Tech Brew Ride Home - (TWTR SPC) The AI Moment With @pt, @miguelisolano & @mignano

Episode Date: October 22, 2022

Talking this week's big week for AI, with Parker Thompson (@pt) plus@miguelisolano&@mignano Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. This is the first time I think I'm doing the show, having just gone off a 30-minute Peloton ride, so I'm probably going to be super hypey. Oh, man.

I always threaten to do that, like go for a run before we do these, but what I usually do is take a nap and then... A nap and a gummy usually, right? Well, the gummies happen half an hour in, so it's not yet. Okay. Cool, all right, so we got Parker here. I swear, at one point I'm probably going to call Parker. I'm just going to totally forget.

By the way, Parker, when you get up here, is it like Parker S. Thompson is in like Hunter? Or is it just sort of like an accident that happened? I don't know. Parker, I don't know if you've, how much experience you have with Twitter spaces, but you will need to be on a mobile device. I thought you usually send people your prep. I guess I just assumed, you know, he's like an OG, so he knows all the things.

By the way, do you see the little magic wand down there now? We all have access to the soundboard. I'm sorry, everybody, I'm going to do this, so it's going to sound horrible. Oh, that's much of it. That's much quieter. Before, there was like screaming. Okay. There you go.

Okay. I see. Yeah, that's, yeah. All right. It's this little softer, though. Okay, great. Parker's up. Okay.

Sorry about that. I have not done this from my desktop. And I thought I would be prepared and have my nice microphone and the whole thing. I don't know. You know what? It's one of those things that I'm still really kind of annoyed at. And if Alon Gavis, and fire 75% of Twitter staff.

I don't know if we're ever going to get a good Twitter space experience in desktop. Sorry. Sorry to break your hope and dreams. Yes. Well, listen, let's not let's not. Yeah, let's not jinx it. Let's not jinx it. To rail ourselves on that one.

Okay. Well, welcome Parker. Brian, you want me just pick it off and we'll just dive in? Yeah. If you're good to go, Parker, you okay? Oh, yeah. Yeah, let's do it. Welcome everybody to the TechMeme Ride Home Experience for Thursday, October 20th.

It's been a couple weeks since we've been on. This is sort of a slow season, I think, for the show. But as ever, lots of stuff going on in the tech world. Brian, a couple weeks ago, maybe even longer, I think, prognosticated, which is the appropriate term here, reading the tea leaves, that there was a kind of vibe shift going on in the VC and investment space towards AI. and that AI essentially would eclipse or replace kind of the hot darling of the, I guess, the tech world last, like last year, which was crypto, NFTs, and that whole jam.

And so I did go ahead. Kind of do that. I want to frame it. First of all, Parker, please introduce yourself in whatever way you want. And then please let me frame how I want this to go down. Sure. For folks who don't know me, Parker Thompson, I speak. spend my time investing in startups. So I do some enterprise, um, enterprise fast investing

at Seed with a partner, a very traditional fund, small fund. And then I run a fund that is more like an index fund that sits on top of Angel List. So that's a little bit funkier where I get to look at all of Angel's deal flow. And then we try to invest in about about a thousand companies a year and about 40 or 50 funds. So that gives me the ability to kind look across the ecosystem and see what's happening at more of a bird's eye view. Yes, since you said that, I'm going to blow up your spot a bit because I'm going to frame it this way, which is, so we recorded that episode the second to last day of September. It came out October 1st, and I said, hey, folks in Web 3 and crypto, I am concerned that all

of your people that have been blowing up your spot on Twitter, the VCs and whatnot, are going to turn their eyes, like the eye of Sauron to AI. And then this week, I'm not saying I'm a genius, but it's become a meme. You're implying me, your genius. It's okay. You can do that. Okay. However, Parker, you just said, you have more insight into, you know, the investing space

than just about anybody that I know. I didn't say that out of just like, you know, sticking a finger in my mouth and holding it up to the wind. I started to see more AI pitches coming my way. So is this a meme or not? All of a sudden, the energy, I did the two big stories about the two big raises in AI this week. What are you seeing in terms of the energy around the AI space right now?

So what's funny is I think that there is definitely a lot of excitement because there's truly interesting stuff coming out. I mean, you see this, all these consumer products that are just phenomenal and fascinating. I have not seen that many AI startups raising money because I think the kind of things that are really fascinating are. are pretty rare, right? Like, this is like, I call these things spaceship technology, which is to say, like, very rarely you see a startup where they show you their product and you're like, holy crap, this is like you just took it off a spaceship.

It's so crazy, right? I feel actually like there was much more, there were many more startups calling themselves AI startups, maybe 2015 to 2017. Like there was a joke then where everybody was saying, oh, we're an AI startup. And what they really meant is, I don't know, we got TensorFlow, and we're going to get some data maybe one day,

and then we're going to do something cool with it. And I felt like that was a modifier people put on their startups then. So I don't feel like we've quite... I mean, it used to be like mobile, right? Like, oh, I'm doing a mobile thing, invest. I'm doing crypto, invest. Mobile was like... Yeah, mobile was a little bit earlier than that, right?

But you're totally right. I mean, there was a thing where everybody was like, I'm making a mobile startup. And then there was when we were all making startups for millennials, right? So you do get these waves. It's possible we'll see a big AI push again. But I feel like all the VCs kind of got that out of their system where it was superficially AI five or so years ago. Actually, can we can we start there?

Because I think this is actually like so important. And I feel like we could and should spend a lot of the rest of the show actually unpacking like that terminology. Like artificial intelligence. And I let me clear. I want to get the investor part of the. this out of the way. That's why I led with it because... No, but that's what I'm asking. So, like, artificial intelligence obviously has been around for decades now. We've been, I think, whittling away at something that feels, I think,

to Parker's point, like space magic, or I'll just coin the term, like, you know, from space, whatever, that's magical. But it's... Yeah, let me... Let's sort of offer this as a, you know, a explanation, right? So I believe it was about 2012 that we saw, some major breakthroughs, real breakthroughs in the technology, right? So what Google bot? Like TPUs and stuff, right? Well, Google bot deep mind, right? Yes.

And you sort of, you saw TensorFlow come out. And so it really actually felt like there was a meaningful technical breakthrough around 2012 or a set of technical breakthroughs and that became available to the general startup populace, right? So I think that's kind of you start there in the modern area, right? That's different than say what was happening in the 80s. and 90s and so on. It's sort of the intellectual successor, but there was a real fundamental set of technical breakthroughs that gave us more than just, you know, like, I remember building

software before that where it's like, yeah, we can kind of build like the Netflix algorithm or something, right, but that was about as good as you can get and now you have more interesting stuff. So I think that there was genuinely a bunch of people trying to figure out what to do with that technology, and that turned in the kind of a hype wave in the, you know, the years that followed. like I say, VCs kind of didn't understand the technology, got excited about it, and we were funding different stuff, right? I looked deeply at that, and kind of my aha moment personally was looking at it going, oh, this is just mass, right? And we all have the same mass. So the math is effectively open source to kind of, you know, steal a phrase there, right? So we're all using the same

math. Some people are better at it than others, right? And then there's data, right? And data is really what differentiates these things, right? So I think the tools were very primitive in 2013, 2014. I did a little bit of investing in tooling. I think the tools are actually quite a bit better now, but still, you can't take a Ruby on Rails engineer and build, you know, open AI, right? Like, the tools aren't that good. The tools are good enough that someone like me can take TensorFlow and build something that I couldn't have built maybe 10 years ago, right? So I don't know if that helps in terms of a framing, but I think if you're looking at startups today, the reality is what's really interesting is being built by PhDs and people

like me would be building things that really aren't that interesting. But if we could get a proprietary data set, maybe I can build a valuable startup if that makes sense, right? But I'm not going to show up with spaceship technology and just make something awesome by crawling the web. Well, okay, I do definitively want to get this away from the framing of investing in it. Sure, sure. Do you

feel like I'm going to bring it right back. I said like, you know, on the show this week that you know, it feels like these could be just games

and like toys in the same way that certain things in crypto haven't like panned out into real companies. But at the same time, what we've seen with these generative AI tools and platforms that have come out, like the it's it's the first time in maybe five or six years since like the chatbot era of excitement about investing in that sort of thing, that I've been like, okay, wow, this is really expanding and iterating faster than I was prepared for. So I'm not sure what my question is, but what is your sense of are we on the cusp of something, or are we in the middle of like, okay, this is a step change in terms of how AI could actually impact the real world?

My sense, and I'm going to caveat this by saying I don't think I'm an expert in this technology, right? So my sense is that this is actually stuff that people have been working on in parallel for the last, you know, three or four years. How old is open mind now, or open AI rather? At least five, I think, right? Yeah. So I think people have been working on these problems in parallel. I mean, I heard an anecdote about, you know, Google moving all of the AI people into the CEO's office, right? Like the CEO sat with the AI folks, right?

And that's been years in the making. So my sense actually is that there's a little bit like, I don't know if you know the story of the, you know, the four-minute mile, right? Everybody thought it was impossible and then somebody breaks it and then a bunch of people do it all at once, right? So my sense is more like everybody's been trying to figure out this technology. We've all got our stables of PhDs. And then somebody releases a product and somebody else goes, we could do that, right? Somebody else goes, we could do that. And so I think that you're sort of seeing this stuff become public, but this is stuff that's

just been happening in all of these larger organizations and to some extent outside of them, right? Like there are with the hugging face people and some of these people that are, you know, more independent doing the same thing. So I don't know. That's my sense, but that's, you know, I don't, I'm sort of saying that from the perspective of a, you know, a casual observer as opposed to an expert in the space. All right. You know, instead of dancing around it, let's get into what I want to talk to you about. So one of your tweets that actually led Chris and I to engage you for this episode is Grady Booch was talking about, how a lot of this AI stuff has been trained on the open web and other people's content and things like that. And I'm going to quote your tweet in response to that, and then this is going to get us right into it.

It can't be overstated how important the coming battle will be between free culture and rent seekers, REAI. We will either decide computers like human artists can synthesize cultural antecedents and produce things. we call new or that cash much must flow, that culture is now owned. Okay, the way into this is, do we know who owns this stuff? Do we know is this just sort of like other things that people have accused technology companies of just strip mining other people's kind of, like what do we know about, like, can you build a business? on this if this is based on other people's stuff.

And forget about art, there's also like the coding AI and things like that. So copyright and stuff like that. All right. Yeah, and copyright's quite complicated. It would be great to get a copyright lawyer into this discussion. And I say that as someone who wrote a master's thesis on fair use, right? And I know enough to know that I'm not an expert in this either, which I'm just going to keep saying through this conversation and then I'm going to tell you, you know, how it is.

We're definitely not experts either and we get the same thing all the time. Yeah. So, I mean, I think this is the interesting open question, right? You know, we've gone through this a couple times. Like the most recent one is I think people forget that there was a serious debate about whether, you know, social media companies that are quote unquote monetizing user data should pay users. And that kind of, you know, faded, right? But that happened, right?

And there's a world in which that could have gone the other way where we passed laws that said, hey, you're going to pay users to monetize them, right? You've got to cut them in on it. Well, in some ways, actually, there is like monetization that's happening. Creators are, you know, making some money. So there has been a little bit of a correction. Yeah, I mean, I think there's a distinction between creators and, you know, it's not like universal basic creator income. Like, that hasn't happened. Well, it's, it's a crappy income as well for most of these folks, right? Yeah. And, you know, and I, I would nitpick there and say, I think talking about the creator, the creator economy is a little bit misleading. I think it's more

more sort of useful to think about it as the attention economy, right? Because you don't monetize your creation. You monetize the attention that you garner for creating, right? Yeah. Just that, I think that's a better framing for thinking about where the value is generated there. Facebook is aggregating the attention and then monetizing it. My data is not that interesting, right? Like, it's not valuable to me. It's valuable to Facebook. Yeah, it's worth like seven bucks to Facebook or something. Yeah. So, I mean, we have.

this conversation, we had this conversation around, you know, the streaming days, right? Information wants to be free. We got MP3s and the sort of the publishing class won that, right? Artist didn't win that. Artist won't win now, right? So the real question is, where do we land on this? Do we land in a world where these companies that may or may not be creating giant pots of money, we actually don't know that yet either, are going to take some of that money and put it into a pool, and that pool is going to be captured by a representative of these artists that then, you know, give the artists some pennies on our dollar. Like, maybe that's where we end up. I mean, you could see a legislative route that goes that way, or you could see us not do that, right? You could

see a loss of their approach to it where these are not considered derivative works, but new creations. So we haven't had, as far as I'm aware, a single lawsuit about this. I'm sure, you know, three, two, one, it's going to happen to that. Yeah. So where would the lawsuit be? It would be because I am a sci-fi artist that the things that are being generated in Dali look a hell of a lot like my book covers and things like that. And then when I can prove that you trained your model on my art, then that makes sense. Is that what you're expecting to see?

Or are you expecting the litigation to come from, I don't know. Maybe we should leave what happens to artists that can't make a living being graphic designers if this stuff takes over. Well, there's litigation and there's policy and they're kind of different things, right? So what's going to happen on the technology side is, I mean, I think you saw this person come out and say, hey, co-pilots stole my code, right? Very clear example. It's like, hey, this code looks like my code. That person has a legal right to sue, right? And I think that the tools are immature.

So the tools are designed today in a way where if you want to, quote, unquote, attack them in that way, you can trick the tools into doing this. I think the tools will get more sophisticated and you won't be able to get them to spit them. out something that looks exactly like a specific instantiation of a work, right? But now you can. That's pretty straightforward, right? You as someone who holds a copyright can sue me, someone who infringed your copyright, exactly, you know, who gets paid how much or whatever is. But who's being sued? Is it the, I mean, this is a lot of stuff? Yeah, I'm not, I'm not sure. I think you could probably actually sue both parties, right? Like, whether the courts would say that the tool

maker is, you know, liable or not. I'm not sure that's an interesting legal question. I think the more interesting thing to me and what that tweet was about was more what you saw in the early 2000s, right, was you saw publishers, record labels, rolling, you know, Metallica drummers in front of Congress going like, they're stealing our stuff. You guys got to do something about this, And so what I'm actually personally not that worried about the idea of individual copyright owners suing companies or suing individuals for infringing their works. I think like that's just always happened and it's fine. That's not that's not a problem if you're thinking about what broadly about human progress, right? I think if you're worried about, you know, how do we structure public policy to enable human progress?

what is more concerning is the idea that you might have people going before Congress and saying, hey, you know those big tech companies that we all hate, let's just go and take their money and quote unquote give it to artists, right? Because then you end up with a policy situation where some of these tools then become illegal, right? And that is where you kind of... And some companies have tried that, you know, with Google scraping the web, as they would say or whenever. I mean, none of us here are lawyers, but that hasn't ever been solved, right? Like, basically, you know, the whole reason that Google got started is because the web became the greatest, like, a giant data set that anyone could ever imagine. And people in the late 90s

in academia were like, you know, mathematicians were like, oh, my God, this is amazing. Like, have we ever settled the fact that these open data sets are out there? Yeah, look, we have robots. We solved this problem. This problem was solved in the early 90s. If you don't want your shit in the search engine, make a robots. Well, but a lot of the stuff that is actually being used in the public crawl and to train these models didn't, like there wasn't sort of awareness of a robots.

combe, per se, on some of those things that are being hoovered up in that process. They want their take and they want to get it too, right? Oh, 100%. 100%. I'm not saying people are rational. Like, please. Yeah. So certainly we can go back and, you know, iterate on robots. But I don't actually think that's what anybody wants, right?

Like if you made a better real estate tax, no one wants that, right? What they want is the giant pot of money that they're not, they don't have access to, right? So wait. So I think we're having two, maybe three simultaneous kind of, you know, conversation of threats here. One is this question of money and who gets paid for essentially what are direct. kind of output or work that can be generated as a result of these massive data sets being hoovered up at a scale that previously wasn't, I have a hard time imagining, contemplated by a copyright regime, you know, historically.

You know, previously it was like, I wrote this thing, this other guy's plagiarizing me, you know, I should get paid, you know, because I put the original work in, therefore my creative work is protected. Now we're doing things, you know, at hundreds of millions of, you know, scale. and we're using those as almost a heuristic or a just mathematical equation to generate more stuff that is similar to the things that came before. And so it's not any one of those particular contributions per se. It's kind of a soup that is generated from those creative works that allows us to

create the models like GPD3, etc., to then produce what ostensibly could be considered new works, new creative works. And then there is this question of, well, who gets paid in this model and who should get paid? and then also who gets to monetize it. And I think that creates a very interesting question because it's very much like an Aurora Boris in the sense that you're talking both about the, what did you call it, the attention economy.

I was thinking about the passion economy. Yeah, let me suggest that I actually think, who gets paid is a more important question than you're really talking about, because that's not the end of this, right? Because if people have to get paid on something, right? Then there's a bunch of stuff on the internet that can't just can't exist. Like I was like to think about like, it says something about our public policy that we could have a hackathon over a weekend and make a much better streaming music client than exists today that you can get on your phone.

Right. And the reason for that is because if we're not constrained by the law, we can do all sorts of cool shit that you just can't effectively do because of the fact. that certain people need to get paid in a certain way. So I think the public policy around... Sorry, but I think that's kind of an interesting and a good example, because are you talking about that media, you know, the music, that would go into that streaming service, or are you talking about the UI and the interaction that lives around it?

Because in some ways... For example, you have a right. If we had a radio station, we can play anything we want, right? If you have a streaming service, you actually have to go and do individual licensing deals for all the music, right? So I can't get my Garth Brooks on Spotify and I'm very angry about it, right? We just make an app that has Garth Brooks and the Beatles and Taylor Swift and whatever, right? The public policy here matters in terms of what kind of art can be created. Another example would be sampled, right? Like sampling was, we were sort of discussing privately this concept

of a platform shift, right? Like sampling for a very brief period of time was just a free-for-all and anybody could do anything, and the Beastie Boys made some amazing albums. And then that became illegal, and you can't do that anymore, right? So there's a lot of art that can't be created in the first place because of how people get paid on these things, because of the public policy, right? So I think it actually, who gets paid matters a lot more to what gets created than I think maybe your original frame was suggesting. Well, I guess what I'm trying to piece apart, you know, is the degree to which things are generated and that are able to be generated from all the things that exist. For example, you can't copyright a musical note, but you can copyright perhaps a sequence of musical notes. And then that sequence will show up in many, many songs, but it's not like the people who put, you know, five notes together the very first time are going to get royalties forever and ever.

That's not really how it works. It has to be sort of a substantive original work. And that's really changed over time as well, right? I mean, think about the 12 bar blues, right? The 12 bar blues is an entire genre that is exactly the same song, right? I don't think people really understand the extent to which the way that we think about intellectual property has radically changed in the last 40 years. It is fundamentally changed, right? I would highly recommend, for example, the Bob Spitz biography on the Beatles.

there are entire sections about the Beatles just ripping off people wholesale, right? We don't think of the Beatles as plagiarists, but they were. Culturally, that was much more acceptable than it is today. Culture has really shifted on this, and I mean, you can all, you know, shake our fist at Disney or whatever, right? But this is why I think we're at such an interesting time, because you're talking about, like, are these algorithms creating new things or are they ripping people off, right? we haven't thought about this at all. It's an exciting time because we haven't accepted, we haven't sort of converged on a common mental model of this process, right?

So I think this is getting to something interesting, right? Because part of this conversation is a bit about who gets paid. And then there's a question about sort of fight clubby, like over what time horizon, you know? And so who gets paid sort of in the short term and immediate term based on, let's say, the model of the last, 40 years versus the VC model where you're thinking about, you know, seven to 10 years, how is the marketplace being structured differently? And so part of the premise of this conversation is looking back to a year ago and thinking about how the economy and money was going to be rewritten by crypto and NFTs. And then now we're either in the build phase or that kind of like

fizzled out. We're not quite sure. But now we're sort of moving over into this other world of AI and generativity and the fact that there's so much compute power that's available and needs to be applied towards something that generative models can be applied to work, to creativity, to art. And I think what I'm very interested in is whether there is a kind of, and not all VCs are, you know, as sort of insightful, I suppose, but thinking about a future model in which work and creative work and creativity and generativity is thought very differently because there's a generational shift. I guess like, is that necessary for these investments to us? play out or will we somehow inhibit the next generation of creativity that is brought on by

these new technologies? Because the stuff I'm seeing is mind-blowing. The types of things that people are able to create that they've never been able to create before, like it does feel like there is something of a Cambrian explosion happening now in that world that I don't think is going to go, you know, the genie's not going back in the bottle, right? Yeah, well, we'll see. I mean, look, if you put on your VC hat, right? Like, it's incredibly rational. to just make this bet, not knowing the answer to the question, how is the culture going to react? Because you just roll the dice and if it goes to zero, that's fine. You did that 20 times. So it's incredibly rational for VCs can fund these things in an environment of extreme uncertainty, right?

But the answer to the question, do these things become big businesses? There are multiple things that need to happen for these to be big businesses, right? I think the cultural conversation, like what is the metaphor that we collectively choose to adopt for this technology that then informs the policy, those will become the ground rules that then these businesses have to operate under, right? Like YouTube is a good example, right? I remember where I was standing when I heard that Google bought YouTube, right? I thought it was just crazy that they bought this infringement machine, right? And it obviously wasn't.

It made a lot of sense. But that was an environment of extreme uncertainty. And then we moved towards a set of ground rules. right and that worked out quite well I think we are in an environment of uncertainty if we decide for example that these algorithms are not capable of unique production that they are just ripping people off as you said right then we're probably going to move towards a world where we steer a lot of that money to creators we're very hostile towards the technology and that's an environment

It looks a lot more like, you know, investing in music. Like very few people were willing to invest in music, and there's been almost no venture-backed music startups, right? Spotify is the notable exception, right? RDO, there were a few others. But effectively, music after Napster was a digital wasteland, right? And the reason it was, or a venture wasteland, and the reason it was a wasteland was because, as a culture,

the Lars Oldricks of the world won. And the technologist lost, right? So I'm very interested in this question of, you know, how are we going to sort of figure out what the right metaphor for this is? Brian, did you want to weigh in here? Yeah, I'm going to take it a little sideways for a second. In the sense that, and I'm coming back to your tweet thread, that, because Chris and I have talked about this, Like you said, you used Dolly yesterday to generate a base image of a thing and a style, and then we, in quotes, iterated.

I think that's an important thing to think about, too, because this is a tool like anything from Adobe, like a word processor. So at what point, we're talking about the tool was formed based off of, in theory, other people's creations. But then in actual functionality, it is a tool, and we can get into this too in terms of whose jobs it's going to be taken away. But you have to be intelligent to use the tool the right way to get something quality out of it, right? And so in the same way that I'm no good at Photoshop, but other people are. So to what degree are we talking about once it becomes a tool, then the creativity is from the inputs of a skilled person using the tool. Yeah, no, that makes sense.

And I failed to make something that was useful because I'm still not good at Dolly, right? So I do think that's a good example, right? These tools, Dolly has, for folks that don't know, you can create an image and then you can iterate on it. You can draw boundaries and like expanded. You can erase things. And so the tools are fairly primitive still, but it will turn into Photoshop. It's going to turn into something quite interesting. And I agree there are getting interesting.

And like Microsoft just announced this week that they have a tool or was it last week, whatever. That's like being built into Microsoft designer where they're actually bringing this stuff in. And Microsoft put another what, like $50 million, $500 million in Open AI. I mean, clearly this is the thing that I'm seeing. and, you know, Brian, do your prognostication, like a bunch of money is going into these startups. And I think what I'm trying to understand with, you know, your clairvoyance is what is motivating that? Is it just because, you know, we're in kind of a dry season and there's not a lot of else, you know, to invest in? And there's a lot of hype and interest and progress.

I don't think it's about a dry season. Yeah, Parker, go first because I have thoughts on this. But Parker, you can. No, this is highly rational, right? This is genuinely, genuinely a generational set of technologies. And it's an exciting time because, like, you forget looking back the sort of the fog of war, right? We're at a moment where just we're going to look at this moment and it's going to be going to, like, who was the MySpace, who was the Facebook, whatever, right? So these companies all have, like Microsoft has massive amounts of money on the balance sheet, the great cash flow.

totally it's incredibly rational for them it's incredibly rational for VCs to make a bet that could be a thousand acts but might be a zero right like we should all be making these bets as it when we put on our venture capital hats so I think that no one knows how it's going to turn out but it doesn't matter this is just rational behavior it's like so exciting as technologists right let me let me let me ask you something about this though one thing that I want to understand and I know this is a little bit speculative but I think it also is rational that there are a number of things that are happening with greater efficiencies with regards to just CPUs,

GPUs. I mean, a lot of the GPUs sort of power was going to mine crypto. And now that we're moving from proof of work to proof of stake, I think. Yeah. Now you've got all these sort of like dormant data centers. And I know this sounds a little crazy, but we are moving to a world where there's incredible efficiencies being created by these TPUs and whatever the sort of AI-focused chips are. And so my question is like with Microsoft,

with Apple, with Google, they are going to continue to develop this compute power. And so they want to keep having kind of like this carrot out ahead of their horses to keep them running faster and faster. And so this is a perfect application of that latent investment that they've already made in those data centers and systems. I don't think that's quite right. I mean, I think for one, when you really start getting in, when we figure out what the use cases are, we will build custom chips that only are useful for those use cases, right? And you already see, like, Google has built custom AI chips. There are startups that are building custom chips.

Well, that's new. Like, that was not something being done. That's not new. People have been doing this for years. How many years? I would say on the order of five-ish years. Yeah, that's not years. I mean, we're going back to, like, open AI and deep minds.

Not decades, but years. Yeah. Yeah. So the people are building custom chips for certain applications. I think Google builds a lot of custom hardware. That's a little bit different than like the GPUs. I would actually be interested. I don't know to answer this question.

There are cryptocurrencies that were mining on generic GPUs and cryptocurrencies that are mining on basics that are only relevant to those cryptocurrencies. I'm not sure how much capacity has been, quote, unquote, freed up by the crash in cryptocurrencies. I would believe it if somebody told me there was a lot who was an expert, and I would believe it if somebody who was an expert told me that there was almost none, right? You might get both answers, yeah. Oh, well, no, I mean, I'm just saying, like, if we ask an expert, we might hear either answer, right?

So I don't think it's that, I don't think it's that, like, you can't build a perpetual motion machine, right? You can't just say, we're going to invent hardware, and then we're going to invent software that uses that hardware because that's efficient. you actually need to be making applications that people want to use, right? So I think the more interesting question is, like, are these applications A, that people want to use? I think the answer is probably yes, but then B, it's like, well, okay, but how do we build businesses out of them? And C is how do we build defensible businesses out of them, right?

So these are more like, again, these are like VC had questions, so we don't necessarily have to go down this route. But those are questions that I think are relevant to VCs. Yeah. before Chris and the back channel, yes, please bring them up. Before we do, real quick, let me frame one more thing, and then we should open this up to the audience, because I think there's a bunch of people in the audience

that are interested in this thing. Can we also throw down the marker for the concept of who loses here and is it the artist? Is it the graphic designers? Chris and I have already talked about this a couple of times, and people have framed it. Whereas, like, again, and I even said that the last time I spoke, where it's like you have to be good at it

in the same way that you have to be good at using Photoshop and things like that. So it's almost like going back to like the original licklider paper of like maybe for the interim period, for the foreseeable future, we have to have a priesthood of people that know how to, you know, horse whisper the machines and things like that. But I'm just curious, real quick, Parker, what are your thoughts in terms of if you're not only a graphic designer or an artist, but I don't know, a software engineer? What do you think about if these tools really are on the cusp of this revolution? Would you be concerned for your job, basically?

Put it starkly. I don't think so. Again, maybe I'm just not. Maybe I just haven't thought enough about this. I think that creative people will be able to find things to do. I think that if I had to pick a loser, this is maybe a lame answer, but like occasionally I go on Fiverr and like pay somebody to Photoshop some stuff because I'm not good at Photoshop. maybe I can now go to Dolly or something for that. And, you know, but I'm still when I go to.

I mean, that's definitely, that's definitely coming. I mean, Canvas is going to be completely. Yeah. When I go to build a web application, though, like when I'm building a startup, I'm going to hire a human and they're going to use these tools. There was literally a product, launched on product on today that, you know, you put in sort of where you want your business to be and like a term and it generates a whole page for you. And it sucks and it's bad. But it's, it's directionally where people are thinking. I just can't imagine that it's not going to happen.

Anyways, I don't think I have a good answer. Ask me again in six months and maybe I'll have a good answer for you. Yeah. This conversation is difficult because there's a lot of things I think that are happening. And I think we're also seeing a lot of different things and making sense of it in different ways. Perhaps because of the way that we relate to those different skills or talents. So, for example, as someone who kind of came up doing web design, seeing a machine being able to do those things. Or, for example, seeing there's a tool called diagram, which works in the world of Figma,

and you apply GPD3 and open AI technologies to it. And now suddenly you have applications that are being designed that are as, you know, I don't want to say like as good as anything else, but now we have two million examples to draw from that it's not that hard to design a relatively reasonable interface for software that used to take, you know, months and super dedicated, super talented individuals to do. So the fact that we can just ingest these patterns and then spit out kind of pretty good, you know, marginal copies, if not improvements is interesting. I think the question that goes to how generative can the stuff become and can it actually

get better at doing these things than we can. Like you said, at the very beginning, I think one of your core insights is that the internet turns everything that humans used to do into math. And computers are always going to, by definition, literally, their name is going to out-compute us. So if they can get better and faster and more capable, doing those things they will. And then the question is, what do we sort of point that fire hose at? And then what happens to the human capital and to culture as a result of an interaction with such a complex system? Well, let me give you an interesting anecdote. I think it's an interesting anecdote.

Sure. So we started investing the Access Fund at Angel's in 2015. And our thesis was, hey, we could use a bunch of data to make decisions at scale that would give us really high-performing venture portfolios, right? And it worked, right? We've been doing it for about seven years now, almost eight years now, and it's phenomenal, right? Sorry, are you saying these are investment decisions using AI? Yeah.

Well, no, we're not, so we're using the data, but then I am the computer, right? So I go and I make decisions, right? But Angel is built out a data science team, maybe 2018 or so, and I'd have just a real funny conversation at like a Christmas party after a couple drinks with the guy who runs a team where I'm like, no, it's like, you guys should put me out of business, right? This is great. Let's just make an algorithm that can do the job. And the guy kind of looked at me awkwardly like, wait, you know what I'm trying to do? I'm like, no, I'll go sit on a beach. I'll go do something else. Right? And they spent a number of years trying to figure this out,

and they actually built a quantitative fund, right? So we're about a 30% a year fund. Their fund's probably about a 15 to 17% of year fund. And the problem with what they're doing, they sort of begrudgingly, like we admit you guys have some alpha. And I said, yeah, well, we have alpha. And there's a couple of reasons we can beat the code, right? One is there's a bunch of data that's just not structured, right? And these machines cannot work well on unstructured data in a lot of cases. And the other thing is like if you look at the current venture market over the last six months, the market is doing things that it has never done in the data set that we have, right? Or another example would be new markets emerge, right? New problems emerge. And we don't

have data to backtest against, right? So for relatively constrained problems, for relatively structured data sets, there's not going to be a lot. Humans are not going to out-compute algorithms, right? I think what ends up happening, though, is I've taken all the work that this data science team does, right? So our historical results are about 30%. Now I've got a bunch of data that I didn't have because the machine learning is running there and giving me these interesting insights and we built some extra tools on top. So maybe I'm a 35 or 40% fun now. So I'm leveling up my game, right? So I think we are going to be able to take these products and simply change the nature of our work, but humans are in many cases, maybe not all cases, but in many cases,

going to be able to add value that's different on top of that in ways, relative to what we were doing a few years ago. Hold on. Let me just clarify it real quick, real quick. When he says 15 versus 35 or whatever, you're saying that your returns are basically double what the bots are. When you say 30%, like an IRA or something like that in terms of. Yeah, that's what we're looking at.

But what you just said is you're standing on sort of the bots as well. You're using that as an input that you're using to achieve your thing. Yeah, a good, I don't know if you know this, but like if you look at chess, right? So the best computers can be now beat the best humans, right? But the best humans can take code and beat the best computers, right? So human plus computer beats computers. Unless the humans are cheating, you know. Yeah, it's, yeah.

No, but that's the thing. It's like you take a human and some anal beads, and then you can go and beat the computer, you know? Yeah, if the computer used anal beads, it'd be totally old. Yeah, we'd be scared. No, the two computers can't be the human plus a computer, right? So there's something interesting there, and I think that's sort of the, that's the world that we're going to figure out.

The thing that I wanted to point out, though, and I think what you're saying is super relevant and valuable, which is that, you know, it's a little bit like, God, what's the baseball thing? Moneyball? Thank you, yes, moneyball. So in certain cases, as you say, when you have kind of a fixed game, like chess, like baseball, where there's a set of known rules,

you can really optimize to a certain point. And then there's like a breakthrough thing that happened. Like you said like the four minute mile kind of thing. I guess what I'm wondering about, and the nature of this conversation is trying to sort of understand and see if there are adjacencies or things to learn from what's going on in some other fields, like investing with the robo advisors and stuff like that,

where there is historical data to look at, but of course it's always hard to predict complex systems going forward. So if they've never seen market dynamics like are currently going on, then of course they wouldn't be able to predict them and they're going to act in a way that is unpredictable and perhaps actually very, very poorly. My, I guess my question thought or insight here is thinking about the aesthetic arts

and the amount of aesthetic content that has been generated to please the eyeballs of humans. And that our ability to synthesize and generate new forms from what has been done before, or similar to what has been done before, actually kind of will be enormous and also perfectly fine. And we don't actually need humans to be remixing that culture. I'm not saying that it's the end of culture or the end of art or the end of new creativity, but that we will be able to create so many different things that will please the Pinterest

algorithm that we really don't need to worry about photographers going out. and taking amazing, you know, staged photos and whatnot because we can actually generate them. And I'm going to take this opportunity also to bring up Miguel, who is working in this space. I think he can at least provide a little bit of, well, first Miguel, come up and introduce yourself. Tell us a little bit about what you're working on and then tell us where we have some of this conversation off based on your lived experience. Hi, Chris. Hi, everyone. Yeah, we just actually announced our, yes, we're done. we're building. We're building

AI computer infrastructure for AI. And I think there's a brother and usually under underestimated aspect about this whole thing is that

the Bt4 is not going to be built. I'm sorry, say it again. Clarify that. The next version, weird, this is as good as it is. Okay.

What is this good? AI is this good at this? Yeah, it's not going to get better unless there's, you know, some fundamental breakthrough. Because, Look, AI is as computationally expensive as crypto. One way to understand stability AI's raise is that, yeah, the raise $100 million. Hold on. Let me let me pause.

You're bringing you something up, which we haven't mentioned yet, which is the stability diffusion or stability AI, which is actually an open source project. You can download the source code on GitHub has just raised $50 million. And so this is one of the things, along with the investment from Microsoft. Oh, $100 million. Okay. Yeah, a lot of money. Yeah. Yeah. And look, and a lot of that is going to go to compute bills. For compute bills, yes.

Yeah. AI is a rare case of a, how do you say, it's a capital-intensive software industry. So you have to think of it in terms of hardware and hardware level expensive. Yeah. And the same thing happens with Jaffer. You know, the, you know, the, you know, the counterpart of the smartness of AI is that it requires an exorbitant amount of computers. and we tips are not going to get better because the the more so is that computers are in as they're small and as fast as they can possibly get so I just want to temper some expectations in that we're used to things getting better and better with time this is not going to happen with AI unless well unless the company that we're working on our but but you know it's it's it's

it's a I don't think people realize how expensive this is okay so wait well I want to pause you though you you said something interesting and I pinned your tweet you said that she GPD4 will never be built. And first, that requires us to define why GPD3 is such an amazing achievement. And, you know, at least in, you know, 2022, we may think or believe or be led to believe that GPD3 actually may be, you know, the most, the largest language model that will ever need to do all of our things forever because building GPD4 at a similar, you know, rate of improvement, I mean, essentially would require like quantum computing or more money than

the human race actually has produced or something. Yeah. Yeah. Yeah, we actually did an estimate that for coming to Get America at NYU in the next coming days. It would take 149 years and $350 billion in expense to build something that would at least reach expert human accuracy. Right? So when people talk about, okay, it's going to be the feature of work, it's going to fully replace programmers. You know, it's hard to think what the scale are intended. And this is a lower amount. It could get even worse.

So, you know, it might be as good as you get. And this is a dangerous, no, people should temper expectations about what these things can do as a result. Wait, wait. So when you say temper expectations, you're saying that like, are we at the height of what we can do with AI? Or do you think we're going to keep getting really? Yeah, yeah, because again, the ultimate limit in fact is compute. And again, like, okay, just, you know, just thinking of a vision, for instance, right? You can have, look at the failure of slow driving cars.

Solves driving cars are not going to get better because they're already at the limit of what computer, how good computers can get. And this might be a controversial thing that, you know, in the chip community, it's well-known, right? So no amount of innovation will get away from the fundamental physical limits that we've reached. So we have to have fundamental new rates of computing, which is what we're working on. But, you know, just aside from what we're working on, the industry right now is, is, you know, it's facing a similar problem to crypto, right? That you can start, you start, you start needing the same amount of electricity that the entire country was trawlers off.

That is the limit that we're going to get if people really want to scale AI. Open AI, for instance, has a massive problem, which is that they have thrown on a hundred million dollars to computer just to serve this few users. They cannot just 10x their users because you record a billion dollars to the computer, which not even Microsoft could afford. Right? So it's very important that people realize that AI is not magic out of anywhere. is because it requires, look, you know, the minimum, it requires about the same equivalent of 100 billion,

or sort of a hundred trillion operations per second, which is roughly the number of operations that a human brain is doing right now. So if you want to get to human level accuracy, you have to have an exorbitant and I've computed, I don't think, you know, the world is not, it's not right for it. Okay, so hold on. So you're building the space. You've just announced yesterday what you're working on. One, why are you doing this? And two, who will be your? your customers, and then three, what are they going to do with the thing that you're ostensibly going to build? Sure. Yeah, we're playing this because we want to make this dream happen.

You know, fully general AI will be the most impactful measure of humanity will ever build, probably the last invention that we'll ever build. But, you know, we saw this massive compute problem and we tackle that in a fundamental way. So, you know, if this hype is, is what it is, is what it is. We will be the main compute provider for the entire world, is what we're working on. Okay. All right. I'm going to bring PT back. See, I did it, Parker back. And just get, you know, kind of your reaction to this. Because here we have, you know, founder. And Miguel, have you raised money to fund this? We have. We have to. Yeah. But yeah, we have. So can I ask you, Miguel, like, so one of the problems that I understand to exist in AI, and I think that's what you're speaking about is the approaches that we're taking are incredibly energy.

intensive relative to, for example, how children learn, right? So what you're saying is that you're trying to take fundamentally different approaches that are going to change the amount of energy we need to produce these kinds of results. Is that way you're saying your start? Well, energy is the same way as compute, right? Or cost. Sure. Yeah.

Yeah. Yeah. It's what a difference about, you know, doing mathematical operations in a different way. So there's, you know, it's essentially an algorithmic approach. It's a smarter way of doing the same fundamental operations because all, yeah, there's something that you were talking about earlier that, you know, the, the, the, the, the numerical operations that that, you know, the bottleneck to AI compute as we know it today are essentially matrix multiplication operations, just multiple applications

and solutions. And, you know, our background comes from mathematics and algebra geometry and numerical computing where, you know, this problem is faced a lot where, you know, you have a to, for instance, simulate a galaxy or about 100 million stars. You know, it takes that goes with a square of the amount of operations. So it would take, you know, even all those computers in the world, 10 times come by, it would not suffice to even do one single step in a galaxy, right? So, so there are smarter ways of things in a number of computation,

and that's what that we're bringing, but that's something that the industry hasn't done at all. You know, the people assume that, you know, that the hardware is what it is, and we just scale harder, but we can't scale harder anymore. And, you know, it's, and you know, we again, outside of our technology, as the interest is right now, there's no way that we can get to general AI. Yeah, I don't even think I mean, a non-goal for me, it short term would be, you know, generally I right? Like we're, we're, I don't, I wasn't even thinking that we're going to get there, right? Or that we're on the precipice.

It seems like when it really shows for the first time that, right? And we have now with Palm, um, the Google source model, 500 billion parameters, we have super human, sorry, approximating medium human performance and expert human performance in some tasks. So for instance, like stuff like stuff like feels really human, like, you know, detecting parity, right, or sarcasm, or even programming using copayage, right? So we're beginning, you know, like any, I think any AI researcher would have thought that it was even feasible before GPT3. DPDC changed everyone's expectations. So yes, I think, you know, it's a matter of scale to get to general And you know, within the next five to ten years, we'll have machines picking up your phone and machines editing other machines, machines when you're taxes, machines, you know, you'll have the bigger with machines. That science fiction world is, you know, the hard part is to get what you can do.

Right. But then it's a matter of scale. Okay. So Parker, Parker, I want to hear your point. Oh, I mean, no, I don't really, I'm just curious. I'm fascinated to have an expert here to ask questions of, right? So I'm curious, so I wasn't thinking about general AI, right? I'm just assuming for the sake of this conversation that like we're not talking about that at all. I mean, you could talk about, you know, the crazy people in Google who think the computers are sentient or whatever, but that's, I just don't believe that to be correct, right? I'm curious to ask a question about what you're saying here. When you look at something like Dolly, my mental model is that the, the computer intelligence, of work is primarily about generating the models on the front end and then the

output is relatively scalable and I heard you say something I want to make sure I didn't misunderstand so can you do you have a can you make up a number for when I go and I create an image on Dolly what does that cost them do you know how many dollars that cost well it would take for them actually would take them about 200 billion operations running on a hundred thousand dollar supercomputer and you take them about but out I can tell you for hour it takes about three hours per hour so and that would serve just yeah I mean a single image would take cents all for that particular operation okay no that's

helpful to understand because as we think about what do we do with these things and what are the I mean we just came through the era of holy shit my Uber is expensive now because VCs aren't paying for it right yeah as we think about this market, it's helpful to think about, well, what is this going to look like when I'm bank for it. Right? And one way to understand Open AI's race is just like stability AIs is that their compute bill is enormous. GPD3 would happen because they got a hundred million dollar, um, uh, essentially, you know,

a credit from from Microsoft in exchange for a license for you p. But that's running on one of what was then one of the largest computers in the world. Uh, and that thing requires GPUs, which is a new, it's a specialized kind of computer that is. Okay. I'm going to pause you on that. I really appreciate you coming up and elucidating this. Yeah, that's really interesting. Congrats on your launch.

It also sort of does, I mean, indirectly sort of support what I was bringing up before, which is that, you know, these are going to be hugely computationally intensive, like applications. And so it does, you know, sort of entrench some of the existing clouds if these products are actually built, you know, for Microsofts or Googles or the rest. But let me pause this part of the conversation. because I want to bring up our good friend of the show, Mike Nano, who happens to be in an Uber well enough.

So we will see how his audio is. But I've pinned, I believe I've pinned a tweet about something that Mike wrote recently. So Mike's been on a tear ever since he came on our show. I take all credit for this, but also since he left Spotify. And he wrote about something that I think is actually very germane to where we started this conversation. And we're talking about culture and cultural production. And specifically, Mike was writing about essentially sort of like the creator supply chain and how the economics around that is changing.

And was asking the question, you know, is the crater economy, which we were so hyped about last year, again, during the whole crypto craze, is that dead? So I don't want to like spoil the whole thing. But Mike, do you want to come up and just sort of like provide a bit of your, you know, insight and we'll see if we can weave these conversations together? Sure. Hey, everyone. I am indeed in and over.

And I will give you guys some credit for kicking off. this writing tear I've been on because it kind of all started after I was on your show this summer. I think maybe inspired you inspired me to like want to get some more thoughts out there into the world. So thanks so thanks for that. Yeah. Appreciate it. So yeah, I did, you know, I wrote this piece, published it yesterday. I've just always been a person who's been passionate about creativity.

I was a musician growing up, a photographer, I was into art. know, I was a, you know, wrote code, I built a company. I'm just like a fan of people sort of making things. And I become sort of like disillusioned by this whole notion of the creator economy because it felt like the more and more people talked about it and we thought about it and companies got funded. It was like all of the attention and the focus was going on this very, very small subset of companies that were serving an extremely small subset of creators. What I like to think of is the 1%. Basically, the people that are already making money, the people that already have distribution, the creator

economy was building sort of this financial infrastructure for those creators. But if you think about the rest of the creators, then 99% or 99% of people, they're all creating too. Like, we're all creating. We all create now every single day. We send tweets. We write emails. We write blog posts. We take photos. We take videos. And then there's this whole demand side on the other end, right, where people are consuming all this content. All day, we just consume constantly. And I actually think it's way more interesting to think about the supply chain or the, the businesses and the business models that get built around everything else, the 99%. I think the creator economy, that 1% is just kind of not that interesting and kind of small. And of course, all of this stuff gets

propelled by everything. It sounds like you guys were talking about when I joined, like, you know, AI coming in here and democratizing creativity. creativity even further, machine learning, making it more efficient for people to find the content, the information in the media they actually care about. Frankly, something as simple maybe as just 5G internet becoming more accessible globally and throughout the world. I think media and creativity is sort of truly, truly ubiquitous at this point. And I think we are all creative. And I think if you really think about all of the software and products and services that power creativity. We're talking about like a multi-trillion dollar opportunity, not just maybe this small

little thing we've been calling the crater economy the past couple of years. So anyway, that's basically what the piece was. Maybe it's kind of obvious to everyone that we're all creative, but I just thought it was worth acknowledging. Well, I think the reason why, one, I wanted to get you up here. And two, it sort of dovetails into this conversation we're having. And it actually started kind of, one, why is there a bunch of investment happening in kind of the AI space? And then we were also talking about creativity and culture and who gets to monetize it and who has a right to participate in the generative output of a lot of these AIs that are being created now and being used. And I think this dovetails into what I've been seeing lately with some of the like super

interesting stuff that's been created by applying those algorithms that again are trained on past creative work. I mean, it's sort of like it is like, I hate to use this metaphor, but just like the raw material or the oil, you know, the unrefined oil or the cruise. I suppose, of creativity to create all sorts of new things now that we have this combustion engine, which is artificial intelligence. And so bringing those things together, and this is, I think, where your point comes in, Mike, that if everyone kind of imagines themselves as participating in this creator economy or a creative economy, like what does that mean? And how do we, if not distribute wealth, like distribute access or privilege or experiences? And Parker, this is

something that you brought up kind of on the back channel, that was really interesting, was just changing the whole mentality around, I think, ownership economics and scarcity economics, and that if we, again, I know when I get too much into like UBI space, I still don't know really what the concept is for how we support humans nurturing their sort of embodied needs, you know, their natural needs. But if the kind of asymptote of human culture is to enable everybody to experience their God-given gifts or greater-given gifts or however, you know, whatever deities you prefer, then if, If that's what we're moving towards, then what you guys are both saying is sort of like bookends of the same kind of concept. Does that sort of land or resonate?

I guess I guess I'm not sure I understand the sort of the framing. I mean, I think we if you're saying, hey, this is really exciting technology because it's going to enable us to do new and interesting things, create more absolutely, right? I mean, I think, you know, you can. I'm not quite prepared to tie that into UBI and whatnot, but I do think that it's interesting that we are, I mean, we are all creating, right? I think people get hung up on this idea of like creating as something you do and get paid on, right? My worldview, my frame is we create because we need to create, not because we get paid to create. I think that's, we get caught. This is why I'm so focused on the policy, right?

I think the policy often starts with the premise of we need to pay creators so that they'll create. And it's like, well, no, we never did that until the last, that's a very recent thing, right? People weren't making cave drawings because they were getting paid, right? We just had to do it. We want to create. So I'm excited. Or there was no, there was no, like, comprehensive framework for it. Like you had patrons, you had, you know, occasionally governments paying for it.

Okay, let me reframe it for you, Parker, in a way that I think you were talking about earlier. In the sense that back to your tweet about, like, using a pen and paper to create an image, using Photoshop to create an image, like using tools to do something, right? Are you a believer in, it's not that these tools will come and destroy the, artist. Are you a believer that, in the same way, that robots have never actually destroyed entire industries, new industries be created? Do you know what I'm saying? Like, is this, again, sort of that man-computer symbiosis thing where we don't know yet? Because it's so so early. I don't think you can destroy. I just think regardless, there's no world in which

computers will ever exist in a way where you know what that's the question no computers can't create can't destroy creativity that's it well we we want you know look I am a really crappy musician I just suck right but I love to do it I get out my guitar and I play it and I like to do it I mean everyone loves auto tune right yeah so yeah exactly So, you know, we, I think humans have a innate need to create, right? And so the computers will do their thing and we'll keep creating. And I think in the short and medium term, it's much, much harder to look further out, right? We will continue to create and in some cases will be better than the machines.

Or we will continue to create and we will do it with the machines and with the machines will create better things than we could have created on our own. There are going to be cases where the machines can do things, you know, I give the example of competing against Fiverr, right? They're going to be cases where instead of paying a human on Fiverr, I spend money on Dali or something, right? I, you know, this is how technology works, right? It eats the low-value use cases and then moves up. And so I just, I'm not worried at all about, I think sometimes people with angst around, you know,

the computers taking away our humanity. I just don't worry about that at all, right? Like, we are creators, we're going to keep creating, and it's going to be awesome. And if the computers can really, maybe this is looping back to what you were saying, Chris, the real problem with UBI, if you look at the economics of UBI is that we need GDP to grow. Right. But that could be possible or enabled by some of these advances in artificial intelligence. Yeah, no, exactly right.

Yeah. So if we're talking about creativity, it's precisely the opposite, which is if the machines are so good that they can do all these jobs for us, we're just going to sit around and jam on our guitars all day. It's going to be awesome, right? Because the machines will just do the work for us. That will be the leisure economy in a way. The sort of ultimate expression of the leisure economy is actually the creator economy, where our needs are met and the drudgery of the things that we don't really want to do

can be set aside to the machines, and we can just engage in our creative pleasures, I suppose. Yeah. No. Mike? No, yeah, yeah. I think we exist at a time. There are eras in culture.

Right? They reflect on technology and I think you can see it and for example how we write fiction, right? So they're like dystoping areas of fiction and then optimistic areas of fiction, right? There's like the Star Trek stuff and then there's like, you know, the cyberpunk stuff or whatever, right? We're in an era where people are really negative about technology. We feel very alienated from it. That doesn't personally resonate with me at all. I think it's fun and awesome time to be alive and it's just going to be better. So, yeah, if we can take all the computers and make GDP 10X what it is,

and we can all just hang out and get UBI. Everyone's just kind of like hang out and complain. Like, that's probably what it would be. They'll just have more creative ways of complaining. But we'll be way better at guitar, you know. I want to bring it Matt Hartman, actually, a friend of mine from BetoWorks. Matt, did you want to chime in? Probably a friend of Michaels too, yeah.

Sure. I mean, I was, I've listened to you guys. This is great. I don't know if you've moved on for this, but I think that the disconnect, I think, between what Mike was talking about in terms of creator economy of the last decade versus the next decade and what GPT3 means for it, I think might be that the last decade was about distribution for creators and the new year is about the tools for creators. And we were calling things that were for distribution tools. Interesting.

Yeah, that's a great point. Yeah, I agree with that. I also very much agree with what Parker was saying. I mean, I know like these these conversations are probably great with there's like tension and disagreement, but I just want to like pick plus one. What Parker was saying, I mean, like if, look, if computers and AI want to want to take away the low value creation from me, like I would be happy to spend my time creating something else or something something, something of more high value that I don't normally have the time to create. So I actually just think of anything, it won't replace the creativity that we as human beings were necessarily born to do. It'll just be additive. Like I think it actually leads to more creativity and more content and more things for all of us to consume.

So I agree. I think this is actually going to lead to a better future. You know, one thing that we haven't talked about that I think is interesting, just as we're sort of riffing on this, is the way in which these tools can actually provide kind of a feedback loop. one of the things that I guess that I'm thinking about is one of the challenges that we run into is that there's just not kind of equal access to all the different sort of talent that exists in the world.

Like if Parker you really want to learn to play guitar, you should really kind of like hook up with like Jimmy Hendricks, you know, RIP, but like nonetheless bring him back and he can like teach you in a way that allows whether it's time travel or just like talent to travel so that you can actually access that resource wherever it is. And AI could actually serve

as that mechanism to learn faster. Let me give you a very specific example of this. I used for the first time a product that I'll be hunting or some version of it soon called Headroom. And it's available. You can go check it out. I think it's goheadroom.com. And it was built by and is being built by Julian who used to work at Google on Google X as well

as I guess his, I don't know if he's a co-founder or sort of like head of their AI stuff, but he was also from Google Brain. And so these guys have a lot of experience with neural nets and language models and so on. And, you know, this type of functionality and feature set, I think, is becoming somewhat more common, given the world that we live in with lots of video calls. So essentially what happens is you'll have, you know, your 30-minute or 60-minute call or meeting. And then at the end of it, there's a summary that's provided to you with highlights of the most interesting parts of the call. And what it does is it actually uses computer vision of the video stream that's coming through to measure. your engagement in the conversation based on voice tonality, based on your facial expression.

Even if you're looking away kind of like up at the ceiling or something, but you're nodding your head in a way that the algorithm knows how to pick up, it'll be like, oh, this person is actually listening and they're engaged. They're not just like disengaged because their eyes aren't on the screen. So there's a lot of subtlety and nuance in that. But if you can imagine that if you are a speaker or if you're someone who is actually running meetings and you have this as a tool for getting reflections on your contributions, this starts to be a way to actually improve and enhance human capability because most of us don't

actually live in a world where we get that much useful feedback. We're kind of like just, you know, fucking around, you know, in the world without that feedback loop. So from an artistic or creative potential aspect, maybe humans actually get better at entertaining each other, like, you know, through the real world by actually having these, these types of tools as another layer to actually understand themselves and i don't know that that's something we've necessarily talked about but i wonder how many artists are actually using this to improve their own capabilities i mean i'll i'll tell you as someone who you know talks to artists from time to time you know they record themselves and listen to themselves and they have said serious artists yeah serious artists you know like i if i remember a

friend of mine like going to a show and he's like hey we plug into the soundboard and record me and we got home and he's like i'm going to go hang out over here and listen to this while you guys you know shoot the shit or whatever, right? So, yeah, I mean, maybe we can all do that. We certainly don't. I mean, I don't record my meetings and go back and listen to them, but that's obviously a great way of, you know, self-reflect. Can I, uh,

tangently, but this is, this is my last, uh, sort of stuff. Yeah. The, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, top of hacker news for a while. One of the things that, the, the, the, the, the, the, the, the things that the huge discussion on that was about was the idea of creativity getting muddied where it's we've been talking about okay do I own this does the the algorithm own this or whatever one of the things in that conversation that I found the most fascinating was what if we're entering

an era where no one can tell the difference in the same way that like sort of like deep fakes and stuff. No one can tell the difference of what is real and what is that. What if this sort of blending of AI and human creativity will make it sort of not only incomprehensible where the creativity comes from, but also it might not matter. Does that make any sense to anybody? Yeah, I mean, so I'll just sort of like try to build on what you're saying, because I saw this thing earlier today, and I'll pin this tweet, where, So Unreal Engine has this technology called MetaHuman, MetaHuman creator, that allows them to essentially create incredibly lifelike avatars.

And so what is being done, and I'm seeing this happen actually in a couple different spots, is they will take these MetaHuman creator things super lifelike. They'll drop them into kind of like a Metaverse or, you know, Fortnite type experience. And then they'll actually simulate kind of like dance moves or other types of emotes. And you really can't tell when these, whether they're NPCs or whatever, non-player characters, you know, come up to you and interact with you, whether they are actually programmed to interact with you or not.

And so gameplay, and I actually think, Brian, you were talking about this with your son playing the Matrix demo. Right. Essentially, like, out of nowhere, he starts just like, I don't know, it was sort of like a Grand Theft Auto type experience. Like playing, you know, how old is he, like seven? Six. Well, as opposed to Grand Theft Auto, he was trying to play by all the rules.

He was trying to stop at every stoplight, not run into other cars. Yeah, exactly. Which I thought is both hilarious and genius, where essentially doing the banal becomes the thing that's so interesting because when you're having these banal experiences, but there's almost like no risk to you because there's no shame. There's no kind of like social dilemma that you're running into as to, you know, the social cost of that, that it could actually become something that people are, I don't know, because like I guess what I think about,

a six or seven year old now growing up in that kind of environment where everything becomes somewhat fungible, it will change the types of things that he thinks that he is able to build for and is able to create and is able to contribute because his set of assumptions are going to be completely rebased at a different level than ours were. I know that's a little bit different than what you were saying, Brian, but that's a little bit, but I was the one that took the gummy, but Mike, you pipe up. Yeah, what I was just going to say was, I mean, her, for the question you sort of pose, Brian, like, is it just a matter of sort of abundance and scarcity? I mean, in a world of scarcity,

people want to really clutch to the things that they have and they want to claim ownership so that they can extract value from them. In the world of abundance, things out obviously lose value, right? Well, it's not losing value. It is what I was saying, like, the muddying of, I think in the conversation people were like, okay, I created this and there was like a graphic where it was like the internet says, no, you didn't. It's just the thing, right?

So like is that kind of what we're talking about where the act of creativity and the same way that like memes go out and no one knows who was the first person to do that? You know what I'm saying? Yeah, exactly. I mean, I kind of feel like in a world where, you know, AI generative media is completely pervasive and anyone and everyone is making content all the time. maybe ownership frankly just doesn't matter. Like maybe I don't care that I made this thing and you're using it and you're

remix thing. I think I think Parker is talking in that direction at the very beginning of this conversation. Well, this is a great example, right? I think that, again, I think that like this is not deterministic. We're talking about this like, oh, the future is going to be a certain way and we're just going to try to figure out which way it is. Memes are a great example, right? Like we had a conversation about information wants to be free and MP3 sharing and whatever, and we decided, no, right? That's immoral. And that's by, you know, this is the reaction that you're seeing around this AI stuff, right?

Is people who are like, that's immoral. Whereas with memes, you could imagine a world. Obviously, this didn't happen. But you could imagine a world where memes were considered immoral. And we all sort of, we go, well, like, you know, you're not making. Or Parker, like you were saying, we were talking about litigating. and things like that. We haven't had huge lawsuits of people trying to claim memes and stuff like that. That could have happened. Yeah, that could have happened. It didn't happen. I mean, you could, you know, we could come up with some explanations for why these things are different and why, you know, why this happened. But I actually think it's not deterministic. I think what happens with AI and how we think about ownership and creativity and these things. Like, right now is the time where we get to sort of put out ideas and everybody gets to choose. on them and we get to decide whether this is moral or immoral or whether it's more of a collective

ownership thing, whether algorithms can create truly unique things or whether they are, you know, stupid and just copying, right? Like this is why now is so interesting and important because we are going to fight a battle that determines what you are allowed, what your children when they are older are allowed to create and do for a living. And we were talking about who gets paid on it. Again, I think that's like that determines what economic models and what creative models can exist.

But like, now is a really interesting time to go write your thought pieces because you have a lot of leverage over where the culture goes right now because no one is made up their mind. Okay, so I'm going to bring this to a bit of a close. But I think I have two thoughts. They're sort of germinated there. What is that Mike has to go and write another piece about this kind of like direction with some regular like some recommendations for either our lawmakers or for the next generation of people

who are going to be writing rules about this stuff because as I understand it, there's a lot of people in Washington right now, you know, helping to write draft legislation around crypto and around money and around those things. And what I find so interesting kind of about this, and I guess it's because we're talking about, you know, Brian Sixth, is because if I think back to my experience in high school, and it sort of has become more and more clear to me how this experience really shaped my life and everything that I did since, as I created my high school's website back in 1999,

and I created all the club's websites, and I hosted it myself. I had my own little stupid server, and one of the websites that I created was for a Gay Shirt Alliance. Now, this was in New Hampshire in 1999, and this was not something that people were ready for. And so as a result of creating this website and putting that club, that club's banner ad in rotation on my high school's homepage,

along with the Art Honor Society and Band and all the others, I was suspended because the school, the institution, the structure that existed for my education and for the furtherance of some kind of knowledge, decided that it was too dangerous to allow someone like me to be able to interact with meme and culture propagation at that level. And I feel like we're having a very similar conversation to the one that was not really going on back when I was in high school now about these AI tools, that there are going to be kids who are writing their college essays using GPD3.

They're going to be kids who are completing art class by using stable diffusion. I mean, a couple weeks ago we had this first story where I believe someone tried to copyright or won some art contest that I'm, used stable diffusion, right? It was a new thing that had never been seen before, and now people are trying to put that cap back in the back. And I guess I wonder if, you know, my high school principal had been successful in shutting me down and shutting the internet down. And if people in thinking like his had shut down this whole world of web publishing, if we would have actually been better off for it? If the Luddites had won, would it be better that we're all making your own, you know, sweatshirts and T-shirts now? Given my life and my experience, I would say no, but I think that

the conversations that are happening now. And so if you guys have thought pieces that you want to go right to help people think about and how to contextualize these things that go beyond just who gets to make money from this stuff, I think that would be incredibly valuable. Yeah, Mike, Chris just gave you a homework. I'm not sure I'm smart enough to write that piece. But one anecdote I will say is I was hanging out with a person tonight who will remain nameless who was currently in college and they revealed to me that lots of students are already using AI to generate their essays for them. This is my point, right? I mean, it's a new form of doping.

It just doesn't have to be in, you know, biking or something. Well, I guess I would say we're going to adapt. Education will adapt to this, right? Like, we'll figure out how to test people in ways that test their knowledge as opposed to. But is that the right thing to do? Or is it actually to put the intelligence? No, that's fine. No, but that's fine.

I think the right thing to do is to say, hey, these tools, like this is like the calculator, right? Like, you don't say we're going to ban calculators. You say, hey, we're going to do. Well, in some context, that's okay. Like, you know, in elementary school, you figure it out. And then later on, you design tests for that. I guess my call to action for people who want to go write thought pieces, I think there's two different ways to think about it, right? One is how should the general public or policy people think about this? Right?

And we sort of talked about that, right? The other one is, how should tech think about this, right? What should tech do? We are in an era where there's a very, we're in a very bad faith era of these tech, right? That's right. And so I think, for example, we talked about the GitHub thing, right? The problem with the GitHub tool right now is that it was not designed to, you know, not be got, right? So it was not designed. It should have been designed in a way where if someone

says, hey, copy this thing for me so I can fuck these guys, right? It says, I'm not going to do that, right? So tech needs to do a couple things, right? One is I think it needs to design these tools so that it can't be got so that you're not infringing specific copyrighted works, right? And that's not that hard. You just need to put it in the backlog and do it, right? I think the other thing that's interesting an idea I've been thinking of that I'll just throw out there is, I actually think that tech needs to go on the offensive, which is to say, look, when these people pop up and they say, I'm going to come after you because you're stealing ideas, a cool thing about these tools is that if a specific artist comes at you, you can say, cool, you draw horses. I'm going to go find all the horses that look like your horses because they're just in my corpus, right?

So if you want to get into the infringement game, I am going to come for you. We're going to build these tools. to dispel the myth of the independent creative genius? Because I think it is a myth, right? I think we'd be much better off as a society if we thought about creativity in a more collective and iterative way. And so I think there's sort of a call to action for the general public or policymakers,

but then there's a call to action for technologists, right? We need to be thinking a little bit differently about how we build these tools and use these tools and approach this conversation. And so, you know, I don't want to write these pieces, please somebody do it and I'll re-tweet it. You know, Parker, you referenced the YouTube history a couple times and the Napster history or whatever, but that problem got solved essentially by allowing stakeholders to make money, right?

Not everyone makes as much money as every stakeholder in the equation, but that's the reason why YouTube survived when Napster didn't. Yeah, exactly, yeah. So, like, that's sort of the thinking, maybe we should be thinking about right now is how to make everybody happy to have a taste of this. That's not going to happen with this. You're totally right. That's a great point that Napster had no way to get people paid.

YouTube did, right? But how do you pay somebody when you're looking at 400 images of horses and making a novel work? There's a question of should we pay somebody? I would say hopefully not, right? If we can lower the cost to society of creating new art, that's better, right? So we don't want to pay people. If we have to pay people, okay, we will. But how do we even think about that, right?

There's these 400 horses and we're making a new horse. Do we, like, and the person, and it's, by the way, it's me creating the horse for my personal use. What do we do with that, right? Like, am I paying money? And now every horse image that I wanted to, I used to draw for free. And now I have to pay every time I make an image. Like, I don't. I want to make a joke about horse e-books and like that's wrong or something like that.

Go on. No, I'm just saying, I think that I don't think that we, at least for the tools that we've been talking about, there's going to be a model where we can viably pay people that are creating the art that it goes into these algorithms or whatnot. There's just not money there, right? So I think what we're really saying is if money has to change hands, then these tools just aren't going to matter. No one's going to use them, right?

These are the stakes. The stakes are, this is a generation of creative tools that are either going to be de facto legal or de facto illegal based on the economics that sort of fall out of this cultural discussion. So the task at hand for everybody listening is to try to shape this cultural discussion in a way that leads to the legality of creative tools so we can all have more art. That's my call to action. I, just to end this, I happened to do a search on AI ethics while we were chatting.

And for some reason, the first thing that came up were these two links that I've pinned to the channel from the U.S. Chamber of Commerce of all places around this bipartisan commission on artificial intelligence. Now, I don't know if it's relevant or interesting, but I think it makes the point that these conversations are happening at the highest level and that whatever kind of goes into their you know, sort of neural net is going to be the thing that's going to determine how or if people can use these things and if they can use them and apply them for, you know, pro-social and positive productive uses or not. So very much to your point, Parker, this conversation is happening

currently and it is not clear to me at least how many of sort of folks in our, you know, world or our community are participating actively in those conversations. Okay, so this is great. I want to bring this to a close, guys. This is amazing. I've got a wrap for you real quick. No, I'm saying I have a wrap idea, which is that you guys just told me that AI will allow college folks to cheat on their exams. And my first startup, EditMeNow.com in 1998, was literally that, except I had to hire a team of 20 to 25.

grad students to do it. So if you're saying that the AIs have taken cheating in college out of the equations, you would never have gotten your start. AI has taken over and so there's no point. Let's allow anyone that's been on stage to jump up and plug anything and then we can wrap. No place. Just want to say thanks for having me. It's just fun as always. That was Michael. Yes. Mike, no, real quick. what's your firm that you're at right now? Ah, yes. That's true.

You announced. Former founder and founded anchor podcasting platform. So that's Spotify for a while. And now, yes, I am a partner at Lightspeed. Yes, that was, I think not known. Thinking a lot about this. I think a lot about creativity and generative AI and all this fun stuff. So, yeah, thanks for the quote.

Thanks for coming on. And Parker, of course, since you've been here the whole time. I've got no plugs. Thank you for having those. I'll plug Parker. He's one of the kindest and most insightful people in the tech space, full stop. Awesome. And I will plug Miguel's startup, which is at vmind.ai.

I'm sorry, vminda.com. It might be both. He's not on stage anymore, but wanted to make sure that he got a little shout out there. Indeed. Yeah. Well, guys, I mean, I guess we'll have to keep this conversation going. Parker, it was amazing having you. I believe you're actually like, are you in Oakland, Berkeley or the Bay?

No. We are occasionally there. We moved to Philadelphia in the COVID exodus. So I'm kind of remote going back and forth, but I miss it there. It is cold as shit here. It has gotten that all of a sudden, yes. Yeah, I'll leave you with this, this idea, which is I saw somebody tweet the other day. It's like there's a bad season everywhere you go.

And, you know, in California, it's tax season. So every other season is several here. Wait. Wait, where are those, Chris, where are those sound effects? Oh, shit. Yeah, let me find it. Yeah, yeah, yeah. Oh, it's not a, wait.

All right. Sorry. Hey, love everybody. Love everybody that came on. Everybody that's listening. Love Chris. Right, Chris? Thank you.

Love you, Brian. Let's go. All right, great. It got me sitting. All right. Good night, everybody. Thanks.