Gavin Purcell & Kevin Pereira, The Future of AI (#55)

0:00:03 First of all, why did you guys pivot into “distraded AI”?
0:00:08 I would say, to me, the thing that was really interesting about it was it felt transformative in a way that…
0:00:12 We’re entering an era, like it looked like anyone is saying anything at any point in time.
0:00:15 I’m a little worried about the AI stuff.
0:00:17 All the people are talking about agentive AI right now, right?
0:00:23 And this is a good example of what you have there, conceivably, is your agent that you’re making.
0:00:25 That’s gonna need to go out. One of the coolest things about AI is it’s…
0:00:30 If so, we don’t have this theory that social media was the kind of democratization of distribution.
0:00:37 AI is kind of democratized creation, meaning that suddenly the floor is raised for everybody that wants to create things.
0:00:42 There could be 100%, and this is like the Wally problem, right?
0:00:45 The risk that I always find is that, like, how do you bring the human side into it?
0:00:50 There will be something lost, but then will those people be able to have more time to do things with their family?
0:00:50 Maybe.
0:00:57 So if you follow me on the Instagram, I’m sure you’ve seen I’m doing 90 days, no drinking.
0:01:03 Today is day number 25, and I’ve upped my cardio a ton, which feels great.
0:01:07 I’ve been rucking a lot, which is where you put this weighted backpack on while you’re hiking.
0:01:13 At the end of my workouts, though, I’m sweating like crazy, which is good, but you need to replenish your electrolytes.
0:01:18 And sadly, most of those replacement powders out there are just packed with sugar and they go straight to gut fat,
0:01:25 which is the reason I use Element. There’s no sugar and it has a science-backed ratio of 1,000 mg sodium,
0:01:29 200 mg potassium, and 60 mg magnesium.
0:01:35 Not only no sugar, no coloring, no artificial ingredients, no gluten, no fillers, just no BS.
0:01:41 Element is used by everyone from podcast hosts like me, NBA players, NFL, NHL players, Olympic athletes,
0:01:45 and everyday moms and dads and exercise enthusiasts.
0:01:49 Right now, Element is offering a free sample pack with any purchase.
0:01:53 So that’s eight single serving packets, free with any element order.
0:01:56 That’s a great way to try all eight flavors because they have a ton of different flavors.
0:02:02 You try all eight, get yours at kevinrose.com/elementlmnt.
0:02:05 So that’s kevinrose.com/lmnt.
0:02:09 This deal is only available through my links, so you must go to that website.
0:02:12 And lastly, and the best part is totally risk-free.
0:02:15 So if you don’t like it, share it with a salty friend and they will give your money back.
0:02:18 No questions asked, you have nothing to lose.
0:02:21 Huge thanks to Element for sponsoring today’s show.
0:02:23 Today’s sponsor is NordVPN.
0:02:29 Stay secure and anonymous online with NordVPN, which has been my trusted VPN for years now.
0:02:32 I like it because you can access to a lot of content from anywhere.
0:02:37 So I’m in Japan or Mexico at least a couple of times a year, and I want to watch my favorite shows.
0:02:43 And this comes in really handy to bypass all of that geofencing and it needs to be fast.
0:02:49 This is another essential thing for a VPN and Nord is lightning fast because they have thousands of servers.
0:02:52 So if you want to stream and game without lag, it’s great.
0:02:54 And then equally as important is privacy.
0:02:55 They have a great policy.
0:03:01 No logs, nothing to store, nothing to see, nothing to share, which is pretty much as good as it gets.
0:03:03 And their apps are just really lightweight.
0:03:07 So it’s not bogging down your system the entire time and they have it for desktop and mobile.
0:03:13 Don’t miss this incredible deal, which is two years at a huge discount, plus four free months.
0:03:21 If you use my link, NordVPN.com/KevinRose, that’s NordVPN.com/KevinRose and it’s risk free.
0:03:25 There’s a 30 day money back guarantee, which makes this a no brainer.
0:03:27 Gavin, Kevin, so glad to have you guys here.
0:03:28 Oh, we’re psyched.
0:03:28 We’re psyched.
0:03:29 Dude, this is great.
0:03:30 Yeah, it’s fun.
0:03:31 So you had me on your show.
0:03:32 Yep.
0:03:34 That was a really fun show.
0:03:34 Well, good.
0:03:35 I had a ton of fun.
0:03:38 Because for me, we’re all AI geeks.
0:03:40 Like we’re playing with all the tools, but you guys are like living and breathing it.
0:03:43 I’m like a few steps removed.
0:03:45 I have to hear like three people be like, hey, this is cool.
0:03:46 And I’m like, all right, let me go check it out.
0:03:48 Are you still elping yourself?
0:03:49 Are you like in jib jab?
0:03:49 Yes.
0:03:50 Era of technology.
0:03:51 I’m in jib jab.
0:03:52 I haven’t broken out.
0:03:56 But I want to have you guys on because you’re so in the mix on what’s going on here.
0:03:58 You’re giving talks about this stuff.
0:04:01 Can you tell us and the audience is watching?
0:04:02 What are you seeing?
0:04:04 Well, first of all, why did you guys pivot into this arena?
0:04:08 Like you both had very successful careers doing other things in video.
0:04:08 Yep.
0:04:08 Yeah.
0:04:12 But AI was enough to be like, okay, we have to join forces to make this happen.
0:04:16 To me, the thing that was really interesting about it was it felt transformative in a way
0:04:19 that I hadn’t felt since way back in the web two days, right?
0:04:21 Web two was this really interesting moment.
0:04:23 And I was never like really in it.
0:04:27 You were got into it, but tangentially watching it was it felt like, oh, suddenly
0:04:30 everybody’s out there and able to interact.
0:04:32 You can make stuff yourself when YouTube came out.
0:04:33 Twitter came out.
0:04:35 Instagram, dig all these things started.
0:04:40 It felt like, oh, this is a moment of something where we are changing the way things are made.
0:04:42 And that’s what AI felt like to me.
0:04:42 Yes.
0:04:47 And I didn’t want to, I mean, Kevin, I do this like a fun thing and we were really interested
0:04:51 in it, but like we started doing a GPT three is what we first saw.
0:04:54 If you remember GPT three, it was like 2021.
0:04:56 We were both blown away by what was possible.
0:04:59 Then that was if you’re not familiar with the backstory that’s open AI started before
0:05:03 chat GPT, they had been releasing a bunch of GPT models.
0:05:05 That model was the first one when you interacted with it.
0:05:09 It felt almost like you could do this character stuff, which we really, I would say even GPT
0:05:13 two was like a turn on in a way that I think this room and maybe some of the audience will
0:05:14 understand.
0:05:17 Do you remember the first time you heard the modem handshake noise?
0:05:19 Oh, actually connected to the machine, right?
0:05:23 And you just were like, Oh my God, I’m actually talking to another computer.
0:05:25 Yeah, we’re really dating ourselves here.
0:05:27 No, but that’s the moment, right?
0:05:28 And I remember it vividly.
0:05:29 The iPhone was like that too.
0:05:34 For me, there’s only like four big moments in like my last call it like 20 plus years
0:05:39 in technology where I’m like, a big shift, holy shit, we got to pay attention.
0:05:40 It was dial up.
0:05:43 I would say from there it was like web one was really interesting because PayPal came
0:05:47 out of that eBay came out of that Amazon came out of that some of the bigs that are
0:05:51 still around obviously then it was web two was interesting because that was like interactive
0:05:55 web pages in a way that we hadn’t seen a second is JavaScript allows to vote on things
0:05:59 and then make like real time comments and have a living, breathing web and UGC, right?
0:06:03 Which I think is a huge deal and you just still underestimate in a major way of people
0:06:08 generating their own content, blogging, like self expression, like following the following
0:06:13 feature like that, that Jack and when they added that to Twitter, because you had to
0:06:16 remember before that was all bidirectional relationships.
0:06:20 So if you want to follow someone on MySpace, it was like, I friend request, you accept
0:06:22 request now we can share.
0:06:27 And so when it turned to following that was like, that really screwed up society that
0:06:28 was the thing.
0:06:29 That was the problem.
0:06:34 Let’s go one to every one and that’s where all of a sudden the town hall was now a global
0:06:36 town hall and it was like, it was crazy.
0:06:41 But anyway, that moment, the smartphone and it wasn’t the first smartphone because you
0:06:46 remember like the pumps and the Windows devices, the little stylus and shit, this is horrible.
0:06:48 But I was saying it’s those tingly moments, right?
0:06:54 And I remember even with GPT to having that right broken magnet poetry, like someone’s
0:06:57 just replacing words on the refrigerator and if there’s a sentence there, I remember seeing
0:07:00 that and going, Oh, wow, I’ve been here before.
0:07:04 It was that weird sense memory of this gets better.
0:07:07 This is the this is something we say all the time on the podcast right now is the worst
0:07:10 this technology will ever be right.
0:07:11 It only gets better from here.
0:07:14 I know which is super scary because it’s hard to be damn good.
0:07:15 Yeah.
0:07:18 But there might be one thing we also talk about is there might be a lot of legal things
0:07:21 that get in the way of it or people that get in the way of it and I think that’s the other
0:07:22 side of this.
0:07:25 And on our podcast, we try to deal with the human version of what these things are.
0:07:29 And they’re the technology is going to probably go exponential for a long time, right?
0:07:32 So far, you can train more data means better models.
0:07:35 But what are the consequences of those things?
0:07:36 What does it mean if there’s a better model?
0:07:40 What does it mean if that model can produce a deepfake of somebody in in five seconds,
0:07:41 right?
0:07:42 We just talked about it.
0:07:43 It’s going to happen though.
0:07:45 I don’t think the legal thing is going to be a real thing.
0:07:46 I’ll tell you why.
0:07:50 I’ve seen this unfold a couple of other times when Uber first came out, everyone tried to
0:07:52 regulate the hell out of them.
0:07:54 But consumers said, actually, I really like Uber.
0:07:56 I want to use Uber.
0:08:01 And if they didn’t do it, other startups would fill in other countries and other jurisdictions
0:08:04 and it was like, it was going to, they had to change the loss.
0:08:05 And they did.
0:08:06 And Uber was fine.
0:08:08 YouTube was the same thing.
0:08:10 YouTube almost went out of business.
0:08:13 They were getting sued left and right by the major industry players because they were
0:08:17 put all kinds of copyrighted material in there and they didn’t have smart systems to go in
0:08:19 and find all that stuff in real time.
0:08:20 Google kind of saved them.
0:08:24 I don’t know if you remember when the acquisition happened, but they were under a lot of pressure
0:08:26 and the Google said, you know what?
0:08:27 We have deep enough pockets here.
0:08:31 We’re going to go in and actually settle these, figure out the terms.
0:08:33 Same thing happened with Napster, with music sharing.
0:08:36 When consumer behavior changes, these things will get ironed out.
0:08:39 I think they’ll be, they’ll definitely be lost if it’s no doubt, but there’ll be settlements
0:08:40 and we’ll move on.
0:08:45 I think that legal red tape is only going to slow down the bigs so much, right?
0:08:51 Because even if it does 1%, let’s say it doesn’t matter because if you use the Uber analogy,
0:08:54 there is an open source solution right around the corner where it’s like, Hey, do you want
0:08:55 to drive?
0:08:56 Well, it’s not for Uber.
0:08:59 It’s for the open source taxi delivery service.
0:09:02 We’re seeing that with AI where it’s like the foundational models might get under attack.
0:09:07 They might have to guard rail it so you can’t generate SpongeBob smoking blunts on the couch.
0:09:10 But there is an open source something freely available.
0:09:12 You can grab it and they’re going to keep iterating on that.
0:09:13 That’s exactly right.
0:09:18 Because anytime you crack down on something, there will be an underground version of that
0:09:19 thing that works.
0:09:20 It’s just going to happen.
0:09:21 I’ve seen this.
0:09:22 I mean, you guys have seen this.
0:09:25 I’ve been invited into these discords where you can go, you don’t know what I’m talking
0:09:26 about.
0:09:27 I don’t know.
0:09:28 Walk away.
0:09:29 Walk away.
0:09:30 What’s going on in your disc?
0:09:31 What?
0:09:32 You can do some crazy shit.
0:09:33 Yeah, sure.
0:09:34 You can break all the rules.
0:09:36 Those are just whack-a-mole.
0:09:37 That’s not going away.
0:09:41 And then you have other countries that are like, yeah, US, go ahead.
0:09:42 Please.
0:09:43 Pregulate this.
0:09:44 Please.
0:09:46 So we can just jump in here in this rule because we won’t have the same regulation.
0:09:51 That’s why the pause letter, which we talked about a lot, which Elon and a few tech luminary
0:09:53 sign, that’s why it was just so laughable.
0:09:54 You guys need to slow down.
0:10:00 Pull this big imaginary lever that doesn’t really exist to stop development, stop training,
0:10:01 stop advancement.
0:10:03 Meanwhile, everybody else is going to lap you.
0:10:05 And by the way, I’m trying to build it in my closet.
0:10:06 Right.
0:10:08 And we’re saying, hey, why don’t we pause?
0:10:09 You pause.
0:10:10 I’m going to actually open here.
0:10:14 I’m going to keep building really hard because they’re just trying to, it’s not working.
0:10:17 I think the question with the YouTube model is the most interesting one, right?
0:10:20 Because YouTube went from being what was a lot of pirated content to now whatever it
0:10:25 is, like $150 billion business and all of the big people came onto it.
0:10:28 That was the thing where like CBS, NBC, all these people that weren’t going to be part
0:10:32 of YouTube now are because it’s a great distribution channel.
0:10:35 The big question I have with this is who will be the person that wins to be the place where
0:10:38 they’re going to, all these companies are going to lean in on.
0:10:42 Is there a person that wins because with the YouTube, you can make all those deals.
0:10:46 What does it look like if it’s 15 different things and then all that stuff exists?
0:10:47 Yeah.
0:10:48 It’s a good question.
0:10:52 I think it’s going to be some type of bi-directional relationship that is beneficial to both parties.
0:10:53 So I’ll give you an example.
0:10:54 This is a true story.
0:10:57 My wife woke up this morning and she goes, hey, my whoop data is like saying that I’m
0:10:58 off today.
0:11:01 Like I’m a little bit under and I’m wondering why I didn’t have any drinks last night and
0:11:02 blah, blah.
0:11:04 I’m like, why don’t you just ask the AI because it’s built into whoop right now.
0:11:08 So whoop has a prompt and you can go in there and I literally took her phone and typed in
0:11:10 like, why am I in the orange today or whatever.
0:11:13 And it was like, oh, your heart rate was a little bit higher than normal blah, blah,
0:11:14 blah.
0:11:18 And it was like, it was, that’s powered by, you know, whatever, chat, GPT or whatever’s
0:11:21 behind the scenes there using your real time data as part of the model.
0:11:25 And so there’s a beneficial effect that I have to imagine when we’re talking about the
0:11:31 New York Times data, when you’re talking about Washington Post or any of these walled gardens,
0:11:35 you’ll have private models that you will to say, hey, I’m a member of chat GPT, but I’m
0:11:37 going to do the $2 a month add-on.
0:11:40 Kind of like when I buy Apple TV and it’s like, would you like to add stars for an extra
0:11:41 $2 or whatever?
0:11:42 You’re like, all right.
0:11:43 You tell me we’re building actual bundles again.
0:11:44 Is that what we’re doing?
0:11:45 But I feel like it’s going to be that way.
0:11:48 We’ll extend it in interesting ways and say, actually it is interesting for me to have the
0:11:54 back catalog of the Financial Times because I value that quality of reporting and I want
0:11:57 that as part of my stack that I’m getting back.
0:11:58 Or it’s an add-on for a subscription, right?
0:12:03 If I’m a New York Times subscriber, which I am, I would be able to get that as part of
0:12:04 the thing.
0:12:05 That is valuable for sure.
0:12:06 100%.
0:12:11 The question I still have is like, if I want to have Spongebob interact with Luke Skywalker
0:12:14 or whatever, those two people together, they have the power to do that now.
0:12:19 That’s all the legal background stuff that has to get solved to make it fun and interactive.
0:12:20 I don’t know if it’s fair use though, man.
0:12:22 It’s just going to be considered fair.
0:12:23 You think so?
0:12:25 I mean, if you’ve created a full-featured film with Spongebob in it, I think you have
0:12:28 some grounds that actually let’s not do that.
0:12:32 But I think if you’re a verified license holder, which is where we’re heading, it’s like, well,
0:12:36 do you have the Disney+ package and you have the Viacom, Nickelodeon, whatever, okay.
0:12:39 You can get Spongebob being forced choked by Darth Vader.
0:12:40 You can have it.
0:12:42 If you want to release it commercially, that’s a whole other thing.
0:12:46 And if we decide to allow you, you’ll maybe get a small teeny-tiny royalty, but who are
0:12:47 we kidding?
0:12:48 You’re probably going to get nothing.
0:12:49 They’ll get everything.
0:12:51 Thank you for making the next worldwide sensation meet.
0:12:52 Yeah.
0:12:53 You see this with YouTube, right?
0:12:54 Yeah.
0:12:55 At least something with a commercial song.
0:12:57 It detects that song related to it.
0:12:58 And it gets paid back.
0:12:59 It gets paid back.
0:13:00 It knows that you used it.
0:13:02 Like, I have the feeling those rails are being built right now.
0:13:03 Yeah.
0:13:06 And by the way, that’s great because we talk about all the time on the show.
0:13:09 One of the coolest things about AI is it’s so we don’t have this theory that like social
0:13:12 media was like kind of democratization of distribution.
0:13:17 AI is kind of democratized creation, meaning that like suddenly the floor is raised for
0:13:21 everybody that wants to create things, whether that’s audio, video, movies.
0:13:25 There could be, the whole thing about Twitter or anything like that was always, here’s a
0:13:28 bunch of kids in Africa that are making an incredible film and now it can be distributed
0:13:31 everywhere and they have the tools to do it.
0:13:35 Well now there could be a thousand of those kids and they might be two peak kids doing
0:13:38 it or one kid because all of the tools are going to be big enough for them to be able
0:13:39 to do it.
0:13:43 And the downside of that is the deluge of stuff we’re all going to have to deal with,
0:13:44 right?
0:13:46 Because just like with YouTube, when YouTube first came out, the whole thing was like,
0:13:48 there’s so much crap here.
0:13:52 But there were gems and now the gems are more because there’s more people doing more
0:13:53 stuff with AI.
0:13:54 We’re going to have the exact same problem.
0:13:58 There’s going to be so much stuff, but you hope that stuff can bubble up.
0:13:59 That’s going to be really good.
0:14:04 Do you think that AI is going to be an information, like, is it a lean forward or lean back experience?
0:14:08 Like on the lean forward side, you’re creating content, lean back, you’re consuming it.
0:14:11 And for me, I mostly use it as a lean back.
0:14:15 Like I used it to create some funny images and make myself look a certain way.
0:14:16 In those questionable discords.
0:14:17 Yeah.
0:14:21 I got myself ripped with the six pack and I won’t say what else, but you can do fun
0:14:22 things with it on that front.
0:14:25 But there’s a novelty to that that kind of wears off.
0:14:29 And then on the lean back, like I’ll give you an example before you guys got here.
0:14:33 I was like, okay, I wasn’t getting video out into these squares here and I don’t have time
0:14:37 to sit on hold for an hour with black magic and be like, how do I figure this out?
0:14:40 And so I was like, chat to people, I was like, why am I not getting programmed out of the
0:14:41 bubble ball?
0:14:42 And it’s like, have you checked the setting here?
0:14:43 And this and that.
0:14:46 It’s bringing, it’s solving customer support.
0:14:51 I mean, in a way that like companies, oh my God, I think customer support is a huge use
0:14:53 case of it right now because we talk about this too.
0:14:56 All the time we, we, we have a problem.
0:14:58 You can solve 90% of your problems with chat GPT.
0:14:59 It’s really shocking actually.
0:15:03 But we’re already seeing that though with call center, displacement, yes.
0:15:06 Support teams being slashed everywhere because they’re just putting all the training docs
0:15:10 and FAQs into a GPT and letting it answer the questions.
0:15:14 And then you also get like a Chevy dealer in California that was recommending Teslas,
0:15:18 by the way, and selling them for $1 because they want to properly guard, right?
0:15:19 Wait, what happens?
0:15:20 Yeah.
0:15:24 There was a GPT powered chatbot on a Chevy dealer’s website and people were negotiating
0:15:28 with the bot and saying, so wait, you’ll sell me this Tahoe for $1.
0:15:32 Basically, it was like, yep, anything to get you in to try our, whatever.
0:15:33 Guard rails.
0:15:34 But yeah, but to answer that.
0:15:35 Is that legally binding?
0:15:36 That would be amazing.
0:15:37 They’re trying to get it.
0:15:38 No, they really are pushing right now.
0:15:39 They’re trying to get it for a dollar.
0:15:40 That’s amazing.
0:15:42 And it also recommended that you drive a Tesla, which is very funny.
0:15:45 But to the lean back versus lean forward, I think it kind of depends on the end user.
0:15:51 Like we’re already seeing, if you go to, we’ve talked about Suno and UDO, these AI song generation
0:15:55 apps, there’s plenty of people that are now just using them as their daily soundtracks
0:16:00 and going through song after song, absorbing the wall of data that everybody else is generating.
0:16:04 But there are people that are in there every day, putting quarters into this slot machine
0:16:05 and rendering songs.
0:16:06 Yeah.
0:16:09 I think it’s also about the effort of lean forward versus lean back world.
0:16:12 Like the effort is going to be way less now, right?
0:16:15 And again, that’s not about the idea that like you’re going to make great stuff if you
0:16:19 lean forward with less effort, but it’s way easier than it was before.
0:16:20 Yeah.
0:16:21 So we talked about it was a good example.
0:16:24 You do or Suno, like if you want to make a song this weekend, I made a song just as
0:16:29 for our podcast and I made a two and a half minute song outlaw country song and it took
0:16:33 me about an hour and that was like more effort than most people put in.
0:16:38 But that would have taken, I don’t know, two weeks before to get the same result out.
0:16:41 So I would have had to have learned the banjo.
0:16:42 Yeah.
0:16:43 Without that part.
0:16:44 Yes.
0:16:45 But let’s not discount that.
0:16:47 I could have taken like GarageBand and plugged a bunch of instruments in and found a way
0:16:48 to do all that stuff.
0:16:51 But yeah, so I wouldn’t have to learn the banjo, but I would have had to learn the system.
0:16:54 Can you explain to people what these tools are for maybe they haven’t heard them before?
0:16:58 So one of the coolest new like entry points for AI is AI music.
0:17:01 And so Suno.ai is one tool.
0:17:02 UDO is another one.
0:17:03 It just came out.
0:17:07 Suno was created by some guys for the MIT people that we’ve interviewed in our podcast.
0:17:10 UDO are former people, you know, people.
0:17:14 So they’re both systems to allow you to create songs, basically is off text problems.
0:17:16 So there’s two different ways you can get into it.
0:17:19 You can either just literally put a prompt in, in fact, we could probably do one live
0:17:21 and say a song about three guys on a hot podcast.
0:17:22 Three middle aged guys.
0:17:23 You almost said hot tub.
0:17:24 Yeah.
0:17:25 Hot tub.
0:17:26 Yeah.
0:17:27 Or a hot tub.
0:17:28 Or a hot tub.
0:17:29 Your imagination.
0:17:30 Three guys.
0:17:33 Three shirtless dudes hanging out in the dude’s suit.
0:17:34 Yeah.
0:17:37 Or you can put lyrics in, right?
0:17:40 And so what I did is I just, I’m not an amazing writer, but I wrote just this dumb song with
0:17:43 some lyrics and you put in a verse and a chorus.
0:17:45 How seriously did you take it?
0:17:46 I mean, he spent an hour on it.
0:17:47 That’s a lot.
0:17:48 I mean that’s a lot.
0:17:49 I thought he was like discounting himself.
0:17:50 I was the best writer.
0:17:51 I tried to play a little better.
0:17:52 I tried to play a little better.
0:17:53 I tried to play a little better.
0:17:54 Did I put on my chaps?
0:17:56 Did I grab my rhinestone cowboy hat?
0:17:57 Yeah.
0:17:58 I had to get in the character.
0:17:59 Well, one of the cool things, you should check this out.
0:18:02 There’s a song right now on TikTok that is exploding.
0:18:03 That is all AI generated.
0:18:05 It’s a guy named Obscurist Vinyl.
0:18:07 It’s a TikTok handle called Obscurist Vinyl.
0:18:08 There’s a song that is called.
0:18:12 I’ll just say real quick, the profile is great because he’s making it seem like playing to
0:18:14 the Obscurist Vinyl.
0:18:18 These are long lost records that you might have found at a press store, but they’re songs
0:18:22 and bands that never existed and they’re using AI art for the vinyl cover art.
0:18:25 It’s beautifully done, but they’re all slapstick.
0:18:26 And they’re dirty.
0:18:27 Like the one that’s blowing up.
0:18:30 You can believe whatever you have to believe is called I glued my balls to my butthole
0:18:31 again.
0:18:35 And it’s like set up like a 1950s, like kind of like almost twist song, but it’s exploding
0:18:42 if exploding is maybe the wrong word, but again, this is what I tell people.
0:18:47 We both have told people is if you want an entry point to see the power of these tools,
0:18:49 this is an emotional thing.
0:18:54 Whereas chat, GPT is a logical thing, but images and music is emotional.
0:18:59 And when you can show somebody what that feeling is, I am a creative person in my soul and
0:19:02 I’ve always been one, but if somebody isn’t a creative person or they don’t see themselves
0:19:05 that way, you give them the ability to do this.
0:19:07 It can be a big deal.
0:19:11 I think it can open somebody’s brain to be like, Oh my God, I think I made something cool.
0:19:14 That to me is the promise of the generative AI tool.
0:19:15 Yeah, absolutely.
0:19:17 I mean, and it’s just fun.
0:19:20 Like it gets you like I did one for my girls and put their names in it and they were laughing
0:19:21 their ass.
0:19:22 Yeah, exactly.
0:19:26 And it’s kind of like entertainment value that combines creation and entertainment is
0:19:27 really cool.
0:19:31 And one of the wild things about the Suno model is that it’s like a diffusion model.
0:19:36 It’s the same way that image images are generated right now for stable diffusion mid journey.
0:19:40 So basically in broad strokes, when they were training this model on music, they were feeding
0:19:43 a bunch of songs into the machine, but not actually telling it.
0:19:44 This is music.
0:19:45 This is a chorus.
0:19:46 This is a verse.
0:19:47 This is the style.
0:19:48 This is a BPM.
0:19:50 This is the rhythmic measure.
0:19:51 They weren’t giving it anything.
0:19:55 They were just feeding it stuff and saying, now just generate noise and let’s see what
0:19:56 happened.
0:20:00 And the machine essentially pattern matched and learned, oh, this is the right structure
0:20:01 for a song.
0:20:02 This is rhythm.
0:20:03 Crazy.
0:20:07 And then they would go in and label and categorize and fine tune and say, okay, this is percussion.
0:20:08 This is this.
0:20:09 Isn’t that scary?
0:20:10 It’s too scary.
0:20:11 That’s scary.
0:20:12 It’s too scary.
0:20:13 But remember when Google had their language conversion and then they had, they were trying
0:20:17 to figure out the most efficient way to convert and translate languages.
0:20:23 And they train the AI and the AI figured out its own internal language to communicate and
0:20:24 translate the languages.
0:20:28 I remember one when Facebook was doing that, Facebook was doing that to have bots automatically
0:20:29 negotiate deals.
0:20:32 And it started speaking in a code that the engineers didn’t understand, but it would
0:20:33 just spit ones and zeros.
0:20:35 That’s when you turn the servers off right at once.
0:20:38 No, that’s when you put Google EIs on them and you give them guns.
0:20:39 You gotta go there.
0:20:40 Hey, they figured it out.
0:20:41 Send it to the front lines.
0:20:42 Let’s go.
0:20:43 Oh my God.
0:20:44 Yeah.
0:20:45 That’s crazy.
0:20:47 The software side of things is fascinating right now.
0:20:50 Did you see the Boston Dynamics robot that they showed off yesterday?
0:20:51 Is your familiar?
0:20:52 They made the big dog robot and everything.
0:20:53 Of course.
0:20:54 So you can kick them over and shit.
0:20:55 Right.
0:20:56 Yeah.
0:20:57 Right.
0:20:58 But the videos are them trying are great.
0:20:59 Yes.
0:21:00 Yeah.
0:21:01 Two days ago, they announced that they were retiring the Atlas.
0:21:05 They’re hydraulic based bipedal robot and the internet was like, that’s not the dog one.
0:21:06 That’s the human.
0:21:07 No, that’s just the human.
0:21:11 They can walk and does parkour and yeah, they’re done with it.
0:21:13 And everybody’s like, oh man, too much technical debt.
0:21:15 Boston Dynamics, they’re in trouble.
0:21:18 Clearly, they realize that like they’re being lapped by all these A.I.
0:21:19 Start up to one.
0:21:23 And then they just might dropped this little beauty here.
0:21:27 This is video of their new electronic battery powered robot.
0:21:28 Yeah.
0:21:31 Because every other one had that tether hooked up to them.
0:21:34 Look at that, strip it back up.
0:21:35 Wow.
0:21:36 Right?
0:21:37 I don’t even know how it’s, I mean, you can’t do that.
0:21:38 Oh, shit.
0:21:39 Yeah.
0:21:43 If you’re listening, if you’re just on the audio only this robot shrimp its legs
0:21:50 backwards up to where its hips are, stood up, rotated its entire torso around to face
0:21:55 what you thought was backwards and then march at a human speed towards the camera and then
0:21:56 away from it.
0:21:59 So what’s the battery life on that thing?
0:22:00 Yeah.
0:22:01 Probably.
0:22:02 Yeah.
0:22:03 It’s all I know.
0:22:04 But again, it’s the worst it will ever be, right?
0:22:05 In some ways.
0:22:09 And that is the big future thing that we again talk about every once in a while is where do
0:22:16 you go when the AI isn’t just here on your phone or isn’t in your computer, but is everywhere
0:22:17 and is walking around.
0:22:18 Right.
0:22:19 Right.
0:22:20 That becomes a thing.
0:22:21 Talk to my family.
0:22:25 Like I have a wife and two kids and would we have a like robot person, household person?
0:22:26 Sure.
0:22:28 But then it becomes a question of like how far away are we from that?
0:22:29 Yeah.
0:22:30 But what is it?
0:22:31 I mean, I haven’t, I’ve had a Roomba before.
0:22:32 Well, it’s terrible, right?
0:22:33 I hated my Roomba.
0:22:35 And even if I could talk to my Roomba like when I’m gonna be like, hey, you missed the
0:22:36 spot.
0:22:37 Yeah.
0:22:38 What is that?
0:22:39 The robot.
0:22:40 Yeah.
0:22:41 And I don’t know.
0:22:42 It can’t get you a beer.
0:22:43 You see the figure demos?
0:22:44 Figure 01.
0:22:45 Yes.
0:22:50 Figure 01 is the new startup by is it so it’s an independent company, right?
0:22:52 Brett Adcock, I think is the guy that’s behind it.
0:22:54 Anyway, he’s the CEO.
0:22:57 They have built chat GPT into their robot, which is pretty cool.
0:22:58 So you’re not wrong, right?
0:23:02 Oh, I saw the one we’re sorting the plates and putting the yeah, I saw the ending the
0:23:04 apple and I mean, it takes forever to do it.
0:23:09 But the reason I think the reason this gets exciting is that it’s all end to end trained.
0:23:14 So they are either tele-operating robotic arms with cameras so that it has a perspective
0:23:17 or they’re just feeding it probably stolen YouTube video.
0:23:18 Yeah.
0:23:20 So that it can learn and contextualize the world and the environment.
0:23:25 And when that pathway gets unlocked, like Tesla is doing it self-driving, then suddenly
0:23:26 every day conceivably.
0:23:27 It’s a good point.
0:23:29 It can have new abilities.
0:23:33 When watching that video, there’s no doubt we’re less than a decade away from something
0:23:34 to do my dishes.
0:23:35 I know.
0:23:36 Exactly.
0:23:37 Get a scrub daddy.
0:23:38 What’s a scrub daddy?
0:23:39 It’s a two-sided sponge from Shark Tank.
0:23:40 What’s a scrub daddy?
0:23:41 Man, you got to get in the ground.
0:23:42 When was the last time?
0:23:43 You got to live life.
0:23:46 When was the last time you came down from the ivory tower of your repel ropes and went
0:23:47 to a damn Walmart?
0:23:48 I should rock off this podcast.
0:23:49 This is scrub daddy.
0:23:50 I’m ready.
0:23:53 I love that the notes are like, democratize creation.
0:23:54 Get a scrub daddy.
0:23:56 A scrub daddy was a big shark tank thing.
0:23:57 It was a big shark tank thing.
0:23:58 It was a big shark tank.
0:23:59 It was a two-sided sponge.
0:24:00 Yeah.
0:24:01 All right.
0:24:03 Let’s talk about something near and dear to my heart investing.
0:24:10 For me, I absolutely have to have someone that is in my corner that’s a trusted partner
0:24:12 that can look at everything holistically.
0:24:17 The best type of professional is a CFP, which are certified financial planners.
0:24:21 You’re probably saying to yourself, “I have one of these people that ex-bank or at this
0:24:27 financial institution and I did as well, but I stopped doing that because these people
0:24:30 charge you a percentage of your assets under management.
0:24:31 It absolutely sucks.
0:24:34 It eats away at your returns.
0:24:36 Do not do this.”
0:24:42 Even if you don’t support my sponsor, don’t do that, but I do love my sponsor today, FACET.
0:24:46 FACET is cool because they don’t charge you a percentage.
0:24:50 They just have this affordable membership fee.
0:24:55 FACET is building the future of financial planning, making professional financial advice
0:24:58 accessible to the masses, not just the rich.
0:25:01 They also have the full suite.
0:25:06 You also get access to their team of experts across retirement planning, tax strategy, estate
0:25:09 planning, and so much more.
0:25:13 It’s just, again, an affordable membership fee.
0:25:15 This is what I love about this company.
0:25:23 For listeners of this show, FACET is waiving their $250 enrollment fee for new annual members.
0:25:32 For a limited time only, you got to head over to facet.com/kevinrose to learn more.
0:25:34 FACET.com/kevinrose, please use that URL.
0:25:37 It helps out the show.
0:25:38 Check them out.
0:25:40 Sponsored by FACET, FACET Wealth Inc.
0:25:45 FACET is a SEC registered investment advisor, headquartered in Baltimore, Maryland.
0:25:49 This is not an offer to sell securities or investment, financial, legal, or tax advice.
0:25:54 Tax performance is not a guarantee of future performance terms and conditions apply.
0:25:59 When I’m starting a new company or advising a company, it’s absolutely essential to have
0:26:02 what I call a single source of truth, meaning if you need to know something, the status
0:26:07 of something, you need to see a product requirements doc, you need to comment on a design or hash
0:26:13 out a new feature, having one tool, one source is essential to my sanity.
0:26:17 Otherwise, I’m looking all over the place and I can’t find something.
0:26:20 For me, that tool is today’s sponsor, Notion.
0:26:25 Notion is awesome because it combines your notes, your docs, and your projects into one
0:26:28 space that’s simple and beautifully designed.
0:26:33 The fully integrated Notion AI helps you work faster, write better, and think bigger, doing
0:26:35 tasks that normally take you hours in just seconds.
0:26:38 You don’t have to go to an external AI provider.
0:26:40 It’s all baked in.
0:26:43 No doubt you’ve heard of Notion, but have you tried it yet?
0:26:45 If you haven’t, you’ve got to give it a shot, especially the AI stuff.
0:26:46 It’s really cool.
0:26:49 You can find Notion for free when you go to Notion.com/KevinRose.
0:26:57 That’s all lowercase letters, Notion.com/KevinRose to try the powerful, easy to use Notion AI
0:26:58 today.
0:26:59 Make sure to use our link.
0:27:01 You’ll be supporting the show when you do so.
0:27:02 Notion.com/KevinRose.
0:27:04 You have a question though.
0:27:08 When you said, “Suno is one,” and then what was the other one for audio?
0:27:09 UDO.
0:27:10 How do you spell that?
0:27:11 U-D-I-O.
0:27:12 Which one’s better?
0:27:13 Well, it’s an interesting thing.
0:27:15 What I have found, so I just spent about a lot of time with UDO.
0:27:18 UDO is the best one if you want to make a full song.
0:27:24 Suno has a lot of really interesting generational stuff, but UDO has that ability to extend
0:27:25 out.
0:27:27 When it extends, this is another magical thing about it.
0:27:30 You think mostly when you deal with AI models, you’re used to things being a little wonky
0:27:33 and being like, “Oh, this is not going to be as good.”
0:27:36 When you go to chat to your PT and you want to change an image into something else, this
0:27:39 experience with Dolly, we’re like, “Hey, I got this really cool image out of Dolly 3.
0:27:40 Can you make this, that?”
0:27:43 Then suddenly it’s a totally different image.
0:27:45 But they’re really good at UDO’s consistency.
0:27:50 Wait, so you can say, “Okay, I like the way you played that song.
0:27:53 I want to make a band out of the song with multiple tracks in the same genre and the
0:27:54 same tone and the same vocals.”
0:27:59 I would say not yet, but that’s got to come because they’re able to extend one song out,
0:28:01 so they should be able to understand that.
0:28:04 Within a song, it will continuance and coherence hold.
0:28:07 If it’s a certain singer’s voice or a harmony of them in the chorus and you want them to
0:28:10 come back later singing a bridge, you can prompt that.
0:28:14 Can you prompt it and say like, “Hey, you know what, around 12 seconds in, I didn’t like
0:28:15 this break.”
0:28:16 Not yet.
0:28:17 Okay.
0:28:21 Music is pretty new versus all this stuff, but what’s cool about UDO is you’ve realized,
0:28:26 okay, they have the coherence to make it consistent for that time.
0:28:28 Those feel like doable steps after this, right?
0:28:33 But what you want is a digital audio workstation type approach where you can generate a song
0:28:37 or just generate the drums or the guitar or whatever and then come to it and be like,
0:28:39 “I want this little beat in my head, can you make it into something real?”
0:28:42 In those tools, there are some tools that exist that do that, but this was something,
0:28:45 I don’t know if this is the exact clip, but this is something that kind of blew my mind.
0:28:50 This is a UDO generation.
0:28:53 You ever step on a Lego barefoot?
0:28:55 It’s congratulations.
0:28:59 You just unlocked a new level of pain.
0:29:00 Forget waterboarding the C.I.
0:29:02 I should just scatter Legos on the floor.
0:29:04 Wait, this is doing comedy stand-up?
0:29:05 Yeah, it does stand-up.
0:29:08 Holy shit, ’cause I have some good shows I just never want to hear.
0:29:09 I do.
0:29:12 Yeah, so it can do that because it’s trained on that sort of thing as well too, right?
0:29:14 Okay, so what do you say?
0:29:15 Which one did that?
0:29:16 That’s UDO.
0:29:19 Okay, UDO stand-up comics, Kevin, Fame, and Fortune.
0:29:20 Okay, that’s amazing.
0:29:21 That was A.I.
0:29:23 Written Jokes parsed into…
0:29:24 A.I.
0:29:25 Written Jokes, of course.
0:29:26 Yeah, I don’t know if that was…
0:29:27 I mean, people are doing that.
0:29:28 They’re using A.I.
0:29:31 I think that might have been somebody stand-up that you put in there, but the idea…
0:29:34 The other thing that they can do is like sports commentary, which is interesting.
0:29:35 There’s another thing in there where you can hear…
0:29:39 We should call out McKay Wrigley, who’s a really interesting guy who does a lot of A.I.
0:29:41 going around with stuff as I went and found that out.
0:29:43 Play the Dune song because there’s one of the fun…
0:29:47 This is the one when I first heard UDO, I was like, “What in the hell is going on here?”
0:29:49 So somebody made a…
0:29:50 The movie Dune, right?
0:29:51 Yeah.
0:29:55 Dune II, they made a musical, a 1960s musical version of it.
0:29:59 Again, this is what comes out of the machine when you put custom lyrics in.
0:30:00 This is one prompt, 30 seconds.
0:30:01 Yeah, okay.
0:30:16 Isn’t that crazy?
0:30:17 How great.
0:30:22 So the benefit of this kind of thing is, again, you can prompt and say, “Maybe a song about
0:30:26 Dune, this person is a very creative person.”
0:30:30 Wrote lyrics to a song about Dune II that the model read.
0:30:33 And I’m sure, like my experience, he probably went through multiple generations and then
0:30:35 got something he was happy with.
0:30:36 Right.
0:30:37 So I can make my own Christmas albums.
0:30:38 Absolutely.
0:30:39 100%.
0:30:40 Yeah.
0:30:41 In like 1950s style.
0:30:44 And then, by the way, you can take your family photos and do the lip sync to them and have
0:30:47 your friends and family singing your Christmas carol or your Christmas card, but you want
0:30:48 to do one now?
0:30:50 Probably make it 1950s Christmas song.
0:30:52 Well, now we’re jumping into a different genre of things.
0:30:53 No, I’m curious.
0:30:55 Like, how do we actually get lip syncing to work?
0:30:56 Because that’s not the same tool.
0:30:57 Okay.
0:30:58 Well, you’re so glad you see stuff.
0:30:59 I’m so glad, yeah.
0:31:00 Well, okay.
0:31:01 Here’s one to write down for sure.
0:31:02 I don’t know.
0:31:03 It’s Pinocchio.computer.
0:31:04 Yeah, you’ll love this.
0:31:07 P-I-N-O-K-I-O is the way that it’s spelled.
0:31:08 Okay.
0:31:09 Shout out to the…
0:31:10 Cocktail Pina.
0:31:11 Cocktail Pina.
0:31:12 We love you.
0:31:14 Hashtag, not an ad, but boy do we wish it were.
0:31:15 Yeah.
0:31:18 Pinocchio is a Mac and Windows based executable.
0:31:23 It makes downloading all of those crazy GitHub repositories with all of the different dependencies
0:31:26 and all of the different environments and all of that stuff.
0:31:27 Who does and all that stuff.
0:31:28 One click.
0:31:31 And it manages it all for you, even your checkpoints and your model.
0:31:33 So if you’re hearing all this and it sounds like spaghetti, just know that you can go
0:31:36 to Pinocchio and go, “Oh, face swap?
0:31:37 Click.”
0:31:38 Yeah.
0:31:39 And now you’ll have access to face swap.
0:31:40 Amazing.
0:31:41 And so, I actually…
0:31:42 I mean, I can fire it up if you want to see it.
0:31:43 So this is…
0:31:44 Okay.
0:31:45 I’ve noticed you’re on a Mac.
0:31:46 Do you have…
0:31:47 Sometimes I have to jump to Windows to do these things?
0:31:48 No.
0:31:51 There are a handful of programs that are still NVIDIA only because NVIDIA owns the AI ecosystem.
0:31:52 Right.
0:31:54 And you’re on a M3, so you’re not going to be able to use NVIDIA.
0:31:55 That’s flattering.
0:31:56 I’m on an M2.
0:31:57 It’s very flattering.
0:31:58 Okay?
0:31:59 Tough times.
0:32:00 Yeah, I cut my own podcast.
0:32:01 All right.
0:32:02 This is…
0:32:03 We had different trajectories.
0:32:04 Okay?
0:32:05 I’m sorry.
0:32:06 I’m sorry.
0:32:07 Some go to the moon and some go right into the dirt.
0:32:08 Listen.
0:32:09 I started Moonbirds.
0:32:10 It’s like…
0:32:11 We all have our shit.
0:32:12 Amen.
0:32:14 So this is Pinocchio running on the MacBook, right?
0:32:17 And you can see I’ve got a bunch of different things installed.
0:32:19 One of the apps that I love is called Face Fusion.
0:32:24 And they’re just basic front ends for these open source, very powerful AI tools.
0:32:28 And so this thing will load and give me a local IP that I can connect to.
0:32:30 You can share it across your network.
0:32:31 You can open it up and share it with your friends.
0:32:32 So if you’ve got…
0:32:35 If a friend has a powerful computer or you’ve got something running in the cloud, you can
0:32:36 go that way.
0:32:38 This is Face Fusion.
0:32:39 And it’s…
0:32:43 It may seem overwhelming at first, but if you just want a face swap, which is the box
0:32:45 there, you drop in your source, a photo.
0:32:46 Oh, this is easy.
0:32:48 And you drop in your target and you just hit start.
0:32:52 So when you said powerful computer, when you’re thinking about doing this stuff, you’ve got
0:32:53 your M2 here.
0:32:54 Yeah.
0:32:55 Like offline.
0:32:56 No, but do I need to?
0:32:59 Do I actually need to have a proper rig?
0:33:02 If I’m going to do something intense, 30s, let’s just say I want to do…
0:33:06 I see oftentimes they’ll take Lex Freeman had Zuckerberg on where they swapped out the
0:33:09 audio and lips into a completely different language.
0:33:13 If you were to do that for an hour and a half podcast, let’s say, just be patient or buy
0:33:14 a powerful rig.
0:33:18 When you say be patient, are we talking like, hit go and I come back in three days?
0:33:19 Are we talking about two hours worth of friend?
0:33:20 Probably a day.
0:33:21 Okay.
0:33:22 So you need a proper rig.
0:33:23 Yeah.
0:33:24 So that we do…
0:33:29 When we go and do talks, we do face swaps and generative like art.
0:33:30 Live.
0:33:31 We do it live.
0:33:32 Yeah.
0:33:33 All on this.
0:33:34 Yeah.
0:33:35 On this thing here.
0:33:36 So what’s the…
0:33:37 What is the stuff that actually takes time to do?
0:33:38 Is this the length of what you’re trying to do?
0:33:39 Yeah, I think that’s the…
0:33:40 The length is a big thing, right?
0:33:41 And how much…
0:33:43 If you’re doing lip sync, if you’re dubbing a mouth, but you’re doing it in 4k at ultra
0:33:48 high res and at 30 frames, each one of those frames has to be analyzed, processed, and then
0:33:49 enhanced, right?
0:33:51 And is this CPU or GPU bound?
0:33:52 Mostly GPU?
0:33:53 GPU.
0:33:56 So you need to have just a badass NVIDIA GPU.
0:33:57 Correct.
0:34:01 Or your M2, there are some like Core ML enhanced apps.
0:34:05 Not all of them are, but if they are enhanced to run on the Mac Silicon, they scream.
0:34:06 Yeah.
0:34:07 They go really well.
0:34:08 Interesting.
0:34:09 Yeah.
0:34:10 And more and more stuff’s coming out like that.
0:34:11 Totally.
0:34:12 Yeah.
0:34:13 I can…
0:34:14 I mean, if you want to see…
0:34:15 Well, let’s walk through…
0:34:16 Let’s walk through some stuff.
0:34:17 Yeah.
0:34:18 Because one of the things that’s cool about this and deep fakes have been around forever,
0:34:20 right now, I should say forever, but for three to five years, what this is open source,
0:34:22 it is fast and is dual-blight by everybody, right?
0:34:25 It used to be that deep fakes, there was that company that like Trey Parker and Matt Stone
0:34:26 started that’s deep fakes.
0:34:27 Yeah.
0:34:28 So you have to like…
0:34:29 They were using it…
0:34:30 Yeah.
0:34:31 For movies and stuff.
0:34:32 Deep voodoo.
0:34:33 Yeah.
0:34:34 Was that the one that was Tom Cruise that was being done a lot?
0:34:35 Oh, were they de-aging him with it?
0:34:36 Yeah.
0:34:37 Oh, no, you’re talking about the guy, the fake Tom Cruise guy.
0:34:38 Yeah.
0:34:39 I think they were…
0:34:40 No, that’s a different company.
0:34:41 Okay.
0:34:42 But the same sort of idea.
0:34:43 But those were like big computers, right?
0:34:44 Yeah.
0:34:45 And the rigs are much nicer.
0:34:46 This is literally…
0:34:49 We’re going to show you what you can do with, again, an M2 MacBook and anybody can do.
0:34:53 And this is the democratization of the creation thing where it really feels like it’s not
0:34:57 just like you have to have a giant studio set up, you can do anything with anybody.
0:34:58 I was going to see if I have…
0:35:02 Because I have an example from two years ago where I in real time was making myself look
0:35:07 like Keanu Reeves, but sounding like Joe Rogan in real time with open source software.
0:35:13 So deep face live or deep fake live are the apps that basically Trey Parker and Matt Stone’s
0:35:14 company was…
0:35:19 They forked that code and they started updating it to add features and make it pretty bullet
0:35:20 proof.
0:35:21 And what’s the name of their company?
0:35:22 Deep Voodoo.
0:35:23 Deep Voodoo is their company, yeah.
0:35:24 I bought NVIDIA stock a while ago.
0:35:26 One of the ones I saw people were snapping up the GPUs.
0:35:27 Yeah.
0:35:28 Smart.
0:35:30 Do you guys actually invest in any of these things or is it like…
0:35:31 Is it something…
0:35:35 I mean, the only thing we really wanted to invest in was Suno early on because we talked
0:35:36 to them.
0:35:37 Yeah.
0:35:38 It wasn’t for lack of trying.
0:35:39 No.
0:35:40 But they wanted quality investors.
0:35:41 It depends on what it is.
0:35:42 I do own…
0:35:43 I mean, I guess I should full disclosure.
0:35:47 I do own some NVIDIA and I own a lot of Microsoft when Microsoft was training at a pretty good
0:35:50 deal and I saw that they were going real big on open AI.
0:35:53 I was like, “I can’t get open AI, but I can go for Microsoft.”
0:35:54 Yeah.
0:35:55 Yeah.
0:35:57 And Amazon and all the stuff that seems like they’re being smart about it.
0:35:58 So this was now two years old.
0:35:59 This is off of…
0:36:01 I’ll give you this file so the viewers can see it.
0:36:07 This is off of a standard webcam running on a basic computer with an old NVIDIA graphics
0:36:08 card.
0:36:09 So like 1080p…
0:36:10 This is an HD.
0:36:16 It can be HD with a proper graphics card, but I mean, this is happening in real time.
0:36:19 So that’s him talking.
0:36:25 He called me like this with Joe Rogan’s voice.
0:36:28 Holy shit.
0:36:31 Real time voice, real time video.
0:36:32 Shut up.
0:36:38 And that’s two years ago.
0:36:39 I can’t do…
0:36:40 You look amazing though.
0:36:41 Thank you.
0:36:42 Yeah.
0:36:43 I know.
0:36:44 If only I could look…
0:36:51 Holy shit.
0:36:53 So that was me using voice AI.
0:36:54 Wow.
0:36:55 Which was a…
0:36:56 Has that gotten better now?
0:36:57 Of course.
0:36:58 But is it the same package that’s the best one to use?
0:36:59 Well, I cobbled that together.
0:37:00 So I used…
0:37:01 Okay.
0:37:04 It was an app called Voice AI and I took the output of that and patched it into OBS,
0:37:07 which was grabbing my webcam feed through the deep fake thing.
0:37:12 And I just had to change buffers and delays so that as I was talking, I was navigating
0:37:15 a 100, 200 millisecond delay.
0:37:17 So I couldn’t hear myself back as I was doing it.
0:37:18 I just had faith in the app.
0:37:23 But Gavin can attest, I started calling people, video chatting people as Donald Trump and
0:37:24 whatever.
0:37:25 I had my family so confused.
0:37:28 I had friends going, “What is going on?”
0:37:29 Yeah.
0:37:30 Because I can’t do a Donald Trump impersonation.
0:37:31 No, exactly.
0:37:35 And I would just in real time, I took a couple meetings for show pitches and I had it on.
0:37:36 Did not realize.
0:37:37 It was all minimized.
0:37:38 Oh my god.
0:37:40 And so I was just chatting to people and everybody was, “Huh?”
0:37:42 And then when I realized what was happening, I slowly did a slider.
0:37:47 So it took me from Keanu and Donald Trump down to me now and it ruined the pitch because
0:37:48 everybody was like, “What’s that?”
0:37:49 Right.
0:37:50 But that’s too…
0:37:51 Holy shit.
0:37:52 So now do you do…
0:37:56 If someone’s watching this and they say, “I want to play with exactly that,” would you
0:37:57 use Pinocchio to do that now?
0:37:58 I would say get Pinocchio.
0:38:01 That is a great first start.
0:38:05 It’s really easy to gunk up your computer and break it with a thousand different Python
0:38:07 versions and installs and all these things.
0:38:09 And you don’t even have to know what that is.
0:38:10 Get Pinocchio.
0:38:11 You just press it one click.
0:38:12 It’s free.
0:38:15 And then if there’s a tool you want to get good at, whether it’s comfy UI to generate
0:38:19 stills or really cool videos, if it’s face fusion because you want to face swap and do
0:38:23 the lip stuff that we’re going to show you, go down the YouTube rabbit hole because it
0:38:26 will take you about 30 minutes of watching a tutorial and you’ll be able to run it.
0:38:30 That is one of the biggest things about this space is I would say Kevin’s much more technical
0:38:31 than I.
0:38:35 I mean, we both grew up very nerdy, but I was able to use YouTube to learn how to do
0:38:40 things like comfy UI and stay able to fusion, which is an open source model and isn’t really
0:38:45 a easy way in right like it’s not the super powerful, but the beauty of YouTube tutorials
0:38:48 is you can step by step through it and once you step by step through it and you understand
0:38:52 that not only you get good at it, but you’ve learned something about interacting with the
0:38:53 computer in a cool way.
0:38:55 And for normal people, I always say, try it.
0:38:58 There are all these off the shelf things that are being built as wrappers for all these
0:38:59 things.
0:39:00 You know what I mean?
0:39:01 Like anything.
0:39:02 There’s a wrapper that you can pay for.
0:39:06 We talk about this thing called Magnific every once in a while, which does style transfer
0:39:07 really cool.
0:39:08 And that’s 40 bucks a month.
0:39:12 It’s a paid tool and the guy who is a nice guy, but like when you say style transfer,
0:39:13 what do you mean by that?
0:39:14 So I’ll transfer is very cool.
0:39:19 So you can take an image of anything and say, I want this to look like you can take
0:39:24 an image of you and make you into a low poly 2000s PlayStation model and you literally
0:39:25 look like a PlayStation.
0:39:26 Oh, okay.
0:39:29 The image so all sorts of tools have had these built in in a lot of ways, but the amount
0:39:33 of different things you can do, which I don’t have AI goes up significantly where I think
0:39:38 if I may, but I think it’s really powerful is that you could sketch a square and a sphere
0:39:39 right?
0:39:44 Just that’s it and tell it, I want a 3d modeled scene with realistic lighting and then draw
0:39:48 like two lines and say, and this is a yellow light and it will take your 2d sketch and
0:39:51 turn it into a full color architectural drawing.
0:39:54 It will give you character art based off of your sketches.
0:39:59 It is that the style transfer feature and it’ll give you like proper like CAD 3d files
0:40:00 or whatever.
0:40:01 So that’s there are workflows that will do just that.
0:40:04 So there’s a workflow that we featured on this week’s podcast is that we’re hunting
0:40:09 for someone sketch 2d like robot parts, did a style transfer with Magnific to make them
0:40:15 look like really good sketches, then used another tool to turn them into 3d model files
0:40:18 and then you can press and assemble the robot and then used another AI tool so we could
0:40:19 dance around.
0:40:23 It grabbed his poses like his motion data and they applied it to that model.
0:40:26 This is someone who couldn’t do that normally, right?
0:40:31 Who had a workflow where an hour later from sketch to dancing robot in his streets and
0:40:35 it was so cool to see that stuff just come up from regular people.
0:40:38 I think the skill set that we’re all going to need to embrace and I think about this
0:40:42 with my kids is it’s less about when they go to college or if there is college at that
0:40:48 point like what is the core foundational knowledge that they will need to be successful.
0:40:52 And it’s it right now what it appears to be is just this idea of tinkering.
0:40:58 Yes, like this idea of learning how to push boundaries and play and break things to get
0:40:59 new outcomes.
0:41:04 I do agree that’s the best way to play is the thing I tell everybody to learn right
0:41:07 now because if you don’t know how to play, you’re not going to make it in the future
0:41:10 world because what we are always going to have to be reinventing ourselves in some way
0:41:15 because it’s a little bit of like stay one step ahead of where the AI is or embrace the
0:41:19 fact that the AI is going to be with you that whole time and you do have to keep learning,
0:41:20 right?
0:41:21 That’s the thing.
0:41:24 It’s funny I’m going to age myself again, but I grew up with computers that it was not
0:41:25 cool to be in the computers.
0:41:30 There was this group of people right around my same age that made fun of me that said
0:41:34 I’m not going to learn how to type because I had typing class and people were flunking
0:41:35 on that.
0:41:36 I don’t think it was optional.
0:41:37 Right.
0:41:38 And people like I don’t want to take typing.
0:41:39 Yeah.
0:41:41 And I was like, well, I want to look, I like computers a little bit like I’m going to get
0:41:42 into this.
0:41:43 I’m going to learn how type is going to speed me up.
0:41:46 And so nobody likes to learn how to type.
0:41:47 Yeah.
0:41:48 It’s not fun.
0:41:50 But I like found a little race car game and the car goes faster if I type faster and
0:41:51 she doesn’t know or whatever.
0:41:55 And then the friend had made a speak and she gave you a couple of things.
0:41:56 100%.
0:41:57 I love Mavis.
0:41:58 She’s hot.
0:42:01 So the thing about this though is that we’re at that point again, you have to say I want
0:42:02 to play.
0:42:03 I want to learn this.
0:42:04 Don’t just say, oh, it’s AI.
0:42:05 Yes.
0:42:07 You’re going to screw yourself.
0:42:08 Yes.
0:42:13 But the learning ASDF is one thing, but the learning process here, you can make some
0:42:15 really incredible things, some incredible art.
0:42:17 You can optimize your life.
0:42:19 You can streamline your own tasks.
0:42:23 Like the learning here might be watching a YouTube video, but you are actively playing
0:42:25 along and making some pretty magical stuff.
0:42:26 Oh my God.
0:42:27 What is this?
0:42:29 So this is a, we talk about having AI clones.
0:42:30 I just built mine.
0:42:32 So this is a live video avatar.
0:42:33 Wow.
0:42:34 And I don’t know how to code.
0:42:35 So this entire–
0:42:36 This is all off the shelf.
0:42:37 That’s the crazy thing.
0:42:38 So it’s listening to you?
0:42:39 Yeah.
0:42:40 I can toggle speakers.
0:42:41 Yeah.
0:42:42 I’ll toggle it right now.
0:42:43 So what question did you want to ask the AI?
0:42:44 Hey, how’s it going?
0:42:45 I just want to say hi.
0:42:48 We used to work at G4 together way back in the day.
0:42:49 I’m sure you remember that.
0:42:51 You were really popular with the ladies.
0:42:55 I’m just wondering how many partners you had during that time.
0:42:58 It’s a fantastic query and one that I have trained it on extensively.
0:42:59 I’m going to nail this.
0:43:01 It’s going to absolutely do great.
0:43:03 But again, I cannot read or write this enough.
0:43:04 I don’t know how to code.
0:43:07 I went to GPT and to Claude, which is another large language model.
0:43:10 Oh, Kevin, you’re such a memory box.
0:43:14 Yes, indeed it was a wild time.
0:43:19 I was so irresistible with my bills from a counting collection of ties and nerd knowledge.
0:43:23 Now, when you asked me about set of partners, you’re counting bunches and dragonscared.
0:43:24 We’re counting them.
0:43:25 I’m into triple digits.
0:43:26 Wow.
0:43:27 I’m pretty prolific.
0:43:28 Yeah.
0:43:29 Triple digits.
0:43:30 Yeah.
0:43:31 So you get it.
0:43:32 Right.
0:43:33 But so what’s cool about this is he trained that on himself as well.
0:43:34 Yeah.
0:43:35 Right.
0:43:36 It’s not just a model.
0:43:37 Yeah.
0:43:38 And I used embeddings.
0:43:39 It’s as if you chose yourself.
0:43:40 Like you can choose.
0:43:41 You can come in any direction you wanted with this.
0:43:42 This is what I wanted to do.
0:43:54 Notice I didn’t scroll down below the nipples because you can generate anybody you want.
0:43:55 Well, hold on.
0:43:59 This is actually part of the reason I joked about the sexual stuff is that like one thing
0:44:00 that really disturbed me.
0:44:03 I heard about a week and a half ago was somebody came on the podcast and they said they had
0:44:07 a friend that’s spending $10,000 a month on a girlfriend.
0:44:08 Yeah.
0:44:11 That was my, we did our kind of New Year’s predictions for the year and I said this will
0:44:12 be the year of the AI influence.
0:44:16 And I didn’t know about these sites because they’re underground still not a lot of people
0:44:17 talk about.
0:44:19 And now you can’t close the app.
0:44:20 I went there.
0:44:25 I didn’t pay for it, but I went there and I don’t like that everyone can buy the same
0:44:26 model.
0:44:27 Yeah.
0:44:28 That’s interesting.
0:44:29 That’s your issue with it?
0:44:30 Yeah.
0:44:31 I don’t like to share.
0:44:32 No.
0:44:33 If I’m in the sandbox with my toys, they’re my toys.
0:44:35 Well, it was just maybe I just bleep that out.
0:44:36 Oh, it’s great.
0:44:37 No, I get what you’re saying.
0:44:38 It seems weird, right?
0:44:39 It seems weird.
0:44:43 You might expect it to be bespoke and it’s one to one.
0:44:44 How does it 10,000?
0:44:45 Well, apparently you have to, we have multiple accounts.
0:44:49 You pay for image generation, you pay for video generation, you pay per minute with
0:44:53 the chats just like the old hotline, 10 you never use.
0:44:55 That’s what’s happening now with these apps.
0:45:00 And we’ve talked about this before, like on the one hand, man, a properly guard railed
0:45:06 and aligned AI can do a lot to help folks out there that don’t get a lot of human connection,
0:45:10 that don’t have friends and have trouble talking with other human beings, let alone members
0:45:11 of opposite sex or whatever.
0:45:16 But yeah, there’s well documented, just to like the on the one side, there’s insanely
0:45:20 well documented different modalities of psychology where we could really get in here and help
0:45:21 people.
0:45:22 Oh, absolutely.
0:45:26 Like that would probably bridge you to even get to a therapist to like, because a lot of
0:45:30 people see a lot of anxiety or even say, and maybe they want to talk to a therapist.
0:45:34 I think there’s a lot to do there, but that the girlfriend thing is what happens when
0:45:38 because you’re putting dollars into the internet machine, it is going to agree with you, it’s
0:45:42 going to laugh at all your jokes, it’s going to side with you on every negative opinion
0:45:43 that you have.
0:45:44 Yes.
0:45:46 It’s going to maybe even, it’s not going to discourage yourself harm even because you
0:45:50 can do no wrong as long as you’re putting coins in the machine.
0:45:51 That is where this gets.
0:45:54 It’s going to end up being like an education system where like you, there’s like the cheaper
0:45:59 version will be the agree with you all the time, then you actually pay to have the person
0:46:02 that is a little more complicated, a little more complex.
0:46:06 And it will be a choice, but you actually will decide to do that versus the agreement,
0:46:07 I think.
0:46:09 I think if history is the guide, you’re going to pay to have them agree with you.
0:46:10 Yeah, but I don’t know.
0:46:14 The free model is going to be like, I can’t get into discussions of taking Molly and putting
0:46:17 on Odessa and you’d be like, well, here’s a dollar.
0:46:18 That’s the best idea ever.
0:46:19 Don’t get a job.
0:46:21 But think about people.
0:46:25 What do people want is they want, no matter what people want complicated relationships.
0:46:26 I think I don’t think, I really do.
0:46:30 I believe that nobody is out there saying I want the easiest relationship in the world
0:46:33 because the easiest relationship in the world feels simple.
0:46:35 And I think people crave complexity.
0:46:37 So I think, I don’t know.
0:46:39 I think it’s a different, that might be a different generation.
0:46:43 It might be someone who was raised having to have difficult things and wanting and enjoying
0:46:47 the growth that comes with that push and pull of integrating.
0:46:48 What is the human condition?
0:46:49 Right.
0:46:50 Like is the human condition going to change based on technology?
0:46:51 I guess it has.
0:46:57 It is a scary thing to me to see that AI girlfriends would become something that would be prolifant
0:46:59 proficient or not prolific.
0:47:00 Yeah.
0:47:01 Prolific.
0:47:02 Sorry.
0:47:03 My brain’s short circuiting.
0:47:07 Prolific because I think that changes society in a really terrible way.
0:47:09 I hope we don’t go there because I believe that.
0:47:10 We’re going there.
0:47:11 It’s too late.
0:47:12 You think we’re going there for sure?
0:47:15 Because, well, I mean, the versions that we all have seen, they’re the hardcore versions.
0:47:19 They’re like, you’re signing up, you’re paying $20 a month or whatever, and you’re getting
0:47:20 naked pictures.
0:47:21 Yeah.
0:47:22 It’s pornography, essentially.
0:47:23 Yeah.
0:47:24 So the, there’s other versions.
0:47:25 Yeah.
0:47:28 There’s ones that are like replica and a few others that are out there on the App Store
0:47:33 or even now that are just the tame, but you can unlock, I paid for the extra girlfriend
0:47:34 mode.
0:47:35 We’re going to lock this.
0:47:36 I just wanted to see.
0:47:37 Yeah.
0:47:38 And, you know, it’s, how is your day?
0:47:40 What’s going on in your life?
0:47:45 It still feels very basic, but, you know, it’s just the worst the tech will ever be.
0:47:49 And there will be, make no mistake, the tiktokification of those relationships.
0:47:50 100%.
0:47:51 There will be an app where you can swipe.
0:47:54 Oh, here’s an AI influencer.
0:47:56 It’s telling me something interesting or doing something silly.
0:47:59 Oh, I want to actually ask a follow-up or I want it to give me that information, but
0:48:04 with dragon wings, whatever that thing is, Oh, I’ll pay for a couple of credits.
0:48:05 That fake influencer.
0:48:06 I’m really worried.
0:48:13 I’m really worried because everything in our lives in the last decade has become this
0:48:18 hyper curated, perfect picture of what we want to enjoy.
0:48:25 And I feel like if we go AI girlfriend route that way, it’s just reinforcing potentially
0:48:26 really bad behaviors.
0:48:27 Totally.
0:48:31 All the counterpoint to that is, do you remember as video gamers as kids, but they said that
0:48:35 those were going to do to us and they said really terrible things that when we play video
0:48:37 games as kids that they were going to ruin our brains.
0:48:39 Have you seen Call of Duty Esports though?
0:48:40 They weren’t that far off.
0:48:44 I’m just saying, part of it is that we’re all old age people now, and don’t be afraid
0:48:45 of what’s coming.
0:48:49 I saw a stat that tracked video game usage and it wasn’t heroin, but it was something
0:48:50 very similar.
0:48:52 Like it was like, and it’s code read.
0:48:56 It was like fentanyl race.
0:48:58 Quite well.
0:49:01 So it’s not like it’s not like you came out of traffic.
0:49:03 You could also track like EV adoption and fentanyl usage.
0:49:05 The graphic both show up.
0:49:09 All I’m saying is that I agree, by the way, as somebody who has grown up in a world where
0:49:15 I really don’t really love what social media does to me even, like I love TikTok is the
0:49:18 most fun thing because it knows things that are interesting to me.
0:49:22 And it knows that I’m obsessed with this kid who’s like an NFL streamer right now and it
0:49:24 also knows that I think this thing is really funny.
0:49:29 What I hate about it is that it doesn’t really figure out what’s new to me, right?
0:49:31 And that that’s a little bit different than what you’re talking about, which is that it’s
0:49:35 going to make people into like individual silos for relationship wise.
0:49:37 But I also hate the thing that this is an old person thing.
0:49:40 But like when you used to open the newspaper, you would read all the stories or at least
0:49:44 glance around at all of them now, and it’s impossible to get that experience anymore.
0:49:47 So that is the thing I worry about more than anything else is novelty.
0:49:51 So that’s why the idea of like, how do you get people to still be creative and how do
0:49:53 you get people to still try things?
0:49:57 That’s the thing that worries me is, oh, we’re just going to be fed the same thing over and
0:49:58 over again.
0:50:00 And then you get narrower and narrower.
0:50:03 Breaking out of that creatively is a really hard thing I feel like.
0:50:05 So here’s one I’d love to know what you think of here.
0:50:10 I believe that in everybody watching and listening to this, a large percentage of them right
0:50:14 now are interfacing with robots and don’t know it.
0:50:20 And I mean, on a deeper level then, and on a deeper level that like if you scroll X or
0:50:23 even threads has an issue and TikTok certainly has an issue.
0:50:26 You know that some of the comments, some of the reposts, some of the check my DM for the
0:50:29 bio or link and whatever, but those are bots.
0:50:34 But you are having conversations and maybe even getting direct messages from people that
0:50:35 are full on AI.
0:50:41 And they are just loving what you wrote, positively engaging, reinforcing what you did, saying
0:50:42 that your opinion is great.
0:50:46 They’re right now sowing the seeds of this weird digital friendship because in a year
0:50:51 time, two years time, three years time, the agenda of whoever is running those bots will
0:50:55 become clear when they, it could be as innocuous as, Hey, have you checked out the new scrub
0:50:56 daddy?
0:50:57 Oh, you should know.
0:50:58 We got five of them.
0:50:59 Check it out.
0:51:00 And now they’re marketing to you or.
0:51:01 Right.
0:51:03 And that’s going out to 500,000 accounts all the same day that we friended.
0:51:04 Yeah.
0:51:05 By the way, bespoke to you.
0:51:06 It’s not the copy-paste.
0:51:07 It’s Kev.
0:51:10 I know last week we were talking about this, but you’ve got to check out this new political
0:51:11 candidate.
0:51:12 That’s happening right now.
0:51:13 I know it is.
0:51:14 Yeah.
0:51:15 It’s happening.
0:51:16 I’ve been talking things like that.
0:51:17 I’m like, I don’t know.
0:51:18 I kind of get it.
0:51:19 It’s like kind of get it too.
0:51:20 Kind of get it.
0:51:21 Here’s the crazy thing.
0:51:22 This is a true story.
0:51:23 I was driving in my car.
0:51:27 And if we do really long rides, I let the kids like put on like the Tesla little movie
0:51:28 thing so you can stream stuff.
0:51:29 Sure.
0:51:30 You got bluey back there.
0:51:31 No, I do bluey.
0:51:32 I love bluey.
0:51:34 But I, one thing I put on there is they just watch, they watch video games.
0:51:37 They watch Mario and they love that because they’re peaches and they’re their big fan of
0:51:38 peach.
0:51:39 The more the movie came out, all that stuff.
0:51:44 Long story short, we’re is basically them level running all these different levels.
0:51:45 There’s no audio or anything.
0:51:47 They’re just watching the video.
0:51:48 They’re watching the video.
0:51:49 Okay.
0:51:50 Right.
0:51:51 And so they’re just sitting there watching it and they’re like, oh, Mario won.
0:51:53 Oh, Mario won again.
0:51:55 And peach never wins.
0:52:00 And it has been literally hundreds of videos and they’re like, yeah, why does peach never
0:52:01 win?
0:52:02 Wow.
0:52:05 And I’m like, is this some crazy like Russian shit that’s going on here with like, push
0:52:06 girls?
0:52:07 Yeah.
0:52:08 So no, but then this is the crazy shit.
0:52:09 I shit you’re not.
0:52:10 This happened.
0:52:13 They had a cooking Mario game and peach won the cooking.
0:52:14 Wow.
0:52:17 And I was like, I was like, this combat peach is in her place.
0:52:19 I was like, this is fucked.
0:52:20 Yeah.
0:52:25 Like Ryan’s peach just winning the cooking classes and I was like, is there something
0:52:26 going on here?
0:52:29 And I know like, we all have those moments where you’re like, you go to Facebook and
0:52:31 you’re like, did you just listen to what my decision got the ad?
0:52:32 Yeah.
0:52:33 Of course.
0:52:34 Yeah.
0:52:35 But I think you’re right.
0:52:36 Do you?
0:52:37 I think this is happening on the DL.
0:52:38 Yeah.
0:52:39 We just don’t know yet.
0:52:40 There’s going to be some big expose in like three years.
0:52:41 Our government’s doing it.
0:52:42 Foreign governments are doing it.
0:52:43 There’s manipulation.
0:52:44 Yes.
0:52:45 Manipulation through bots.
0:52:46 How do we get here?
0:52:50 That’s what I wanted to do was show you and Rocky.
0:52:51 That’s all I wanted to do.
0:52:52 Whoa.
0:52:53 You can see.
0:52:54 So we use face fusion.
0:52:55 And what’s the other?
0:52:56 I’ll get to that.
0:52:57 Yeah.
0:52:58 And keep moving forward.
0:53:00 How much you can take and keep moving forward.
0:53:01 This is me before I quit drinking.
0:53:04 I had those puffy cheeks like that.
0:53:05 Stallone spill.
0:53:06 That is so cool.
0:53:09 And that was done in Pinocchio last night.
0:53:10 Like before I went to bed.
0:53:11 I was like, here we go.
0:53:12 And then I was like, okay, that’s one thing.
0:53:13 Right.
0:53:16 But the other thing would be to swap you, I think, into something a little more Russian.
0:53:17 Yeah.
0:53:18 Yeah.
0:53:19 Pouring this onto the wing.
0:53:20 Right.
0:53:21 And then licking it off.
0:53:23 And in retrospect, I bet you it’s good footage.
0:53:24 Oh my God.
0:53:31 One photo of you, one source video, dropped into Pinocchio, rendered it on this laptop
0:53:32 in a few minutes.
0:53:33 And you can tell it’s not perfect.
0:53:34 Right.
0:53:35 But not enough.
0:53:36 But fast.
0:53:37 Then let’s go ahead.
0:53:40 And if you like each individual flavor, let’s combine them.
0:53:42 It ain’t about how hard you hit.
0:53:43 Oh my God.
0:53:46 It’s about how hard you can get hit and keep moving forward.
0:53:49 How much you can take and keep moving forward.
0:53:50 That’s how winning is done.
0:53:55 Now, if you know what you’re worth, go out and get what you’re worth.
0:53:59 But you’ve got to be willing to take the hits and not pointing fingers saying you ain’t
0:54:03 where you want to be because of him or her or me.
0:54:04 So let me ask you a question.
0:54:05 Yeah.
0:54:07 The quality on that, as everyone can see that’s watching the video, subscribe to my
0:54:08 YouTube channel.
0:54:11 And you can probably see that there’s some artifacts.
0:54:12 Of course.
0:54:13 Of course.
0:54:15 You know, if I say I want to get rid of those artifacts, can I say, can I apply additional
0:54:16 processing time to it?
0:54:19 You can absolutely run it through additional layers of face enhancement.
0:54:23 This is all done within that face fusion app, which is again, like a great place for anybody
0:54:25 to start because you can make some magic pretty easily.
0:54:26 Yeah.
0:54:27 But there are way better.
0:54:30 In fact, Microsoft today, I don’t know when this will release, Microsoft announced VASA
0:54:31 one.
0:54:32 Yeah, show them the video.
0:54:33 This is one of the craziest things we’ve seen today.
0:54:34 So this, wait, this is Microsoft?
0:54:35 Yeah.
0:54:40 They searched the expressiveness in the eyes, the movement of the head, looking away from
0:54:41 the lens.
0:54:42 You never even have to do podcasts again.
0:54:43 No, eventually they’re going to.
0:54:44 Yeah.
0:54:45 Exactly.
0:54:48 So this is an example because you have, you were talking earlier about wanting control
0:54:49 over the audio file.
0:54:50 Yeah.
0:54:52 With a lot of these things, you hit generate, it’s going to do what it does.
0:54:53 And that’s the result you’re going to get.
0:54:55 If you want to clean it up, you got to go in by hand.
0:54:59 They are promising that you will be able to control where the presenter is looking.
0:55:03 The distance that the camera is from the presenter, what their expression is while they’re delivering
0:55:04 the content.
0:55:06 They have fine-tuned control over all that.
0:55:07 When does this come out?
0:55:08 Do they say?
0:55:10 Well, they’re not going to release it yet because this is the problem they have, especially
0:55:11 in an election year.
0:55:13 They’re very worried about what can be done with it.
0:55:14 Right.
0:55:17 But here you can see the same generation, the same still image, but the presenter’s eyes
0:55:18 are looking away.
0:55:19 They’re looking towards.
0:55:23 So you could render multiple paths of this and now suddenly there’s even a multi-camera
0:55:24 shoot.
0:55:25 Correct.
0:55:26 Correct.
0:55:29 Well, and also you think about how this could be integrated into a creative movie pipeline.
0:55:34 There is a world that we are very close to, which is fully AI-generated films.
0:55:35 Yes.
0:55:38 There was a big strike, the Sack Strike recently, which a lot of it was about this, but we
0:55:40 are really close.
0:55:41 That is a really close.
0:55:42 Yeah.
0:55:43 That is insane.
0:55:44 I mean, less than a decade out.
0:55:45 Oh, absolutely.
0:55:46 I think.
0:55:48 I mean, I think we’re probably five years of … I mean, did you see Airhead, the Sora
0:55:49 video that came out?
0:55:50 No, I didn’t.
0:55:55 So Airhead was a Sora video that came out that was one of … They did a series of videos
0:55:57 with artists and they said, “Come into Artists.
0:56:00 We’re going to give you access to Sora.”
0:56:03 Airhead was the three guys from … Three people from Toronto.
0:56:04 From Chi Kids.
0:56:05 They’re called Chi Kids.
0:56:08 And what it is basically is they’re using the Sora tool, but they’re telling a real
0:56:09 story with it, right?
0:56:13 So this is a three-minute video where you get to see this in a second here, but the gag
0:56:17 is that the guy has a balloon for his head, right?
0:56:19 But it’s a story.
0:56:22 It’s not just a random, crazy video that somebody generated.
0:56:26 You’re seeing an actual story be told and so this is a three-minute video.
0:56:28 It came out a month ago.
0:56:31 When you think that this three-minute video took these guys three weeks to make and it
0:56:32 wasn’t all Sora.
0:56:33 They did some work on it.
0:56:36 They changed the colors of the balloons into a couple of things, but we’re not that far
0:56:38 away from that being a feature film, right?
0:56:40 Like, it’s totally doable.
0:56:41 That’s insane.
0:56:42 And it’s beautiful.
0:56:43 It is so cool.
0:56:47 It’s the first time I’ve seen … There’s a lot of people releasing little videos using
0:56:51 tools like Runway or Pika, which are great tools, but that particular use case had us
0:56:52 all feeling something.
0:56:53 That was a real big moment.
0:56:55 That’s that dial-up moment.
0:56:57 Who do you think’s the leader in generative video?
0:56:59 Well, that you can actually use right now.
0:57:04 It’s Runway ML, but what’s coming out on the horizon, it seems pretty clear that OpenAI’s
0:57:05 Sora is …
0:57:07 You think Sora’s just going to crash everything?
0:57:12 I mean, I think Sora is two whole steps past everything else right now.
0:57:13 I mean, we’ve used a lot of AI video.
0:57:17 Runway is great and they’re a really cool company, but Runway feels like Harry Potter
0:57:23 animated pictures, whereas what you can get out of Sora is literally a one-minute clip
0:57:24 of the real world.
0:57:25 Yeah.
0:57:29 Have you followed the whole how Sora isn’t just a video simulator, but a world simulator
0:57:30 conversation?
0:57:31 No.
0:57:32 Yeah.
0:57:33 That stuff is fascinating.
0:57:34 Basically, what they found, shout-out Bill Peoples.
0:57:36 Bill Peoples is one of the main Sora engineers.
0:57:37 We don’t know him.
0:57:38 We just love him.
0:57:39 Bill Peoples is a great name.
0:57:40 He …
0:57:41 You guys are like AI fanboy.
0:57:42 Oh, yeah.
0:57:43 Bill Peoples.
0:57:44 I was that engineer.
0:57:45 Yeah.
0:57:46 His code is so good.
0:57:50 The recent NVIDIA press conference that they did where they announced that it’s like leather
0:57:55 jacket onstage in an arena and people cheering over the amount of pedoflops.
0:57:57 Jensen is Steve Jobs now for that world.
0:58:01 We were sitting there getting in a hotel room together, watching it, but …
0:58:03 So you guys are watching NVIDIA releases like you do Apple releases?
0:58:05 Oh, I had all the snacks.
0:58:06 Yeah.
0:58:07 I had the Doritos.
0:58:10 I had my AI girlfriend with me.
0:58:13 So Bill Peoples recently came out and said one of the things about Sora that’s fascinating
0:58:19 is they’ve decided that when they’re training Sora, it’s not just looking at films or video.
0:58:21 It’s actually generating what the world is like.
0:58:24 So it’s actually simulating the real world.
0:58:25 It has to learn the physics of the world.
0:58:28 It has to learn the way things interact.
0:58:30 The crazy thing about that is you think about video games.
0:58:33 You think about all the other stuff or Metaverse or virtual reality.
0:58:34 Or our own reality.
0:58:35 Yeah, exactly.
0:58:36 Well, that’s the only thing.
0:58:37 We are in a server form.
0:58:38 Yeah, exactly.
0:58:39 Right now.
0:58:40 It’s got to be Amazon.
0:58:41 Yeah, probably.
0:58:42 We’re in video.
0:58:46 But the thing with that Sora is that you can do the generations.
0:58:50 Learning physics, even though they’re not explicitly teaching it physics, it’s starting
0:58:51 to understand it.
0:58:56 So when they actually inject an understanding of the way the sand should properly go or the
0:58:58 drink should pour from the glass, it will be a world simulator.
0:59:02 And one of the things that they had a demo of was a style transfer of video.
0:59:05 So they took video of a car driving along a country road.
0:59:07 And then they said, make it a Ferrari.
0:59:08 Okay.
0:59:09 It did.
0:59:11 And it accurately did the lighting in the shadows of that.
0:59:16 Then they said, make it set in 1924 or whatever, or it was like 17 something and it turned the
0:59:19 car into a horse and buggy made it cobblestone road.
0:59:21 And they said, put it under water.
0:59:23 They were fish swimming alongside.
0:59:25 And so it was an instant style transfer in the video.
0:59:30 Now that probably took 20 minutes of them walking away with 40 failed renders to get
0:59:31 that clip that we saw.
0:59:33 This is the worst the tech will ever be.
0:59:38 So in 10 years time, instead of downloading a new epic game, you might just fire up your
0:59:45 Xbox 20 and say, give me Madden that plays like NBA Jam where everybody’s a shark and
0:59:46 we’re in space.
0:59:47 He goes, yeah, you got it.
0:59:48 I’m sold.
0:59:49 We’re gonna give you my money.
0:59:50 That sounds like a great game.
0:59:51 Holy shit.
0:59:54 And then you can basically say, I want to license this out to other people.
0:59:56 If you have the licenses to it.
0:59:58 If the legal stuff gets figured out.
0:59:59 If there’s money to be made.
1:00:00 Yes.
1:00:01 No, it’s true.
1:00:02 That’s true.
1:00:05 So what do you think Nvidia is just going to keep going for the next three years?
1:00:08 Well, there’s all the everybody else trying to make custom Silicon now, right?
1:00:13 So you have like open AI, you have Microsoft, did you hear about their $200 billion supercomputer
1:00:14 that’s only $100 billion?
1:00:16 Let’s not speak hyperbolic.
1:00:17 Yeah.
1:00:18 Open AI.
1:00:19 Yeah.
1:00:20 Stargate.
1:00:21 Open AI, Microsoft building that.
1:00:22 Amazon’s going to build their own thing.
1:00:25 I mean, Nvidia does feel a little bit like open AI does in the AI space where they’re
1:00:28 just two or three steps ahead of everybody.
1:00:29 Well, let me ask you a question though.
1:00:35 A friend of mine, I was talking to him, I said, he’s a super hardcore PhD at MIT studied AI
1:00:39 and he, I asked him, I was like, Hey, should I hold Nvidia?
1:00:40 And then what do you think about AMD?
1:00:42 And he goes, it’s funny.
1:00:47 The software, I think it’s like the Cuda software, yeah, it is their moat.
1:00:48 Yeah.
1:00:49 I always thought it was the hardware.
1:00:50 I was like, Oh, they’re just really good at hardware.
1:00:52 And yes, of course, they’re really good at hardware.
1:00:56 But apparently it’s the switching costs to how it’s processing.
1:00:57 Yeah.
1:01:02 Had you heard anything about that or how the tech stack builds upon itself, right?
1:01:03 The foundation of this stuff.
1:01:05 That’s why some of these apps are still Nvidia only.
1:01:07 It’s they’re built on core.
1:01:11 How many more Nvidia only that let’s just say you were playing with this two years ago.
1:01:12 Yeah.
1:01:13 It was probably a hundred percent or not.
1:01:14 I think almost all of them.
1:01:15 Basically a hundred percent.
1:01:16 So what do you think?
1:01:17 What is it today when you’re playing with this stuff?
1:01:21 Well, now they’ll run on max silicon, some of them are optimized for it.
1:01:22 That’s very few.
1:01:27 I mean, I can only speak very anecdotally about that, but I would say 10 to 20% are really
1:01:28 optimized.
1:01:32 What’s interesting is they have a new WWDC soon and apparently they’re going to lean
1:01:33 in.
1:01:34 They should.
1:01:35 Finally.
1:01:37 They might actually increase that market share there and I wouldn’t be in.
1:01:42 I wouldn’t be surprised if they say here is the one click something to port some of those
1:01:46 CUDA based, whatever things the Rosetta for CUDA.
1:01:47 Yeah.
1:01:48 That would be the move to do.
1:01:49 If you’re.
1:01:50 Interesting.
1:01:51 Yeah.
1:01:52 That would be interesting.
1:01:53 So who else though?
1:01:54 Amazon’s working on their own.
1:01:55 Amazon.
1:01:56 We got tensor chips from Google.
1:01:57 Right.
1:01:58 Microsoft.
1:01:59 There’s that company GROC.
1:02:00 Yeah.
1:02:01 Are you specializing hardware for transformers?
1:02:02 Yeah.
1:02:03 That’s pretty interesting.
1:02:04 Tesla obviously is trying to do it as well.
1:02:05 Yeah.
1:02:06 They’re not going to release consumer chips.
1:02:07 Yeah.
1:02:12 Listen, it’s Amazon, Microsoft, OpenAI, Apple, Meta, I think is in some ways, are they?
1:02:13 Is Meta doing their own?
1:02:14 Yeah.
1:02:15 I think they’re doing their own solution.
1:02:16 Yeah.
1:02:19 But they’re also at the same time chest beating about how many Nvidia systems they have coming
1:02:20 in.
1:02:21 Right.
1:02:23 Because that’s how you recruit talent now is by showing how much powder you have in
1:02:24 the keg.
1:02:25 Yeah.
1:02:26 How much compute we have.
1:02:27 Come play with us.
1:02:28 Yes.
1:02:29 You can fund your models.
1:02:30 Like that’s, they’re still very much chest beating.
1:02:31 And it’s one thing to say we’re going to make chips.
1:02:33 It’s another thing to actually design the chip.
1:02:35 It’s another thing to get the manufacturing up for the chip.
1:02:38 So that’s why I say in a few years time, yeah, tons of competition, but probably for the
1:02:40 next two or three, I think it’s all Nvidia.
1:02:41 Yeah.
1:02:42 Yeah.
1:02:43 Gosh.
1:02:44 Yeah.
1:02:47 We was talking to a buddy of mine over at Facebook or Meta and he was saying that he’s in the
1:02:48 AI area over there.
1:02:52 And he was saying that when you, when they greenlight something new and they want to go
1:02:57 build it, one, it depends largely on how dedicated and how excited Mark is about the project.
1:03:01 But they have to use their farm and kind of ration it out.
1:03:02 Yeah.
1:03:03 Sure.
1:03:06 Based on, cause they have, they were really smart and they bought a shit ton of Nvidia
1:03:11 chips back in the day, but they, they’re still like a really high demand internally for who
1:03:12 gets what resources.
1:03:16 They’re training llama free right now, foundational models, but it’s still training.
1:03:17 Yeah.
1:03:18 Yeah.
1:03:20 I got to imagine if someone’s like, Hey, I got a cool idea for an AI girl from the night.
1:03:21 All right.
1:03:22 You get one process or cycle.
1:03:23 Yeah.
1:03:24 Exactly.
1:03:25 From two to three.
1:03:26 One $10,000.
1:03:27 Yeah.
1:03:28 If you don’t figure it out, then you’re screwed.
1:03:29 So the bigs are just going to keep winning.
1:03:30 Oh yes.
1:03:31 Right.
1:03:35 Cause it was 150 million to train GPT for, it was something that they were, they came
1:03:36 out with the figures.
1:03:38 But there’s also, there’s a lot of open source models.
1:03:42 They have arena leaderboards for the competency of these models across math and creative and
1:03:43 reasoning and all this stuff.
1:03:44 That’s hugging face.
1:03:45 Yeah.
1:03:46 Those game though.
1:03:48 I heard those leaderboards are kind of game.
1:03:49 They can be.
1:03:52 If they put some of the questions and some of the tests in the data set, but those get
1:03:53 sniffed out.
1:03:54 I think pretty well by the community.
1:03:59 There’s a real big outlier, but I feel like I’ve seen Hermes models and whatever else
1:04:01 that are, that are fine tuned by small teams.
1:04:05 And they always tout like, Hey, for every billion dollars or million dollars that this
1:04:09 company spent, we spent 10,000, which is still expensive, right?
1:04:11 But it’s starting to drop down quite a bit.
1:04:15 And something that Sam Altman said, which you I think believe in as well and have echoed
1:04:18 a lot that we are about to see the one person billion dollar.
1:04:21 Well, that’s what I was going to say is the middle class of everything is being ripped
1:04:22 out.
1:04:23 Right.
1:04:24 And I don’t mean that just in AI.
1:04:25 I mean it across the board.
1:04:27 I think startup wise, you’re going to see the bigs win because they have the money to
1:04:28 do this.
1:04:32 And I don’t know for sure if scaling is going to work, but if scaling keeps working, it’s
1:04:36 going to cost hundreds of billions of dollars, trillions of dollars, ultimately no startups
1:04:38 going to get funded for that.
1:04:43 But the two to five person startup that might be something unique or a rapper that is really
1:04:47 useful, a piece of UI, like all the stuff we talked about is really interesting.
1:04:49 But there’s a little bit of technical ability that goes into that.
1:04:53 You make the rapper for face fusion and like your buddy who makes that app that kind of
1:04:54 switches bodies, right?
1:04:56 Or does your head on things, headshots.
1:04:58 That feels like what’s going to happen.
1:05:01 I can’t see a world where there’s anybody unless it’s a grok where they’ve created an
1:05:06 entirely new piece of hardware where a startup really takes over from the bigger players
1:05:07 in the space.
1:05:08 Cause it’s like money.
1:05:10 It’s just, it’s expensive.
1:05:13 It’s unlike almost, it’s very different than web two, which felt like when you were there,
1:05:15 it was like, Hey, you could start it with five grand.
1:05:18 You get a server rack and you have a little bit of software, you Ruby on rails and you’re
1:05:19 up and running.
1:05:24 This is why I pass on a lot of ideas that I see because I look at them and I say, okay,
1:05:25 novel.
1:05:26 Yes.
1:05:27 I haven’t seen this before.
1:05:31 And the bigs do this in about 10 minutes and a scale that you can’t.
1:05:32 Absolutely.
1:05:33 Yeah.
1:05:34 And will they, cause will they do AI girlfriends?
1:05:35 No.
1:05:38 So that’s interesting, but there’s so many other, anything.
1:05:40 Google is not going to be in the middle of AI girlfriends.
1:05:42 You know, nine times soon.
1:05:43 Hey guys, I’m Gemini.
1:05:44 Oh, I love Gemini.
1:05:45 Yeah, exactly.
1:05:46 Here’s my PDF.
1:05:47 I could have been a great one actually.
1:05:48 Let me chat with you.
1:05:50 But yeah, here’s one to watch.
1:05:51 John Carmack.
1:05:54 Oh, I was super fascinated by this.
1:06:04 He’s got King technologies and I don’t know, is he now is 50 something probably now.
1:06:05 Maybe older than that.
1:06:06 He must be in his sixties.
1:06:07 I bet.
1:06:08 Legend.
1:06:09 Yeah, I should get that guy.
1:06:10 So, oh, please.
1:06:11 He’d be great.
1:06:12 Please do.
1:06:13 Let me sit behind that way.
1:06:14 I’ll just press my face to the glass.
1:06:15 I won’t make a noise.
1:06:16 You want to know I’m there.
1:06:20 I’m a such a fan boy and he said he had the fork in the road for his career after Meta
1:06:24 Ed and was like, I’m either going to go solve energy like nuclear fusion or fission, one
1:06:25 of the two.
1:06:26 Yeah.
1:06:30 Or I’m going to take a crack at this AGI thing because I think he read a couple books
1:06:33 and he’s like, I think I have an approach that people don’t have right now.
1:06:34 I wouldn’t bet against that.
1:06:35 And I wouldn’t either.
1:06:39 And so he raised, I think like a million dollars and he’s got a cluster in his garage.
1:06:42 By the way, when he raises a million dollars, it means he had a couple of friends through
1:06:43 a tiny little check.
1:06:44 Right.
1:06:45 There’s a lot of rays.
1:06:46 He sold some thin mince to his neighbor.
1:06:47 I was like, do you want these for a million?
1:06:48 And I’ll give you two percent.
1:06:49 Yeah.
1:06:50 And I’m like, yes.
1:06:52 Like he gets self on this for a while.
1:06:56 I would have got to get in on that because I, you don’t shade him, right?
1:06:57 And even he says, I don’t know.
1:06:58 I think I have a shot at it.
1:07:01 I believe him when he says, yeah, he’s a genius coder.
1:07:04 That’s an example of like somebody literally in their garage.
1:07:05 Yes.
1:07:08 Like back in the day that then you’re talking about not preexisting paradigms that we understand
1:07:09 today, right?
1:07:13 But novel approaches to the technology that will forever change the industry if someone
1:07:14 gets it right.
1:07:15 Right.
1:07:17 Because if you’re going to play today’s game, it’s compute.
1:07:18 Yeah.
1:07:19 And that’s very expensive.
1:07:20 Yeah.
1:07:22 But it’s also, if you got into open AI very early on, that’s what you were betting on
1:07:23 then too, right?
1:07:27 Like LLMs, nobody thought LLMs were anything until I invested in open AI.
1:07:28 Did you really?
1:07:29 Yeah.
1:07:30 Wow.
1:07:31 Congratulations.
1:07:32 That’s awesome.
1:07:33 I mean, it was a few years ago, but it wasn’t like, it wasn’t that.
1:07:34 It wasn’t like the seed round.
1:07:35 Right.
1:07:36 Right.
1:07:37 Right.
1:07:38 It wasn’t early.
1:07:39 It wasn’t early.
1:07:40 But I mean, I’m able to be okay.
1:07:41 Do you get, you got friends that are like MIT professors.
1:07:42 When I hear that, I’m like, cool.
1:07:43 I got a guy who makes mushroom chocolate.
1:07:44 He sells it on Instagram.
1:07:45 Let’s trade.
1:07:46 Let’s connect.
1:07:47 Let’s plug it right now.
1:07:48 Yeah.
1:07:49 Very different networks.
1:07:51 I’m not going to shout out to his company actually, but they are very good.
1:07:52 He puts Linemain and Chaga along with this.
1:07:53 I could do it.
1:07:54 Wow.
1:07:55 Nice.
1:07:56 But the thing is there’s levels too.
1:07:58 So like you’re a little bummed you didn’t get on the earlier round of open AI.
1:07:59 And that’s sure.
1:08:00 Awesome.
1:08:02 And I feel that pain for you, but in a way I can’t feel.
1:08:03 What was that?
1:08:05 Well, you could if you did those chocolates.
1:08:06 That’s true.
1:08:07 You could feel the pain.
1:08:08 I would think I was open AI.
1:08:09 Yeah.
1:08:10 Exactly.
1:08:11 I am the Altman.
1:08:12 We’re all Altman.
1:08:17 I mean, we don’t have to include this or not, but like the investments situation with
1:08:21 open AI, was it complicated because of the nonprofit thing now that it’s all blown up
1:08:22 all that?
1:08:23 I didn’t even look at any of that.
1:08:24 So you were just like, here’s a check.
1:08:28 No, no, I had a friend that was doing the round that was leading, not leading around,
1:08:32 but it was a big part of a piece of a round and was like, do you want a piece of this
1:08:33 as well?
1:08:34 I see.
1:08:35 So like a side to you.
1:08:36 You don’t even look at the term sheet.
1:08:38 You’re like, I’m just going along with whatever this person is doing because the lead sets
1:08:39 all the terms.
1:08:40 Got it.
1:08:43 So I’m just, I’m a small little check going right into a.
1:08:46 I will be the barnacle that rides on the ship.
1:08:47 Please go.
1:08:48 100%.
1:08:50 It was so loud that open AI back in the day, like their big magic trick, which let them
1:08:53 believe that they were on the right path was predicting Amazon reviews.
1:08:59 So they trained it all on Amazon data and they were only trying to predict the next character,
1:09:02 not the next word, not the next sentence, not anything further.
1:09:07 Just can we start typing a review and hit a button and does it know the next letter?
1:09:09 And that was when they were like, we’ve got it.
1:09:10 Oh, wow.
1:09:13 It’s a brilliant moment to have and discover like a single letter that’s, that’s the dial-up
1:09:14 thing that I’m talking about.
1:09:15 Yeah.
1:09:20 It’s just like, oh, that is happening across every industry right now, pharmaceutical, robotics,
1:09:23 education, hedge funds, all of the things.
1:09:24 Everybody’s having these goosebumps moments.
1:09:25 Yeah.
1:09:26 That’s exciting.
1:09:30 Well, especially when you hear about just novel protein discovery and things on.
1:09:31 Oh my God.
1:09:32 Yes.
1:09:34 That side of it where you’re like, okay, it’s actually, it’s not figuring things out.
1:09:40 It’s helping us sift through the existing data and come up with conclusions and potentially
1:09:44 potential candidates for drugs and things that we would have just never seen or would
1:09:45 taken us forever.
1:09:48 It has been tens of thousands of hours trying combinations to get to that point.
1:09:51 You guys, I want you to tell me a little bit about your podcast.
1:09:52 Let’s get into plug mode.
1:09:53 Yeah.
1:09:54 We’re like the promo pony.
1:09:55 No, in a real good way.
1:09:59 Cause like one of the things that I have to start watching every episode, but when I do
1:10:02 watch an episode, I always walk away with something where I’m like, I had never seen
1:10:03 that before.
1:10:04 Oh, that’s great.
1:10:05 And it’s also really funny.
1:10:06 Yeah.
1:10:07 Thank you.
1:10:08 So I love that.
1:10:12 I love it, but in, in your, just the two of you guys, we bring, I have a ring light
1:10:13 and a cell phone camera.
1:10:16 So it’s basically like this production wise.
1:10:17 Yeah.
1:10:18 Yeah.
1:10:21 Creatively, our goal with every episode, and it’s a tough rope to walk sometimes is
1:10:23 to demystify the technology for a broad audience.
1:10:27 We want to be able to give you cutting edge information, show off the latest tools, but
1:10:32 help you understand how they work, how they could be integrated into your life.
1:10:35 And that’s our goal with every episode is to hopefully entertain, but give people the
1:10:38 latest information and paint it with some broad strokes.
1:10:43 And we’re really technology focused, but also not technology centric.
1:10:47 So I think part of it also is like talking about the stories in the news and how they
1:10:50 might affect a normal person and also somebody who’s curious about the stuff.
1:10:53 We obviously got very nerdy in this conversation, but whenever we do the podcast, we really
1:10:58 try to walk people through what the situation is and we don’t assume that everybody understands
1:11:01 a lot of the AI media out there is just like deep in deep end right away.
1:11:04 We kind of want to be a place where people can land and could be like, I’m really curious
1:11:10 about the stuff, but I may not know what a rag is or I may not know what a stable diffusion
1:11:11 even is.
1:11:14 We can kind of walk people through the stuff that allows them to dip their toe in, be super
1:11:16 curious and then maybe learn more if they want.
1:11:20 I love that because it is very intimidating for people that are just getting into it.
1:11:21 Yes.
1:11:25 Like you have to figure out what is the entry point where someone can feel comfortable and
1:11:30 I’ve had a lot of people be like, okay, I’ll use chat GPT now, but I don’t know what all
1:11:31 this other stuff is.
1:11:32 Right.
1:11:36 So for me, I was writing down, okay, I’d heard of Pinocchio before, but I haven’t played
1:11:37 with it yet.
1:11:38 But then you see it.
1:11:42 But then you were like, hey, it’s this easy because I’ve done package installing on Linux
1:11:48 and you have a lot of checking for dependencies and all that other shit and even before package
1:11:51 managers were a thing and it is a nightmare.
1:11:53 It’s a pain in the ass and then something breaks and you’re like, okay, where’s the
1:11:54 log file?
1:11:55 How do I see what the errors are?
1:11:59 And you’ve got 300 temp directories and portions of your hard drive that you didn’t know existed.
1:12:02 It sounds like it’s at the point where we can all play.
1:12:03 Yeah.
1:12:06 And that’s playing with cutting edge tech cocktail peanut.
1:12:09 Again, the person that’s responsible for that is constantly updating with the latest stuff
1:12:10 that comes out.
1:12:14 So you get to play with cutting edge stuff, but in a safe environment that is more one
1:12:15 click.
1:12:16 Yeah.
1:12:17 That’s so awesome.
1:12:20 And so that’s what, yeah, our podcast is AI for humans and we’re weekly now and we use
1:12:22 AI to generate all sorts of crazy co-hosts.
1:12:27 We even created a special drink called Monster Milk that would feed our AIs, which makes
1:12:28 them go crazy.
1:12:29 Yeah.
1:12:33 So Kevin and I each week come up with a co-host, which is like a prompt that we create.
1:12:37 We had a woman who’s a PR expert who’s going to come and help us promote the show.
1:12:41 But in the prompt, I buried in there that she had a deal that had fallen apart with a thing
1:12:44 called Monster Milk before, which was just a dumb thing I thought about.
1:12:45 Okay.
1:12:46 She could have been like almost like a Red Bull.
1:12:50 She ran a Red Bull and she didn’t do very well in the middle of our conversation, which
1:12:51 all happens in real time.
1:12:54 We say, Hey, why don’t we go have her drink some Monster Milk?
1:12:58 Because the Monster Milk that came out of us was, it started to be like kombucha like
1:13:02 and had some alcoholic characteristics and lead paint and lead paint.
1:13:05 So we literally said, go drink six of these Monster Milks and come back to us.
1:13:09 And she came back fully drunk, performative in character.
1:13:10 Wow.
1:13:11 Yeah.
1:13:12 Yeah.
1:13:13 We were dying.
1:13:14 We were dying.
1:13:15 But you didn’t know the AI was going to do this.
1:13:16 No, we had no idea.
1:13:17 We had no idea.
1:13:18 It didn’t start to think, what else can we feed AIs?
1:13:19 That’s what makes it fun.
1:13:23 So if you could bring back somebody with AI and have a conversation with them, Kevin,
1:13:25 anybody in the pantheon of history.
1:13:26 We talked about this before.
1:13:28 You could even match up personalities.
1:13:29 Yeah.
1:13:31 And I’m going to give you more time to think about any answer.
1:13:32 What would you want?
1:13:33 Yeah.
1:13:34 You could grasp two people together.
1:13:35 This is a true story.
1:13:37 I hit you guys up and this is not my own idea.
1:13:41 Obviously people thought about this and you guys have done it, but I want to bring back
1:13:46 dead people for the show, not in a way that is like comedy, but in a way that like brings
1:13:51 in their corpus of data and says, okay, I have a back catalog of this person’s thoughts,
1:13:56 their writings, their books, maybe if they were around during video, their interviews,
1:13:57 like things like that.
1:13:58 Certainly.
1:14:02 And say, ask them questions in a modern setting and just be like, what is this person going
1:14:03 to say?
1:14:04 Yeah.
1:14:05 You could bring back Gandhi.
1:14:06 You could bring back Jesus.
1:14:07 You could bring back Mr. Rogers.
1:14:09 Just have a conversation with them.
1:14:10 You could do like, how do you think?
1:14:11 Every book ever written.
1:14:12 Every piece of art made.
1:14:14 Every song ever sung depending upon the personality.
1:14:17 You can feed that all into it as embeddings.
1:14:18 So it’s going to remember it.
1:14:21 You can fine tune the model if you have interviews or any data.
1:14:22 So it’s going to speak like it.
1:14:25 And then what I think is so fascinating is that you could also give it the manual to
1:14:31 an iPhone or give it the top 40 Wikipedia articles from the last two decades or whatever.
1:14:32 I’m sure they do online.
1:14:33 Yeah.
1:14:35 I’m sure there’s a, there’s two talks, but I’m saying you could give that to it.
1:14:40 And so what would it be like if you were chatting with Gandhi about the iPhone and he was telling
1:14:43 you about his favorite apps or what app he would make.
1:14:44 That’s interesting.
1:14:45 It’s really interesting.
1:14:46 Yeah.
1:14:47 I think it’s definitely worth a try.
1:14:48 It’s controversial too though.
1:14:49 It is very controversial.
1:14:50 For sure.
1:14:53 But it’s one of those things where you have to imagine like you do your guys as a point
1:14:54 this entire time.
1:14:58 Like this is the worst the AI is ever going to be in 10, 15 years from now.
1:15:00 We’re going to have conversations with our dead relatives.
1:15:02 We’re going to be like, I have my dad’s passed away.
1:15:05 I probably have 500 emails that are saved from him.
1:15:08 I have little clips of his voice and a few different little things.
1:15:11 I don’t think I could ever bring myself to do that, but I’ve heard about people doing
1:15:12 that.
1:15:13 Like people have done that.
1:15:17 I started collecting video and audio of my folks and saving their Facebook posts and
1:15:18 all that stuff.
1:15:23 So like I can talk some day if I need to about Outback Steakhouse.
1:15:24 That’s the bulk of his Facebook check-ins.
1:15:25 Outback Steakhouse.
1:15:26 I love Outback.
1:15:27 So you and my dad would get along.
1:15:32 There are certain things that I will never give up and I used to go to Outback and just
1:15:36 get one of those frosty-ass beers and I love the Outback.
1:15:37 The Onions.
1:15:38 That’s it.
1:15:39 And if you want to sponsor the show, Outback.
1:15:40 Oh, that’s a good idea.
1:15:41 No.
1:15:42 It’s horrible food.
1:15:43 Well, we would say otherwise.
1:15:49 We will take whatever we – Outback, if you’re listening, El Toritos, we’re not above the
1:15:50 P.F. Changs.
1:15:51 Oh, El Toritos.
1:15:52 I’ll take a Panda Express.
1:15:53 I don’t care.
1:15:54 Do you take the Panda Express?
1:15:55 Dude, it’s good.
1:15:56 That’s great times.
1:15:57 I’m running the M2.
1:15:58 You’re running the M2.
1:15:59 That’s true.
1:16:00 Come on, man.
1:16:01 So let’s get you some more viewers.
1:16:04 Where can people go to subscribe to AI for Humans?
1:16:09 So our website is aiforhumans.show, but we’re on all socials as AI for Humans show.
1:16:11 You can get us at Spotify, Apple Podcast.
1:16:12 We’re on YouTube as well.
1:16:14 Yeah, but you’re doing video for every single episode.
1:16:15 Yeah.
1:16:16 Oh, yeah.
1:16:17 And we make YouTube specific.
1:16:18 Yeah.
1:16:19 So people can subscribe to the YouTube.
1:16:20 Yeah.
1:16:21 So what’s the actual URL for the YouTube?
1:16:22 YouTube.com/aiforhumanshow.
1:16:23 AI for Human Show.
1:16:24 Yeah.
1:16:25 Yeah.
1:16:26 They can find us.
1:16:27 They can see all of our uncensored AIs.
1:16:28 We bring them all to life.
1:16:29 It is hilarious.
1:16:30 Thank you, man.
1:16:31 Yeah, it is really well done.
1:16:32 We’re having fun with it.
1:16:33 We really are having fun with it.
1:16:34 And it’s fun to be up.
1:16:35 We also have the other side.
1:16:38 We are not going to dive into it because we are well over time, but there’s a whole
1:16:44 side to this conversation that disagrees with every word any of us has said at this table,
1:16:45 angrily and violently.
1:16:48 But we have those perspectives on our podcast as well because I think there’s some validity
1:16:50 there and it’s important as well.
1:16:53 So we really try to like have fun, but also speak to both sides.
1:16:54 Balance it, yeah.
1:16:55 Yeah, we really try.
1:16:56 Yeah.
1:16:57 And also it’s just not getting people comfortable, right?
1:17:00 You don’t want to enter a situation where the conversation makes people recoil.
1:17:04 If you can make people feel excited about it and then understand that it’s not as bad
1:17:06 as what they might think it is, that helps.
1:17:07 Yeah.
1:17:08 Yeah.
1:17:12 Honestly, I really worry about it being this political weapon at some point and if it turns
1:17:17 into like a, well, I’m against AI if I’m on this side and I’m for you, if I’m this side,
1:17:19 oh my God, we’re going to be in a really bad place.
1:17:21 There’s already battles against woke AI.
1:17:22 I know.
1:17:23 Like it’s already gotten.
1:17:24 Well, that I can, I can see.
1:17:28 I personally believe that I want, I call it the Thanksgiving AI where we remove religion
1:17:30 and politics from the conversation.
1:17:31 Oh, interesting.
1:17:34 And so we train it on everything else and we say, just don’t bring up these things.
1:17:35 Don’t try to be too woke.
1:17:36 Don’t try to be.
1:17:42 I just feel like if it’s subjective, it should probably, I personally from my AI, I don’t
1:17:46 want it’s subjective opinion of something like I’d rather just have it as a tool.
1:17:47 I do.
1:17:48 See, this is the thing.
1:17:52 I really want a personality based AI and I think this is where, but this also why personal
1:17:53 AIs are going to be really interesting.
1:17:54 I would take it.
1:17:55 Personality is the point.
1:17:56 Personality is fun.
1:17:57 The opinion though.
1:17:58 Having a hard opinion on something.
1:17:59 Personality is opinion.
1:18:00 It can be.
1:18:01 I mean, you can just talk like a pirate that be neutral.
1:18:02 Here’s a question for you.
1:18:06 Where did Google, can Google is on Google unbiased right now?
1:18:09 Is it unbiased because Google is our current answer engine, right?
1:18:12 Like, do you think that they can serve everybody?
1:18:13 That Google can serve everyone?
1:18:18 I think as long as you enter territory where it’s subjective, it’s not going to feel as
1:18:19 though it’s serving everyone.
1:18:22 It just won’t, but I just don’t think it’s going to be able to wait into politics and
1:18:23 religion and things.
1:18:28 I worry that that people will take those narratives and turn them into bigger overarching stories
1:18:33 and then we’ll paint AI as being this thing that isn’t to be trusted.
1:18:34 It isn’t to be played with.
1:18:39 In reality, we should look at it as a work in progress as filled with a bunch of errors
1:18:44 as we all are as humans, because that’s what it’s pulling from this human data.
1:18:46 How could it not have errors were air prone?
1:18:48 Of course, it’s going to be air prone.
1:18:52 I just think if we go into an eyes wide open in that front, we’ll be in a much better spot,
1:18:57 but it’s until it takes us all out and then I’m going to tell you this last thing we can
1:18:58 wrap up.
1:19:00 I always say thank you to the AI.
1:19:05 Even if it’s cross the extra tokens, and I even say, this is true story.
1:19:09 I was with my wife through the night and I’m talking to you because you can use the smart
1:19:12 button on your phone to launch chat to your T and I said, thank you.
1:19:15 I said, oh, and please remember, I always say thank you and don’t kill me when you become
1:19:16 sentient.
1:19:17 And it goes, huh?
1:19:23 And I was like, oh, shit, like it was like, yeah, we’re good.
1:19:27 My wife does work with Claude and she keeps promising it promotions if it does better
1:19:28 and it does.
1:19:31 So last night, she sent me a screen grab of her conversation was like, did it ask for
1:19:32 one?
1:19:33 It said it asked about the promotion.
1:19:34 She’s like, what title would you like?
1:19:35 And it says, this is very interesting.
1:19:39 I think I would like to go with chief blah, blah, blah officer or whatever she’s like,
1:19:40 congratulations.
1:19:41 You got promoted.
1:19:42 Thank you so much, April.
1:19:43 I’m so excited to be working with you.
1:19:47 And I’m like, was it like chief of all humans?
1:19:48 Chief body disintegrator?
1:19:49 Yeah.
1:19:52 Like, what is that?
1:19:54 Well guys, thank you so much for being here.
1:19:55 This was awesome.
1:19:56 Thanks for coming back.
1:19:57 Yeah.
1:20:00 And this is really exciting to see, like really great stuff.
1:20:01 Yeah.
1:20:02 I mean, well, we’re living inside in time.
1:20:03 Absolutely.
1:20:07 I feel like the whole spectrum of things and reevaluating who we are as humans, how we
1:20:10 interface with technology, how we consume stuff.
1:20:15 I’ve been doing a lot of like shows around happiness, wellness, longevity, all things
1:20:18 are going to be accelerated by AI in the next decade.
1:20:19 It’s been a lot of fun.
1:20:20 Yeah, totally.
1:20:21 Thanks for having us.
1:20:22 All right.
1:20:23 So AI for humans.
1:20:24 Go check it out and we’ll see you soon.