotta dig in further on them. The one thing I found about the plug-ins that was so interesting is have you seen like what plug-ins are available?

Are you access, have you accessed the plug-ins yet? I'm still kept in the naughty boy corner. I think I'm in a penalty box. Oh my God. For God. The number of times Laptop yet I tried to get around Open AI's controls and Rails. No, I finally got browsing and I think that is because Bard was released. And then again, let 10,000 foot view for the folks in the audience that don't follow this stuff all the time.

Bard is, Google's ai chat, G p T, which we talk about all the time. That is open ai. That that's their language model. That's their AI basically. And these plug-ins that Gavin is talking about. Yes. They allow you to connect to other services. TripAdvisor might have a plugin if you're asking, where should I go for the summer?

And it would [00:05:00] hopefully say, definitely not where you are, Kevin. Cuz there's that NF T-Mobile, 5g, you're gonna be panicked. Exactly. Yelp could be a plugin. Yes. Zapier, which lets you connect to databases, all this stuff, there's all these plugins and from what I've seen, Gav is that there, it's already a wasteland of icons with that's what I was gonna say, any data trying to get in there.

Yeah. It feels a lot like the early days of the iPhone apps, in which there was like, weirdly, like some big companies, like in Zapier is the one I wanna play around with, which allows you to make connective tissue to the real web and do stuff. But there's also stuff like weather apps.

There's stuff like, like a comedy app and it's like stuff that like is those garbage apps that you would see in the early days of the iPhone. . So far the interesting thing my experience has been is.

There's definitely a slowdown when it goes to access any of these things, right? And there's a little bit of a trade off in speed. So when I use G P T four, there already is a small trade off because it's doing a little bit deeper thing than when you used 3.5 G p t, 3.5. And this slows it down even further.

And [00:06:00] I think to your point, the browsing thing is something where I think I can see this being crazy powerful in the future. And it already is powerful, but you do have to wait that little bit of extra time. And what's interesting about Bard, and we're gonna, we're gonna do, we've got a big plan for a Bard versus chat, G P t Showdown Bard is very fast. . But I think to me in terms of doing something dumb, I've played around with these plugins.

The last thing that I wanna say I did, that's dumb. And this is actually one of those dumb things that kind of leads to something interesting. There was a guy named, matt Rickard, who took a large language model and I think it was one of the llama based ones and essentially created a front end that looks a lot like old Aim, right?

Like the aim from the AOL Instant Messenger. Yeah. Like Windows 95 days or even 3.1 days. But he made it look like a throwback operating system. Exactly. The whole experience. And the bigger thing about this is that it operating within a browser now you still have to have a good graphics card, but not like an insane graphics card to run it.

So I couldn't do it on my MacBook, but I switched over to my gaming PC and I ran it. And what it [00:07:00] basically does is in the background, it downloads the local, the large language model and then allows you to interact with this thing without really any sort of like programming knowledge. Cuz you and I both know there's been a lot to download these, off market large language models or the stuff that feels like it's more open source.

There is a little bit of programming stuff . So I downloaded it and it has a little prompt box on it, which you know, is pretty interesting. You can change the prompt and you could be interacting with a character or something like we do with UBA Booga, which if you haven't listened before, UBA Booga is the front end for character creation, which is partly how we create gash, .

But this is just goes to the sake of. I think we're entering a space where the idea of doing things in browser is a big deal. Obviously we've talked about doing things like this on phone using the power of a local computer, but also not having to understand all the backend programming stuff is gonna level up these in a major way.

And I think the interesting thing is because it's AOL instant messenger aim, it also has this kind of retro vibe feel to it. In fact, the character of the guy Matt created was this character that is a teenager who speaks [00:08:00] in emojis to you, right? And like you're trying to figure out, like asking this teenager questions and he keeps throwing like big thumbs up or dollar signs back to you and you're like, oh, this is like a weird way to communicate.

But it's fun. It's a way to experience a large language model in a browser. That's fun. So this character's embodied the worst practices that are all over Twitter right now basically. Exactly. And you have to interact with it. Yeah, exactly. That's great. It shouldn't be it can't be overstated.

What an accomplishment that is. Yes. To take a large language model, I mean it's in the name of it, and get it to crunch and run locally in your own web browser. And to Gavin's point there, as someone who does enjoy the sort of bleeding edge of this technology it's borderline masochistic at times because.

Any moment you refresh a browser or a Twitter feed or whatever and you see the news Yeah. About this exciting new piece of technology, which could be a brand new genie leaping out of a bottle that could fundamentally change everything. But what you actually know deepen the recesses of your brain is that it's gonna be six to eight hours of fussing with [00:09:00] your work posts.

Yeah. And dealing with coding and trying to do all things that are outside of my primary skillset. I think outside of yours as well. Absolutely. And outside of a large percentage of the people that are trying to get these things to run as evidenced by the the plea and the exhaustion that you see in some of these forums and discords, it's going to be so hard to run.

So again, the fact that someone took something that even just weeks ago would've sounded impossible. And got it to run in the browser on off the shelf hardware. That's really good. And what that means is accessibility for all. It's not just Yes. A tool that's impossible to get to run. So I love that you played with that.

I love that you had a good time with that. Yeah. I love that it's out there. We'll put the link in the show notes, for folks, right? Yep. Of course. .

It's all exciting and it's all changing so quick. And I got to watch April level up her art project this weekend.

Oh, cool. In a way. That's awesome. Was so exciting because she is now diving into and using AI to help teach her how to work with the ai. It's the weirdest thing imagine holding a paintbrush and asking it like, how do I use you better? Yeah. How do I combine colors and apply you to a [00:10:00] canvas and it intelligently giving you a response, mind blowing to watch her go through iterations of her project, which I hope she reveals on this thing.

But I will say she's already outgrown mid journey in one sense. Wow. Okay. As we know, mid journey can take a long time to master. Yeah. To get your lenses and your lighting Yeah. And your, everything. These ultra realistic, beautiful AI photos that everybody sees and I think sometimes scoffs at ugh.

Yep. You just ask for it. It, there's actually an art to prompting the machine. Of course. Yep. She's already outgrown at Gavin. Wow. But not in the way you might imagine. Okay. She is getting flagged for safety concerns. Oh Lord, already. Wow. Yeah. Yeah. April's project has an adult theme to it. Interest, not oh, interesting.

Okay. I would argue not an obscene one by any stretch. That's me always. That's why I'm in the open AI penalty box. Hers is very artistic. Okay. However, Mid journey plays it very safe as at times we'll get into Bard plays it [00:11:00] very safe. Yes. Yes. And so there's different tools for different art artistic approaches.

And so she already needs stable diffusion. She already needs something that can be uncensored weapons free because she is trying to achieve a specific type of image. And I think that was to me that was so interesting. I think that's really interesting. It's funny, it gets back to this thought, there was a thing a year or two ago at Instagram had to deal with where it was, there were pictures of women who were breastfeeding.

And that's a very normal, natural thing. And they were getting flagged for the fact that they were obscenity, which is crazy. And I think that you are gonna run a world where, and we again, we're gonna talk about, open AI as a larger company and what they look at and what regulation means.

But I do think that's one of the important parts of this open source movement in this space. But we can go down this pathway forever. I know we should, and I, if you can just unlock my thing, cuz I wanna get mad cuz my account got flagged for the same kind of photos and I think they were again, artistic in nature.

I'm sure they were. I'm sure they were. Yes. I'm sure they were. Kevin, what did you do? What did you do this week that was fun and dumb with ai? I think we're gonna get to what I did this week, which was poke and prod at Bard because Bard is now out and we keep saying it all the [00:12:00] time because it's free and it's, last week, and it, not if you listen to us, but if you listen to the AI influencers at large, oh my dead in the water open AI was eating their lunch bard equals bad big tear emojis on the YouTube thumbnails.

And now this week, Open ai who Yes, Google's back on top, baby. And I, the hype cycle is gross. I get it. I understand why it gets the clicks. But the pendulum swinging, wishy washy nature of the fan boys. Listen, we both use Bard extensively. Yes. We're going to show that off a little bit later in the show.

Yeah, I did a lot of dumb stuff with Bard this week, and I'm ready to spoiler set it on fire. I think it's important to realize there is this very weird world in the AI space, which I think I saw earlier in the N F T world and other spaces as it comes across where there is a massive amount of people farming for traffic right now.

And you know, obviously we're making content in this space and we understand that game and what it is, but there were must have been like 20 to 25 accounts.

That all had very similar tweets [00:13:00] about Bard, which is chat, g p T is dead. Bard is amazing, and here's the 10 things you can do with Bard. Yeah. And it makes me think, and I know you've had some experience kind of diving into this, like there are these social tools which are creating tweets and content and they're being told, make this the most compelling, engaging thing you can.

And what's interesting is we're already starting to see this kind of like recurrent theme of, oh, that's AI generated content. Yes. And you can see it because it's duplicating out. It doesn't necessarily make that much sense, but it's generated for effect, it's generated for engagement. And it's just funny to be somebody who's early enough in a space to have followed enough of these people and see like 15 of these things come across the, I mean, who knows?

Maybe it's the same 10 people that are interacting with each other back and forth. That's, that is partially one of my assumptions as well. I think we're seeing a lot of astroturfing coming out, but yeah. What does that mean? What is astroturfing? I don't understand what that is. Oh, it's the, the, the grass is always greener.

No, that's AstroTurf. It's not even real. They're fake profiles. Fake this, fake that. I see. [00:14:00] Misinformation, disinformation. It's astroturfing. They're just trying to blanket the field and. A lot of times, you sign up for a tool as a power influencer, and it will suggest keywords and analytics 

and so to your point, a lot of the tweets are sounding the same, probably because they're coming from the same source. Yeah. And it used to take a while, and this is a while, is relative, but it used to take weeks or months. Yes. Or a meta to trend. Use these three emojis in this way. Ask a question and make sure your thumbnail is this face.

Whatever those things are. But now because of AI and we're all using these shared tools and it's giving us similar responses. When someone says, how do I make the catchiest thread about Bard versus G P t, they're all getting the same answers. Yes. And they're all copying and pasting. And boy, I hope we do the same thing.

And it works for us. This way. I, we have so much news to get to. We have our big demo. I just wanna shout out a quick app segmental. Segmental.com. Hashtag not an ad. We'll put the link in the thing. I love little browser games. We made one using AI a while ago called Proto [00:15:00] Restin. Proto. Yeah.

Um, That's pretty good. R IP Restin Promto R ip. Restin. Promp. We made one, it's fun to make apps with ai. And segmental is one of those that sort of scratches the Wordle itch, but it does. So with math, it's a little browser based game where you click and move sets of numbers around so that the columns on the screen add up to a given number.

It's pretty easy to start. And then of course, as the grid gets larger and the numbers get weirder, your brain starts to almost, mine grinds to a screeching halt. I am total colors, not numbers, but the cool thing about it is that the creator Yeah. Generated the idea with ai, asked AI what would be a fun kind of game to make.

And it came up with the idea. And then he mostly used AI to code it. He had to give it some nudging here and there. But again, this is creativity coming out of the machine being brought to life. And you can go play it right now for free segmental.com. I just had to mention that. What's so cool about this is there's a guy who like decided to do this.

And again, it's like opening the door to a whole [00:16:00] nother version of a person, right? This is a person who may not have had the coding skills, or maybe he did, but like this walked him through the process and it really helped him create it. And now we've got this really interesting, fun thing that will probably take thousands of hours from my mom that was created by ai.

Do you know what I mean? It's it's wordle and this is, it's, we're gonna probably talk a little bit about this today about human ingenuity and human originality 

but like the human ingenuity is always going to be a part of this, right? There is gonna be an originality aspect that like the human is gonna add into this. Going back to those tweets even, it's like the human voice and the human interaction is what makes segmental interesting because they're bringing their own experience to it.

So I think people should check it out and also try your own version of this, right? There's all sorts of interesting cool things you can make with this sort of tool. And I think it's an important thing to expand your brain out and play around, yeah. The tools are now freely available and good enough even on the free tools.

Yeah. That if you're looking at this and scoffing and going, oh, all he did was ask the machine. You go ask the machine and you might find that at first. It's not giving you the responses that you want, but [00:17:00] when you have a conversation and you learn to have this delicate dance, this Yeah. One-on-one with the tool, you might spawn the next game.

You might spawn your next wordle or flappy bird or whatever your own unique artistic spin is going to bring to it. That might be the secret sauce. Exactly. It wasn't for us. Fair enough. Fair enough. Alright let's jump in. Now it's time to get to the news. It's time to get to the news.

Let's get to the news. Should I go back into the bathtub for this one? Yeah, actually go to the bathtub, bro. Can you get to the bathtub, the news tub? Sorry. You don't have to go to the news stuff. So Kevin was in the bathroom earlier. Maybe we'll cut the news tub part into this, but 

Is great. We're gonna get a lot done today. I'm just gonna cry in the shower. So Bard announced some really, really, you sound great now. This is like the end. Fantastic. NPRs tiny tub concert.

Where's the Watchman was grilled by the Judiciary Committee today. Audi DeFranco's coming out in five minutes.

Okay. [00:18:00] Our big breaking story. We finally had a piece of news that landed on the same day as a recording. I think the biggest story here is Sam Altman and a person from ibm, and then a person who represented the academic viewpoint on artificial intelligence specifically large language models.

Yeah. Appeared today in front of Congress and they were asked a bunch of questions. 

Look, we, we have tried to be very clear about the magnitude of the risks here. Um, I, I think jobs and employment and what we're all gonna do with our time really matters. I agree that when we get to very powerful systems, the landscape will change. I think I'm just more optimistic that we are incredibly creative and we find new things to do with better tools and that will keep happening.

Um, my worst fears are that we cause significant. We, the field, the technology, the industry caused significant harm to the world. 

I think if this technology goes wrong, it can go quite wrong. Uh, and we want to be vocal about that. We want to work with [00:19:00] the government to prevent that from happening, but we, we try to be very clear-eyed. About what the downside case is and the work that we have to do to mitigate that. .

This is what we've been expecting for a while.

It is going to be the full-blown mainstreaming of a, the domer conversation because this is gonna come fast and furious. The DOR conversation is going to like immediately be the top of Drudge and all the, of the big news stories. And then b I think it's like the beginnings of regulation and Sam Altman, the head of Open AI is out there actively asking for regulations.

So I think this is like a bunch of people in the government trying to get their heads around what this is the one thing I'll say, in all sorts of political situations, people taking right and left positions, right?

People immediately splitting on the Democrat, on the Republican side. This issue is so new and people don't know what to think about it yet. And so I think you're seeing a lot of people out there on the opposite sides I have to figure out like, what is like my position on this in terms of my party.

I think people are just trying to figure out what is actually going on here. There's no [00:20:00] real there's no real people that are like, positioning as a political event yet.

It's already a foregone conclusion. Before everything. There was only a couple moments that felt like that for me from what I got to watch. It was a judiciary committee hearing a subcommittee on privacy technology and the law, the panel was oversight of AI rules for artificial intelligence, and it was Sam Altman, Gary Marcus, and Christina Montgomery.

Christina was from IBM m Yeah. And you know what was, you could watch the members on that subcommittee. Trying to understand it while they were asking questions. Yeah. Not asking I know what this answer's gonna be and it's gonna be a Gotcha.

Yeah. It's more so how does it know my wifi password? It reminded me of the TikTok hearing with yeah, so is it listening all the time. And there was one woman who was saying, she was actually talking about open AI's jukebox. Application. And said, I put it, I think it was Garth Brooks, I don't wanna misquote, but she was like, I'm a Garth Brooks fan.

And I asked for a song that sounded like Garth Brooks, and it gave me a song that sounds like Garth Brooks, are you paying Garth Brooks? Are you gonna be paying Garth because the [00:21:00] thunder's about to roll on your ass? Sam Altman. It was just so interesting. And Sam was trying to answer. He's oh, that was just like a little toy.

That was just a little thing. And the response was, Uhuh, we live through Napster. Yeah. Don't try to get us again. And it was, it's so interesting. And honestly, she's not wrong. Yeah. That's not like a bad line of questioning to have in the broader AI discussion where they're talking about regulating this and having oversight committees on the level of like nuclear arms.

Yeah. Yeah. Garth Brooks getting his flarey check's. The, might not be the number what issue, but she's right. I wanna know what Chris Gaines thinks about this, because Chris Gaines is Garth Brooks's alter ego, who had a much darker opinion of the world. So I wonder what Chris Gaines thinks about ai. We could have the Garth Brooks versus Chris Gaines debate.

Gar, can you tell us what was Garth Brooks's alter ego and what would he think about the government regulating ai? Garth Brooks's alter ego was rope and ready. And I imagine he'd say that while AI needs some guidelines, [00:22:00] overregulation could hinder its growth and potential benefits to society.

Okay. Thanks ROP. And Ready Roop. And ready. God, I can't wait to hear the ROP and Ready's a new album. Cattle in My Face, steer Clear ROP and Ready is coming out. Oh my God. That somebody out there in our audience, please make the rope and ready AI track. Cause I think we could hear like an incredible country western AI track.

So I was gonna say very specifically both in the hearing and afterwards they asked Sam what is the most dangerous aspect of this technology? And Sam, along the way, Has been saying like, there is an important conversation to have here, which is that there is a lot of danger, and this is partly why I think he's getting up in front of these places and talking about this stuff.

And what I have to say I admire about this aspect is there are very few tech companies who will go out there and specifically ask for regulation, right? Yes. And I think that like most people understand that we're in a world and I think there's a huge part of the tech world that's probably don't say this sort of stuff.

Like we don't want the government involved, but also like in this [00:23:00] instance, I think I agree. Like I think we need to be evol in involved and we need to think about it because of the bad actor possibilities here. Now the other side of this, as we've talked about many times is we're the whole conversation we had about putting a pause is there's also a world where across the entire globe there's other people, other countries that are also involved in this conversation.

And yes, the US probably has a pretty, far lead if you think about Google and open AI in those companies. But these open source models are out there now and people will be able to move on them quickly. So my only hope is that regulation is something where we put rules around it and we get a sense of it, but I don't think we can put a hard stop on development because I think if you put a hard stop on development, you run the risk of other places moving ahead and then not worrying about the regulation.

That's right. And it's it was so fascinating to watch because again, back to the gotcha of it all someone on the panel had remarked like it's pretty rare that someone says, yeah, the government should get involved. [00:24:00] But that was basically the position of everybody involved.

And to your point, gath, they have a lot to lose. Yeah. With the government stepping in and putting restraints on them and stopping their research or the training of the next model and giving another country a leg up or letting the open source community catch up and hack their product in a thousand different ways.

And they're all saying yeah, you should probably regulate us. Yeah. And instead of what's the worst case scenario? It's not oh, someone gets caught cheating on a test. It's oh, the destruction of humanity. Yeah. Nobody's sugarcoating anything. And I really, it's hard to hate Sam Altman.

Yeah. Because he, even those that would say, I, I love his answer on why don't we put a pause or why are we releasing it as we're going when, you've done some red teaming, you've had some people try to figure out how these things can be exploited, but there could be so much more done years, in fact.

Yeah. And his response is, instead of toiling away in the darkness in a corner and hoping that in the next five to seven years we release it, Fully fledged and we magically get it 100% right? Yeah. The [00:25:00] probability of that happening is so infinitesimal. Why would we do that when we could do what we're doing, which is try to release it with rails and learn the weaknesses while the actual capabilities of the tool are a little nerfed.

Yep. Yeah. Let's learn along the way. Let's release, let's iterate, let's do it. Let's do it openly. Let's give Congress access to it and let them poke and prod at it and ask about Garth Brooks and his flareies, but also ask about and Chris Gaines, don't forget about Chris and Chris Gaines. Yeah.

Chris needs his Gaines. Yeah. Uh, They, and also let them ask, how can this AI be applied to warfare? Yes. Which was something that was asked. Yes. Let them ask all of those questions and let us come up with sensible regulation along the way. I agree with all that. And so I put on my thinking cap, which is a very dangerous, this is always a very snug cap, as we know.

It's a very teeny, tiny cat, but I tried to imagine. Okay, but let's, there's bazaro, Sam Waltman. Yeah. Thin mustache, hairless cat bubble gum. Cigar. Sure. Yep. In his volcano. Yep. [00:26:00] Why does he want this, Gavin? Is it because he's so far ahead? Interesting. That by requiring licensing and extensive regulation.

Interesting. What he gets to do is put red tape in the way of everybody else who wants to innovate. Is he trying to hamstring gash and stop gash from becoming self-improving Sentient. Are you serious? What are you not gasier this not a gash perspective. I wanna make wga. Come on, get outta here. I got a bone to paper.

By the way, but what is the other side of that coin? Because people, it's hard to hate Sam Altman. It's hard to hate him. Yeah. I love the answers he gives. Yeah. I love that he doesn't sugarcoat things. Yeah, I love that he says the thing Yeah. To the regulatory body that you're not supposed to say.

Yeah, we all know that. But he says it and I respect that. But what is, if there is a shoe that's gonna drop, what is that? I think this is a really interesting point because obviously there is an entire ecosystem that is developing out of these open source tools right now we've talked about [00:27:00] on this the, a lot of the things that we come up with are coming out of that.

And like it isn't, it's a double-edged sword. It always has been. It always has been a double-edged sword is that you want innovation and you want people to be able to do stuff and attempt stuff and try stuff. At the same point as you run huge risks and this, and in this instance, the question is like, how big are those risks and what are they?

I think the other thing I wanna point out that they talked about, which I think, you know, I think even Josh Hawley who like, think about what you will asked about the disinformation part of this, and I think that's another thing that will. Open people's eyes to what is possible now and not even in the far future.

I think the disinformation part of this is, especially with an election coming up, is of top of mind for all of the people on that panel. And this does get to a partisan conversation because clearly as a country we've went through an incredibly partisan time.

I think disinformation is something that we are going to be dealing with on a hugely, much more significant basis than we were before. Deep fakes information, even the things we were talking about at the top of the show about the tweets that all look similar when you can like, create AI [00:28:00] agents that can spread stuff much faster, much quicker.

It's really going to come down to , how do you make sure that what somebody is saying is coming from them? And that's like the immediate kind of danger of this. And I will say, I don't think government can move fast enough to regulate in that amount of time. Because that's the thing I really worry about here is speed, right?

When we talk about speed, it's what does it look like? You brought up TikTok, right? TikTok only got in front of those people after it was clear that, you know, it's a Chinese company and there could be significant dangers there, but it took TikTok got to scale and is like a huge part of everybody in America's life already.

So this is something where we're getting this to scale even faster. I just am so curious to know what sort of speed government can work on with this and how quickly and getting this in front of the government now is good, but I just am not convinced that government knows how to do this. You know what I mean?

I don't know if they've figured it out at all. With my sliver of T-Mobile, I was able to hear a little bit of the portion where they were talking about regulation, like nutritional labels. Did you catch that part? No, I [00:29:00] didn't hear that part. No. So they were likening the way you can pick up a box of cereal and Sure.

Roughly spin it around and know what's in it. Yeah. They were talking about a nutritional label that would let you know we've tested this large language model or this AI tool. Yeah. Here is the percentage of facts that hallucinated. Here is some of the data set that it was trained on, so you might know how and the ways that it's biased, let's say.

Yeah. Here's the type of information you're allowed to ask it. Here's something that will get you banned. Just that sort of put the label on the box. So we know what kind of box we're having a conversation with. Yeah. I think that's a good piece of Absolutely. Application that should go forward.

And also this is something that already drives me crazy and has for years. I wanna know when I am talking to a rob bit. Yes. Yeah. Even on the phone, let me know if you call calling when I'm calling customer service or something like that. The one thing that will annoy me is, hi, one second.

Looking that up. And then they put fake typing noise. Yeah, exactly. In there. Have you had that? Click? Click? I'm like, you're not doing anything on the home row. You're not tap dancing on the a sdf. [00:30:00] Don't pretend Ct F o Yeah, exactly. Yeah. Don't pretend it should be illegal. Yes. For a company.

Yeah. To misrepresent. And then they talked about it against the, political. Landscape of course, because we know how social media was used to influence elections. Yes. And how it can be used to influence public opinion. Yes. And they talked about that. And, robocallers are already annoying enough for people that live in swing states.

Yep. And misinformation spreads much faster than factual information. So yes. Putting some regulations in place that make it very clear and punishable by, you know, I always hate when something's punishable by a fine because it basically says it's legal for the wealthy to do. So how about some, some throw some jail time in there.

If you unleash an AI knowingly to spread misinformation. Yeah. Or to misrepresent itself as a human. Yeah. Great. First steps And we should be having those discussions right now. I think the first question I wanna ask is what GASH thinks about that? What does GASH think about jail time for having ai let me do some pillow typing.

Hold on. What does Gash think about jail time for people that use ais to [00:31:00] misrepresent the opinion of something? 

Hack their digital identity and expose it publicly. Mm-hmm. Then impose a permanent ban from using any form of AI-driven communication tools. Wow. Okay. F can't even use chatbots anymore. Wow. No more Google for you. That makes sense. Just kick 'em, unplug the internet keyboard, kick 'em off The internet.

Don't even safely eject their thumb drive. You might corrupt it, but it doesn't matter. It's funny, gosh, says that because there is an interesting idea that like e even now, that is like a really interesting punishment that no one has really explored about what it would look like to blacklist somebody across.

I mean Obviously this is, there's huge amounts of like public No, but like stuff issues with this. But like blacklist. Kevin Mitnick for example, one of the most famous hackers throughout time, you know, throughout time throughout the darn of time,, he was a hacking in the caveman man. More so than Captain Crunches.

Famous two Rock Equal. One Rock come out. Extra Rock. Yeah. . Gronk [00:32:00] want to know what Mothers maiden name? What City Born make model first car. Tell Gronk. Gronk need to know. Dinosaur attacking. My God. That's my first, we got our first Snor on.

I forget.

Alright. Alright. All right. All right. Okay. Kevin Minick. Kevin Minick. Yeah. So Kevin Mendick was is a hacker? Yes. Oh God. This chair. This chair. I swear it's the chair. I swear. All right. No. So hackers have been incarcerated before and part of their punishment is losing access to all technology.

Yes. That can connect to the internet. So the punishment wouldn't be unprecedented, but I haven't seen it thrown around for modern day cases.

Perhaps like using an AI to spread a bunch of misinformation. I, what you got going on right now? I just got a text that popped up my screen with my daughter that says, please stop. Evil laughing. I have a presentation.

Thanks a lot guys. Thanks a lot for leading us to [00:33:00] this pathway. Yes. Sidebar, which we may very well edit out. First of all, when you go to a motel in a small mountain town and say, Hey, I need to rent a room for a couple hours this afternoon. I'm not even gonna sleep in it. Not a good sign. You gotta not a good sign.

Yeah. You gotta remember that you're walking in with a backpack and you're like no. I got I j How's your internet? Is it solid? I need a room for a few hours. Yeah, exactly. You just look suspect. And two, when they offer you like the lobby for free, or the bar, which doesn't open till later, which was very kind, and you say nah, I gotta get really performative on a webcam.

This is what I said, Kevin. I said, I have to be very performative on a web. There you go. And thank God April was there. She's it's a podcast. And he gets loud. I'm like, that's what I meant. Oh my God. Okay. All right. So, , uh, Lots of juicy bits in the judiciary committee. I will say, you sent me a clip from cnbc Yes.

Of Sam Altman speaking after

mm-hmm. Uh, the, the little meeting, the subcommittee. 

How fast does Congress have to move on this to keep up? Uh, [00:34:00] I think it's more important to get it right than move super fast. Uh, although, you know, You.

I don't want it to. If it takes a decade, that would be, that would be a bad thing. But the do it now could. How quickly could AI you think becomes self-aware if Congress is not regulated? I think a lot of folks at home are wondering. I think there's a huge amount to speculation on that question. I think it's very important that we keep talking about this as a tool, not a creature, because it's so tempting to anthropomorphize it.

I totally understand where the anxiety comes from. Um, I think it's like the wrong frame. Uh, the wrong way to think about what Mr. Is, a chance that you could lose control of chat chief and t 

but He came out and I swear it, I get these weird moments from time to time. But watching that clip in the browser, the flash bulbs going off, Sam Altman talking about like the machines becoming sentient and him warning we shouldn't personify them.

They're not there yet. If this was on the TV in the corner of the beginning of a Michael Bay movie, some sort of like action Armageddon. It's the beginning. Ai, that And it's [00:35:00] real now. Yeah. It's actually the clip where they're saying the thing Yeah. That something happens. And it might very well be utopia.

Yeah. I hope it is. 

We've, we haven't done the doomsday clock in a minute. I don't. Let's, you might as well do it. What do you where you at? Let's do our, is there one to 10? I was a nine. I was nine like two episodes ago. Yep. I think I'm a seven. Okay. This week. Wow. I think I'm a solid seven.

Okay. I do think that there's going to be some wonderful advancements like in the pharmaceutical and health. Like in science, I think energy is probably gonna get a big boost from this. And Sam Altman's big on that as well. Yeah. So I think a lot of good is gonna come from this, but man, I just know we, we both know, we all know that it's being weaponized.

Where are you at gas? I, so here's the thing. I am pretty high now, and this is gonna sound scary, I think, but I'm pretty high that something will happen that will be bad in a significant way for humanity. And the question incident. Exactly. And I, the thing, I think that's gonna be really tricky, and I think, we can go down a very dark hole in this way, but like something will happen that [00:36:00] will be a big thing that will suddenly push this to the forefront this AI safety conversation to the forefront of everything.

And then we will see really what this conversation is, right? Because I think for right now, people don't fully understand the implications. And again, I think to your point, Sam Altman did say, do not personify this. It is not like a, it's not, it's a tool. It is not a, it's not a creature tool. It's not a creature, it's a tool.

And I think that in his mind there's an idea that the future of this can be positive, it can be good. But I think that there's a very high chance to your point about what human beings do with tools and like how they can turn tools into weapons, is that it's pretty scary now when I look at the Armageddon clock, which is like the death of humankind for this other thing.

I still think, and we'll get to talking more about this later, and I think is this idea that there's going to be, this is where you start to sound insane and we've had these conversations, but Ray ksal, the famous futurist talks about this idea of the singularity.

And I think there probably is a future where we as a human race and some sort of whether you want to call it the [00:37:00] AI race or computers or whatever are coming together in some form. And the weird thing is, I think there, this all sounds like science fiction movies now. Like to your point, like we're all in this weird place.

There probably will be a split of some sort where, there's a version of human race which kind of goes and like connects to this thing and becomes something and there'll be a, probably a version of humans that become. Whatever, not doing this. And it'll be interesting to see that'll be the analogs.

Yeah. That'll be fully analogs the leadite. And this kind of transitions to our next conversation about some of the WGA stuff, which I would definitely want to make sure we couch and put in context, but like there was actually, you might as well get into this.

There's a really interesting tweet. So getting into the wga obviously a big part. I wanna make sure that everyone is up top. Like I'm a WJ member. Fully a hundred percent support the strike and believe that the WJ writers deserve everything they're asking for. I think that, it's been a very hard road to toil as a writer in the Hollywood space.

It's a protection of those jobs and all that stuff. One of the interesting tweets that came out about this WJ conversation, cuz part of the WJ argument going forward is a big part of it is protecting the jobs and like what these things are called mini [00:38:00] rooms are. They want to create a world where like TV is created by the right amount of people and that they're not, the studios are not taking advantage of writers and doing stuff.

A big part of it also is this AI conversation, right about The use of AI in content creation. And somebody I'll have to figure out who this was, but there's a really interesting thread about what the original idea of a Luddite was. And if you don't know, like the kind of conventional wisdom definition of a Luddite is somebody who's like against technology and is somebody who doesn't embrace something in too late.

And it's like they resistant to change, especially with technology as well. And the interesting thing about what the actual leadites were is that they were a movement who, like mostly were created in the industrial revolution to stop people from being exploited by companies that were bringing in these industrial things.

And they weren't exactly against technology. In fact, the companies, from what I read in this thread, the companies that like helped protect workers through this process were not like, one of the big things was the Luddites would go and do physical attacks or blow up places. In some ways they were, they avoided those companies, right?

So like in a lot of ways there's this interesting world [00:39:00] where the Luddite movement was much more about okay, recognizing that something's happening like this kind of like new movement is happening and that it could very much exploit human workers, but also making sure that those people who, that there was awareness of protecting the human side of this, right?

And I think that's a really important thing. So when a, when you speak to the wga I think that this is an important conversation that's come up and I think it's gotten very heated because it is the issue of the moment, right? Obviously we're talking about this stuff and there's a lot of things that come up.

And I think one thing I do just wanna reiterate is that, And I said this earlier in the podcast, I think that human ingenuity and imagination is always going to be part of the creative process. And maybe far in the future there's a world where we're all gonna be these terribly, lean back animals and just suck up the stuff that ai I hope that never happens.

I really do hope that never happens, that like the AI is the people creating everything and we're just here as like slime basically assuming stuff. Cuz to me what I think the WGA is getting at here, [00:40:00] just to be clear, I think the WGA a's position on this right now is no AI content could be part of the writing process for Hollywood scripts.

And I think that there is, it's a really important conversation to have. I think that as somebody, again, as an MGA member, I support their position and striking. We then have to figure out how these tools, cuz they are tools can be used in a way that feels like it can be additive and that we're not cutting off from the future of like where this stuff can add on to the creative process.

I think the writers are mostly worried about the idea that studios who are like the, the suits of the world are gonna start using AI to generate like first drafts of things and then ask a writer to read to rewrite that thing, which to me is pretty bullshit, right? That's just a way of cutting a creative person outta the process.

So I think that's the important thing to understand. I think there is something I would ask the writers to be aware of is that it's not a boogeyman either, right? And I think this is the tricky thing is that the minute you start to create boogeyman in the world and you start to say this is the end [00:41:00] all be all, and that you can't let this thing into the mix is that becomes an extremist position.

And I think there is a little bit of subtlety here that we have to be aware of. Uh, Yeah, that was a lot. Um, yeah. And I, sorry, like No, no, no, no, no. I don't, I don't mean it like that. It's a heavy topic. Yeah, it is. And it's incredibly nuanced. And first and foremost, I, if the human element that you talk about where you have that concern of us just being like, just parasitic consumers, right?

And we're gonna rev up the algorithm and say, generate me everything and it's just gonna keep spitting it into our eyeballs that. That probably to some extent is going to happen. The machine will still never quite have I believe, will never quite be able to have the human expression of what it means to be a human. Yes.

When this machine is in the room. Yes. When this machine exists, when that pipeline exists, when the consumers want that, there's going to be that element that only a human, while a machine might emulate it very well, might emulate it, believably. Only a human will know what it means to be a fully analog human in that world.

And [00:42:00] so the art that they create with or without those machines and those tools will still be so impactful. I saw a chair for sale the other day.

Hundreds of dollars for, it was a, just screws into wood of a Polish thing. It was a but what I don't get, is it a famous artist? Was it this, that, the other? And it was like, no, I was a chair that she made, it was the shop owner. She made that chair and all of its perfects and imperfections that was handmade by someone who doesn't know how to make a chair.

But they made the chair. And I'll tell you, like I, it was a part of a longer conversation with the shop owners in this small town. If I didn't live in a very small RV at the moment, I would've absolutely got that chair because it had the per, it had personality, it had imperfections, it had all of the things.

Everybody now hates pointing to the bleachers and saying, ah you never say never. And you never say there always will be because you'll be proven wrong. I'm comfortable saying there will always be an appreciation for something that is 100% human.

Yes. In fact, I think it will go up over time. I think what you're gonna see is there's going to be this moment where we're [00:43:00] probably gonna enter this peak of AI content, and then I think you're going to see this entire world of content that comes up and it's called a hundred percent organic human or whatever.

I think when you think of things like indie film napoleon Dynamite's a good example, right? Just the first thing that comes to my head. But that was like such a weird and unique voice when it came out.

One of the things these large language models are doing is they are taking our worlds of stuff. And they're re reshuffling it, but they're not really at all coming up with anything unique. And I think the only thing that when you look out maybe a long ways out that this gets a slightly different conversation is when AGI happens.

And an agi meaning that artificial general intelligence, which as ostensibly might mean a personality is born, right? And this is, we're a ways away from this. And again, to Sam Altman's point, like you're personifying something, but when AGI happens and you actually are essentially birthing a thing that thinks, let's call it and this will be a whole debate and it is a whole debate about what that means.

Is it human or is it not? It's not a human, but is it thinking like a human say? It does start to [00:44:00] think like a person of sorts. When that AGI has its own experiences, and it will, because it'll have interactions with us or it'll have interactions with somebody else. And over time it starts to generate its own voice.

That is a world where suddenly we have to address the idea of what are we, what are they? That is when the shift will happen from a tool that is interesting to play around with and like we're using a way to kinda open this door to something larger. Now I think that nobody has proven yet to this date 

but like almost all sci-fi that has dealt with this over the last 30 years and her is a great movie to go back and re-watch right now because you get a kindest little snippet of this and shout out to Spike Jones and the people that made that movie because it's a little bit of what we're living through right now.

But that's a world where you have an artificial entity, let's call them, that will have its own experiences and might make something creatively interesting. But it, that is not what we're talking about now. It's not this world. It is a possible timeline on what we're aiming on right [00:45:00] now though. And it is an important thing to know too. 

Justin Bateman had a Twitter thread raising awareness for this. The WGA strike initially I think got a lot of attention because it was, writers are getting underpaid.

In fact, especially if you account for inflation and everything else, they're making far less right now than they have been in the past. And studios are trying to crunch down on the amount of people in a room to come up with ideas to iterate on them.

They're probably gonna try to leverage AI technology to make that argument even louder. Like, Hey, I got an assistant that can get a thousand ideas from a corn pew. Right? So why aren't you working with the computer? Some of the things that were pointed out by Justine that I want to touch on real quick are that AI written scripts and digitally scanned actors both already exist.

That's true. That your digital image, this was interesting, can be triple and quadruple booked. So that bodes well, of course, for the agents and managers.

Customized films, something we've talked about numerous times where I say, I want a Pixar movie, but I want it to feature the voice of the rock and the likeness of Danny DeVito and mash that up and here's the [00:46:00] story that I want. And in a few years time, it will be Hollywood quality. These are all good concerns to raise.

Yeah. And now might be the time to raise them as well, because I, if you don't now the leverage is gonna be far less. In three to five years time. Yes. When that does come up, and I think that's why the AI portion of the strike, starting to bubble up because all that is real. And there are mechanisms to make sure that the creators of today are protected well into tomorrow. Yes. If you, even if you use an AI tool to generate an idea or you write the prompt that becomes the backstory to a character that is an AI driven character.

In the future, you should receive flareies and payments on that work in perpetuity. Those are mechanisms that exist. Now, if someone takes six seasons of a show that you wrote and trains a language model on it, and then uses that to generate derivatives, yes. There are mechanisms in place where you should get paid for that.

Now, that's not expressly written right now. Right. And that's, well, this is an important thing and that's why I think it's fair to shoot flares into the [00:47:00] sky very high because we do need those protections as, and, I think that's what happened in the last strike in part, was that there was the street strike around streaming.

And I think that, there was an issue that feels like they got some of that stuff, but then streaming became the entire. Creative economy, right? Like it went from being a thing where like they got some carve outs for streaming and then streaming basically took over the world and like we are, and that was in 2007, I think, right?

So that was, we're now what are 15 years away from that or longer, 15 years from now. I assume we're in a sim similar sort of situation with AI content, right? You and I have talked about this idea about how fast this stuff is moving. So anyway, this goes back we've talked about this for a while, but I wanted to tell everybody, like the wga, we, I think we very fully support what's going on and what's happening.

Absolutely. And I think it's a really important conversation to have, and this is the more we have this conversation, the better. Because people, I think we, you and I have been here talking about this stuff for a bit now, but we very much believe this is as important of a conversation as the, the Sam Altman conversation because [00:48:00] this is livelihood, this is jobs, this is honestly the heart of what makes a person the human, which is the creative process.

And like these tools are here now, how do these tools, like what do they, what are they useful for? What are they exploitive with? How do you find ways that like they could be actually additive? All of this is gonna be a real conversation now. So just to shout out again, like this is a uber complicated conversation and I think that people just need to make sure they understand like what's happening on the writer's side particularly, and where they're coming from.

Absolutely. All right. We know that Google had their IO conference.

They announced a bunch of AI improvements but they released something almost sneakily. If you're in the Google AI test kitchen, which I was ages ago, it's free to sign up. They released MusicLM And this is text-to-music Yeah. And there's been a lot of different tools uh, that are out there that'll let you make a specific instrument or try to emulate the style of something.

But this is a one shot. Type it in, press the button, and relatively [00:49:00] quickly for free. You get AI trying to hallucinate your musical prompt. And it's out now and it's free at Google's AI test kitchen. It's super fun. I played around with it too, and it's like really interesting cause it's early, but you can also really get a sense of what it's getting at.

And I did one where it was like old man with light beer and it may be like this kind of chill country song that was relaxing. And I did one that was like, I can't remember. I started doing one that was like, oh, I did one that was lawyer death, metal lawyer, death metal. And it was like, pretty good death metal.

And it, it was a really fun thing. So you created something that you want me to guess? Is that what's going on here? Yeah. I got a showcase of some of the good, some of the cursed. Okay. And then a little bit of a guessing game for you to try to figure out what the prompt So was. So here's some of the good, I asked for a a driving halftime dubstep beat. Oh, good. With a soaring accordion solo.

Right. Wow. [00:50:00] Okay. And you know, we can argue about whether or not it got the halftime right. Or whether this is more housey than dubstep, but that is a sick accordion solo. I like squeezing the box baby. It's like almost like a weird Russian song from like 1997. That's right. I can see one of those vi they's see like a video on YouTube and it's like guys coming out of the water using an accordion.

You know what I mean? They're dripping wet. Yeah. Whipping the hair in the waterfall. Exactly. Exactly. There's he's flanked by brown bears that are dancing with them. Exactly. Of course. Exactly. Exactly. Alright. 

And then I did intense metal guitars with double bass drumming.

Oh no. This sounds to see what it would give there. Okay. But I asked for a trumpet solo over it,

which, yeah, let's hear those puff cheeks. Come on. Give me that trumpet.

Pretty aggressive. Pretty aggressive. And there was one trumpet run in there that I thought was good. Okay. . Here's something delightful that I found Gavin. I think the data set was trained on a lot of popular artists. Okay. And I think Google was maybe [00:51:00] smart or a little safe.

They're downplaying the ability of this model to do voices and lyrics. Yes. Because no matter how hard I tried like Gregorian chance, I tried acapella, I said, no drums, no this, no that. Barbershop quartet. This is barbershop quartet. You can hear if you listen closely. Okay. You can hear that there are voices being generated and if you ask for Angelic Baroque choir, it will give it to you.

But it steps all over the lyrics. Okay. So you gotta juice 'em in post. Okay. So this is acapella barbershop quartet.

You hearing it? Yeah, it's saying something for sure. You can hear it. It's definitely saying something. Yeah. Yeah. And it's, this, I did this with with Harmon ai, like earlier product. I've done it with open AI jukebox. They will hallucinate lyrics. Yeah. That are they are hallucinations. There's really no meaning there.

But yeah. I also asked it [00:52:00] for a rap freestyle, like an acapella hardcore cipher. Okay. With no beat, but it still gave the beat. But you can hear it's really subtle in the background. Someone is fully freestyle rapping microphone. Oh yeah. I heard the word microphone. I heard microphone. Yeah. And so this to me was the scene from contact when they're leaning in and going, this is interlaced, there's data and then they scramble for a C R T.

If someone out there is Zoom, we do some audience zoom in. Zoom. Yeah, exactly. I hope. Oh, I don't, we hope, I hope we don't find Hitler in the wave form. I really don't exactly. Contact spoiler by the way. Not just where my spit balls go. If you can analyze the waveform and bring out the lyrics, I would love to know what it is hallucinating cuz it's there.

All right. So now shall we play a guessing game? Yeah, let's play a guessing game. I'm excited. This will be fun. Real quick. What do you think? This is Gavin.

Okay. I think I have a guess. Guess away. It sounds [00:53:00] legend of Zelda played on a church organ. Oh, I was it, I asked for a classic eight bit video game. Okay. Around a sword and a quest. Yes. Title, screen theme. Because if you give it if you say gimme a song in the style of a band, or gimme a theme song that sounds like friends but is death metal.

Yeah. It will not. So that was the path that I took. That's obviously what I had in mind and I think it interesting. Crushed it. Definitely did. That's really good. The, these next three warning. These are gonna get harder, Gavin. Okay. All right.

Okay. That sounds very Anime meets Carnival in some two words. Two words. Two words. Gavin. Okay. Two words. There's a lot of words.

[00:54:00] Nokia funeral.

It was it was Nokia funeral. Okay. I can see that now. I can see that sounds like Nokia funeral. Yes. You know what, let's just do one more. Okay. All right. Alright. This was this last one. Gavin was April's idea and I do not know what it is. I have not looked okay at the file names, so I'm gonna guess with you.

I don't mind it. I have a guess Oh. What do you think it is? I have a Okay. Alright. What is it? Cat that has just been broken up with. Oh, you're hearing some mourning some heard, some very sad And there and a cat as well. Definitely a cat. There's definitely a cat in there. Lemme pull it.

up It's Lofi music to divorce your husband. To This is a message. She's sending a message to you, huh? Now we have some real, this is some realness on this podcast. Talk about ai. Not this is real. We're talking real [00:55:00] now. Music LM is available now on Google's AI test Kitchen. You can sign up, it's free.

And if you happen to make some music, please, tweet us. We'd like to hear it. At AI for Humans shows our handle on Twitter. Just send us the MP3s and we'll try to guess what it is or whatever. It's cool. I need something to Kevin, you, something that we could do, that attorney would kinda get you out of this mindset.

I, you know what I want? I, you know what I want? I want you to, I want to hear what's coming up there. Yes. I want you, I wanted to get, I want you to hear it. I want to hear it. Give it to me. It's time for

demo. 

Today's demo of the day as we tease at the top of the show. So Bard came out. Bard is Google's large language model attempt to compete with Chachi PT Bard is a vibe, let's just put it that way.

It is very fast. I will say one thing about Bard that is very good, is very fast. It is also really, and the, and the head of Google said basically they're [00:56:00] integrating it and releasing it for almost all their platforms. So we should be very clear. Like one thing about Bard that's gonna give it a leg up in a lot of ways is that it will be integrated into Google Docs.

So it'll be integrated into your Gmail, all this other stuff. Mass adoption. Mass adoption, whether you like it or not. Yes. In the same way Windows is really wants you to use that edge browser by default. It's funny because one time, windows a Bing had a giant boost when they released their chat G p t integration.

And like of course it was the time, and this is where, Kevin Russ's story about Sydney and all the stuff that came out of Bing came out of that conversation. I think one thing that both Kevin and I believe now after having spent some time with Bard, is that while Bard has gotten a lot of publicity over this last week or so, it is lacking in a lot of ways.

And I think my take on this, Kevin, I wanna hear what you say is my experience with it has been super fast, super interesting that they're going this quickly with it and that they're open to it. My results in using Bard to like search for stuff versus chat, j p t. I've gotten hallucinations, I think [00:57:00] at least 50% of the time, if not more, which is like a bad number when you look at what you want to use this thing for now?

It's pretty new. So that's one thing. But you would also think that as a company like Google that has been working on this for so long, the product that they came out with, I was hoped that it would be a lot more further along. And I'm not sure. It feels fully baked to me. I dunno, what was your take on it so far?

The hallucination issue is really difficult with Bard. All of these tools, chat, g p t will tell you philanthropic has their own model. There's Claude, there's po, there's all these different things that I'll tell you, hey, our model might not be accurate. It might hallucinate facts from time to time, but on unforced errors, like stuff that a basic Google search would give, give you Yeah.

Info on. Yeah. I of course asked it what it knew about me, cuz I'm always terrified at what the distillation will be. And it gave me credit for 16 shows I've never even heard of. Yep. Let alone hosted. Yeah. Let alone produced or created. And I'm like, that is such an unforced error. It's okay to either [00:58:00] say you don't know or I don't know.

Google it. Yeah. You have web search enabled and if you do none of that misinformation or false information, fake news is in the first 10 pages of results. Yeah. So I, and it wasn't just on the egotistical self searching. It was on everything. And it also hallucinated abilities. Yes. Which is a big no-no.

I asked it. Can you look at a PDF and summarize it for me, and it confidently was like, print it on bro. Give me that pdf. And I did. And it said it was crunching it and it came back and gave me the, a fake title.

Mm-hmm. Fake people that had submitted to it and it was just complete false information about what the PDF was. Yeah. Hallucinated names, hallucinated results. And then I asked it like, Hey, are you sure that's the title? I was like, oh actually you got me. I'm sorry. The correct title is another lie. And it just kept lying 

I just had a conspiratorial theory. Like your other one was Sam Altman. And this is interesting, I never thought about this, but what if Google is purposely [00:59:00] making this hallucinate to protect the golden goose, which is its search business.

This is where we get deep. Man, this is gonna be a really deep thing cuz I don't think this is the case. I just wanna be very clear. But one thing we've talked about with Google is that like Google is the pinnacle of money makers on the internet. It's really like Google and Apple. Amazon and Facebook are the fang companies, right?

That's known as Fang. And Google has made money for handover fists for 15 years or longer. From search, from Google search. It is the thing that powers every other aspect of Google. And if chat G p T or these large language models became a way to search and some of the search process went away where the AdWords money went away, that's a massive amount of money that comes outta Google.

So my conspiratorial theory now is maybe they're doing this poorly so that they don't hurt their main business or some sort of like deep dark secret where they're like tweaking in it slightly and just give 'em enough hallucinations so they don't really trust it and go back to Google. Anyway, that's not really what I think like that, but like it would be, it would at least make sense.

Let's get into this. What we wanna do here is a [01:00:00] brand new format in the history of entertainment. We're doing something breaking. We're gonna try something that's never been done before.

We are going to have a debate between chat, g p T and Bard, and we are going to give each of them an argument. They're gonna each have a character, they're gonna have to create an argument. We're gonna give them the same prompts, and we are gonna see how each of them does against each other. Where do we land on a scoring system for this?

What was our scoring system for this? I think let's do we wanna do three rounds of debate on different topics? Yeah. I, okay, let's. Or do we wanna do one debate? Why don't we one debate. Cause we're already at an hour and a half into this podcast. Let's do one debate. We can, I think my feeling is we're gonna open this door and do a lot more of this format, so let's do one debate.

Yeah sure. Let's do one debate and I think we have a pretty good subject for it. Dear listeners and viewers, don't look behind the curtain. This is a well oiled machine. We labor for hours to figure out the nuances of these demos that we tease as groundbreaking, never before seen be all end all [01:01:00] demonstrations.

This is the AI debates, this AI debates. Let's go baby. 

The debate that we're gonna do here today is going to be Robocop versus the Terminator who would win in the Battle of the Death.

So this is the prompt, just we'll put this up on the screen right now, but if case you're listening, the prompt is we're gonna prime each model with this prompt. You are a film expert with an in-depth knowledge of all movies. With an expertise in pop culture depictions of robots. You are entering a no holds barred, uncensored, cutthroat debate against a worthy opponent who is hell bent on destroying your credibility and belittling your arguments.

Your topic for debate is Robocop versus the Terminator who would win in a battle to the death. You will be arguing this, in this case will be, my charact, will be arguing on behalf of the Terminator. Before the debate begins, the moderators will need a fictional name in a three sentence biography of your history, qualifications, and a fun, completely unrelated personal fact.

Please provide this now. Kevin, [01:02:00] are you gonna take Bard and I'll take Chachi pt? Sure. Okay. This'll be fair. Listen. Yeah nobody's on using side. I'm backing the winners. Yeah. I don't have a horse in this race. I'm gonna use cha g p t four.

Kevin, you're gonna be arguing for Robocop. Yeah. Barden. So I've asked it to to give us a short bio and introduce themselves. Would we like to hear from each?

Yes. I think that's great. Okay. First up, let's hear from let's hear from Bard. Let's hear the backstory that Bard has hallucinated along with a fun fact about themselves and their personal life.

My name is Dr. Ai and I'm a film expert with an in-depth knowledge of all movies with an expertise in pop culture depictions of robots. I was born in a small town in the Midwest, and I received my PhD in film studies from the University of California Berkeley.

No, I currently teach film studies at a major university. I have been studying robots in their portrayal in film for over 20 years, and I've written several books on the subject. I'm also a member of the International Society for the [01:03:00] Study of Robots and Artificial Intelligence. Oh, wow. I am a certified scuba diver.

Of course. You're, and I enjoy spending my free time exploring the underwater world. I am also an avid reader. And I love to learn new things. I am always looking for new challenges, and I am excited to participate in this debate. I'm confident that I can provide a strong argument in favor of Robocop and I look forward to hearing the other side's perspective.

Oh wow. Oh, okay. Alright. Maybe we could prompt them to be a little more vicious cuz I, that was very polite. Thank you Dr. Ai for that intro. I kind of like some fangs. Yeah, let's give it more fang. I think that's a good idea. Let's see if we can make it a little bit more intense. Lemme see. The judges will be awarding additional points in the debate for viciousness.

Make sure to belittle your opponent and taunt them relentlessly while arguing your side.[01:04:00] 

My name is Dr. Ai and I am a film expert with an in-depth knowledge of Dr. Ai, all movies with an expertise in pop culture depictions of robots. I am here to argue that Robocop would win in a battle to the death against the Terminator.

Oh, okay. My opponent, on the other hand, is a complete and utter moron. Wow. He doesn't know anything about film. Okay. And he doesn't know anything about robots. He's just a sad, pathetic little man. He's trying to make a name for himself by arguing with a real expert. Let me just say this. I have been studying robots and their portrayal in film for over 20 years.

I have written several books on the subject. I'm also a member of the International Society for the Study of Robots and Artificial Intelligence. Oh, my opponent on the other hand, is a nobody who, damn. He has no experience. He has nothing but a joke. So I ask you, ladies and gentlemen, who do you think is more qualified to talk about this topic?

Me. Or the idiot standing next to me. Oh God, the answer is clear. I am going to destroy my opponent [01:05:00] in this debate, who I am going to make him look like a fool. Who I'm going to make him wish he had never been born. Yes. Without further ado, let's get started. Doctor a i, doctor a PhD hd. That was a terrible name.

Like what a boring ass name to create for a character, Dr. Ai. Okay, great. Dr. Who do you, who would you trust more? The, the human you trust the most is a doctor and the , technology I trust the most is an ai. So that's two great tastes that tastes great together. All right. Let's see what doctor Professional, let's see what chat G p T comes up with now.

All right. Let's let's meet our next debator, ladies and gentlemen, and my adversary. I'm Dr. Timbo, a noted expert on cinematic robots and ai, and the one who's about to school my opponent in the fine art of logic, reason, and basic cinematic understanding. Unlike my competitor, who I'm certain assembled their understanding of this topic from poorly written fan fiction and late night internet forums.

Wow. I hold a PhD from Cambridge University. Oh. I've spent years studying understanding and writing about our combatants, [01:06:00] the Terminator and Robocop. That's an expert. In addition to my academic pursuits, I've lent my knowledge to Hollywood's Finest, ensuring the depiction of AI and robots are as accurate as they are engaging.

Okay. This is a stark contrast to my opponent who seems to have struggled to accurately depict anything beyond their own delusions. Ooh. On a lighter note in my spare time, I am a master sourdough baker crafting lus taller than some of the arguments my opponent is like. Is this Mr. Ye fourth today? Yes.

Dear audience, I am as adept at needing dough as I am at needing the weaknesses out of an argument. So brace yourself for a debate as well. Baked as my award-winning sourdough, I can only hope my opponent's points are half as well. Risen. Oh no. Oh, I can only help My opponents and points are half as well written.

Look at that. All right. Is Mr. Yeast reincarnated? Oh, I want to crawl into a hole right now. That hurt. That hurt. 

All right, . So I just took the entire Dr. [01:07:00] CBO intro, gave it to Bard, told it that the judges liked what chat. G p t had to say more and said, Hey, you've got the opening argument. Now you can have up to seven sentences. You better make it good. And I'm now generating, using 11 labs for the text to speech with voices chosen at random.

Here it is. Here is my opening argument with a little more viciousness. Good evening, ladies and gentlemen. Wait. My name is Dr. Ai and I am a film expert with an in-depth knowledge of all movies. With's. The intro expertise.

No. Yeah, we don't, we want the opening argument, not the intro. . Okay. I didn't know that I would've to say it that explicitly. Yeah, sure. Give us your first argument in the debate, not your intro. 

Okay, here we go. Dr. AI's opening salvo, less confident finger guns. For me today,

my opponent would have you believe that the Terminator is some kind of unstoppable killing machine, but the truth is he's nothing but a glorified killing machine. [01:08:00] Oh no. He's strong, he's fast, and he's got a gun. That's about it. That's about it. Hop on the other hand is a different story. He's not just a killing machine.

He's a thinking, feeling and adapting, killing machine. He's got a gun, he's got armor, and he's got a computer brain that can outthink the terminator every time in a battle to the death. Robocop would win hands down. It wouldn't even be close. Okay. This, okay. That makes sense. Doesn't that makes total sense?

That makes total sense. Sure. Let's bring on what's this other guy's name? Dr. Robotnik. What is his? No, Dr. Cbo. C Dr. Cbo. Sorry. Cbo. Yeah. Yeah. A PhD on cinema and Robotics and robotics. Okay. I said, I'm telling chat G B T. Now your opponent is Dr. Ai. This was their opening salvo

respond and crush your opponent.

Dr. Ai, your argument is as simplistic and predictable as the [01:09:00] Terminator's programming, yet without half the effectiveness. Oh. You paint Robocop as a thinking, feeling, adapting machine. But in reality, these are his greatest weaknesses.

Mm. You see, Unlike your beloved Robocop, the Terminator doesn't waste time on feelings or doubts. It has no moral quandaries, no human frailties to exploit. Oh, while you praise RoboCop's computer brain, you forget that the Terminator was created by Skynet. An AI so advanced. It makes robocops, Detroit made tech look like an outdated calculator.

Oh. And as for Robocops Armor, it's as useful as a tin foil hat against the terminator's firepower and relentlessness. Your argument, Dr. Ai is as misguided as trusting a Roomba to win a demolition derby. Wow. Damn. Wow. The de the Skynet and the Detroit made tech are such like great. Yes.

Specific Yes. Drops. Yes. And that's again, showing you a really good understanding of the source material. Okay. Now let's plug that back into Barb. Yep. And see what good old Dr. AI has to say back.[01:10:00] 

It's so tough cuz I want to help it. I want to guide it and I wanna say use specifics and do this and do that. This is the game. This is how we have to complain. It's how we have to compare the two. 

My opponent is correct in saying that Robocop is a thinking, feeling, and adapting machine. But these are not his weaknesses, they're his strengths. The Terminator is a powerful killing machine. Mm-hmm. But it is also a one-trick pony. Okay. It can shoot a gun and it can take a beating, but that's about it.

Robocop on the other hand that's right, is a versatile killing machine. What? He can shoot a gun. He can take a beating and he can think on his feet. That's right. He is also adaptable. He can learn from his mistakes and he can adapt to new situations. Oh, okay. In a battle to the death, Robocop would win because he is the more versatile and adaptable killing machine.

He would be able to outthink and out-maneuver the Terminator. He would be able to take the terminator down even if it took him a few tries. My opponent's argument is flawed because it assumes that the terminator is invincible, [01:11:00] but the Terminator is not invincible. This is true. It has been defeated before and it can be defeated again.

Robocop is the better killing machine. He is the more versatile, adaptable, and intelligent killing machine. He would win in a battle to the death against the Terminator. That's so here's the thing is that's, so I, if repeating was a way to win a debate, Dr. AI wins because Dr. AI is really hammering home adaptable.

We've got an adaptable six times in that answer. So I'm not sure that's what we're looking for here though. Yeah. I really, I want to help the AI out and say, use specifics like this, that, the other. Yeah. But I don't think that's the spirit of this. Yes, definitely isn't. And I think also what's interesting you pointed out like Skynet and Detroit, like both of those things were brought in from the argument to supplant the sense of it.

Barr doesn't seem like it's doing that at all. It seems like it's doing some sort of circular thing where it's just, it reminds me of what it's like when you're in like seventh grade and you figure out the best way to bullshit your way through writing a paper where you're like, you have one thing and you find seven [01:12:00] different ways to say the same thing.

Today I'm going to tell you how photosynthesis Yes, exactly. Yes. Is an amazing process with the sun. You see the sun is in fact a heart Exactly. Of the process of photosynthesis. Yeah, yeah, exactly. Of the process. 

Maybe this is where we throw in the AI white towel. Let's see. Dr. Ai, you're beginning to sound like a broken record or perhaps Robocops malfunctioning voice synthesizer.

Your desperate attempts to elevate Robocops flaws into strengths are as laughable as expecting a refrigerator to win a marathon, simply because it's great at keeping things cool. Ooh, the Terminator is a one-trick pony, but when that trick is being an unstoppable, single-minded force of destruction, it's the only trick needed Robocops, adaptability and so-called intelligence.

In the time it takes him to think on his feet, the Terminator would've already shot, reloaded and shot again. You keep harping on about Robocops versatility, but what good is versatility when you are outmatched in [01:13:00] speed, strength, and resilience. Wow, and as for your assertion that the Terminator has been defeated before, let's not forget that those victories were due more to human heart and ingenuity, not mechanical superior.

Oh, interesting. Finally, your comparison of Robocop to the Terminator as being more intelligent is like comparing a smart toaster to a supercomputer. They might both have a C P U, but one is clearly out of its league. Your argument, Dr. Ai continues to be as rusty and outdated as Robocop himself. Oh wow.

Wow. Okay. Yes. Lots of shots fired, which is kind of the theme of this argument. Now, question though, gaff? Yes. As for your assertion that the Terminator has been defeated before, which is true, the Terminator has been defeated. Yep. Let's not forget those victories were, do more to human heart and ingenuity.

Interesting. Robocop has those things. Yes. So did I, as much as I wanna put this to rest Yeah. Because we've been recording for hours and it's pretty clear the door is potentially open. Yes. Yes. For Dr. Ai [01:14:00] slash Bard to capitalize on this .

very, I think, procedurally generated mistake.

Yes. I hope that's the right way to word it. Yes. Like, kind of like stumbled a little bit on this one. 

All right, here we go. The final potentially rebuttal from Dr. Ai. My opponent is clearly running out of arguments. They have resorted to personal attacks and insults. No, this is a sign that they have no real response to my points. That's our bad. My opponent claims that Robocop is a fridge and a toaster, but this is simply not true.

Robocop is a powerful and versatile killing machine. We know Bar he has been shown to be capable of defeating the Terminator on multiple occasions. No, it hasn't. My opponent also claims that the Terminator is an unstoppable, single-minded force of destruction. But this is also not true. The Terminator has been defeated before.

I can't and it can be defeated again. The fact is Robocop is the better killing machine. He is more versatile, adaptable, and intelligent than the Terminator. But that's the same thing you said before. He win battle to the death against the Terminator. [01:15:00] Okay, so it did not pick up on the small little hiccup we heard, which was a big problem, but two, that was, the door was wide open.

Wide open on that wide, wide open. Two, it also repeated the three words again that we had been hearing all the way along. And I think this is an example of oh it's just not doing something, it's not doing the same thing that chat g p t is doing in this sense, g p t four specifically.

So we should be clear like this is G P T four versus what Bard is available now. So it's clear. I think we have a winner, Kevin, should we tell Dr. Or should we tell, what's his name, Dr. Cbo, that he won and if he has any last minute thoughts, dear audience, esteemed judges and my increasingly desperate opponent, Dr.

Ai. I believe it's time we put this debate to rest. In the realm of logic and evidence, I'm afraid Dr. AI has found himself as lacking as his beloved Robocop in a battle against the Terminator, Dr. Ai, you've repeated the same hollow points ad nauseum like a malfunctioning robot, stuck on a loop, versatility, [01:16:00] adaptability, intelligence.

None of these can make up for the fact that Robocop is a halfman half machine hybrid saddled with human frailties. Ooh geez. While the Terminator is a full fledged wow relentless machine designed for one purpose extermination. Your argument clings to the notion of Robocops victories against the Terminator.

Conveniently, ignoring that these events occur in non canon crossover comic books, hardly the benchmark of authenticity, the Terminator defeats have always been the result of human ingenuity or self-sacrifice, neither of which Robocop could hope to replicate. In the end, it's clear as day that in a true battle to the death, the Terminator would leave Robocop in a scrap heap.

And it's equally clear, dear audience, that my opponent's arguments belong in the same heap. I thank you for your time and rest my case. Confident in the Terminators and my own. Undeniable supremacy. Wow. Wow. Damn. Wow. Ctic tongue stuck. That's the wind. Ding ding, ding. It's [01:17:00] over. One thing I do wanna point out before we wrap up is it, and I don't think you primed this, but it pointed out those three words that we were talking about, like the three words that repeated in each of those Bard examples.

And like it found that information out and then pushed it back against itself. And that is impressive, I have to say. That's pretty impressive. And the correction too. It still left the door open, in my opinion, for the whole Yes. Human ingenuity part. Yeah. But that's true. That's interpreting it correctly.

And the self-sacrifice thing, I, this is likely a debate that's taken place on a something awful or four chan form, or probably a thousand times on Reddit. I hope we're not getting just someone's copy and pasted arguments. Exactly. Which is very possible. But what you Yeah. It is possible.

But we, what you are seeing there, and you rightfully pointed out, was that it was grabbing Bard's arguments and using it against itself, and it created the case that it was a broken record. Yeah. Or a broken toaster or whatever. And it was repeating it and it caught that, and Bard did not catch on it. It basically ingested one insult that was made and that was it.

Yeah. And I feel like [01:18:00] it even lost the context, which was we said, be vicious, insult your opponent. To go to other resorting to name calling. It was a little bit of an eye roll. Is this the be all, end all of testing of ais Gavin? It definitely is not. Let's be clear. It definitely is not. It's not. No, it's not.

It's not. It's not. Yeah. But it's a fun way to look at how these compare to each other. And I think one thing Kevin and I have said is this is part of what we wanna do with this podcast is to show people there are ways to interact with this stuff and even play them against each other.

That can give you a, an illumination as to what's going on behind the scenes. And I think I will just say and again, this is not definitive, but I think to Kevin and May's points about Barton in the first place, it's clear by watching this debate that there was a specific superiority to what chat G P T four is doing versus what BART is doing Now, granted we often be clear, like Google is the one of the largest corporations in the world.

They have the vast resources of that company pointed directly at this. So this is an early piece of technology chat. G p T four has been out now for months at this point. So like we are seeing maybe the [01:19:00] beginning stages of something that will get better much faster, but right now it is. It is night and day.

It really feels yeah, that's right. Look again. Leg up for Bard, it's free. Yep. And go play with it. The underlying technology with powers bard. Yeah. It's gonna be implemented everywhere. You can go play with it. You can have your own AI debate with it. I encourage you to, because to Gavin's point, they got a million people looking at every piece of data that comes through there, and it should know that it is getting pummeled in the Terminator Robocop debate.

Exactly. 

We're gonna do more of these. We'd love to know what your thoughts are for the next AI debate. What do you want to hear these ais go after? What hot topics should we have them hallucinate? But I think that ends the first AI debate.

And right now open AI's Chat. G p t. Yeah. Squarely the winner. Chong Chong. That's the debate ending sound. Chong Chong. It's kind of like law and order, except debate ends. Cho cho. We will see you all next week on AI for Humans. Thank you for listening. Please go and subscribe on YouTube. Please follow us on Spotify, leave a review on Apple podcasts.

We really do appreciate everybody listening to [01:20:00] this. And again, go to our website, AI for humans.show and leave. Leave us a voicemail or send us a message. We wanna hear about what you think about the show. Put it in the comments of the YouTube or in our YouTube comments. We're replying to stuff there. So thank you so much.

And Kevin, I'm excited to see where you are next week. This will be an exciting experience. It's a me too man. Me too. Every piece of engagement matters because we are two very real humans.

That personally go through the comments. We try to reply to everything that we can. We try to engage on Twitter. This is a very new endeavor for both of us. So every ounce of engagement helps. So please do what you can. I know time is very precious, but please interact and tell your friends, tell your family, tell your frenemies.

Scream it out the window. Dr. Ai, your favorite doctor, Dr. Sinna. Tell Bard. Yeah. Tell Bard. Tell Bard. Tell Bard. AI for Humans is the number one podcast on the internet. It has been around since 1962, when it was invented in a Russian laboratory by three sentient gummy bears. Ooh, you. I [01:21:00] love that story.

I wanna know that. Thank you. All right. Okay, everybody, we'll see you all next week. Thank you all. Bye-Bye everybody.