Subscribe
Share
Share
Embed
NET Society is unraveling the latest in digital art, crypto, AI, and tech. Join us for fresh insights and bold perspectives as we tap into wild, thought-provoking conversations. By: Derek Edwards (glitch marfa / collab+currency), Chris Furlong (starholder, LAO + Flamingo DAO), and Aaaron Wright & Priyanka Desai (Tribute Labs)
00;00;16;15 - 00;00;18;12
Aaron
Fable's back. What do we think?
00;00;18;15 - 00;00;24;04
Chris
I'm using a lot of bad words at it, so I don't think much at the moment.
00;00;24;06 - 00;00;26;13
Aaron
More or less bad words than you normally use, Chris.
00;00;26;14 - 00;00;53;06
Chris
Well, there was a certain tinge and harshness to my bad words right now, because they're filled with frustration and disappointment, because all this thing is doing is chewing through tokens and not communicating thinking. 15 minute shell wait times is cool and just generally being an asshole. And so I'm a maybe I could I shouldn't have dropped it in at the end of a huge build and very long context window.
00;00;53;06 - 00;00;56;09
Chris
But yeah, I've not been happy.
00;00;56;10 - 00;01;13;22
Aaron
I think meta should pull you over to run their reinforcement learning like move over current team and you should just you should just get in there and just yell at the machines more directly. It could be like a like a parent that were the models never good enough. You're not going to make it, buddy. Got more work.
00;01;13;22 - 00;01;17;21
Pri
To do. The tiger dad. Tiger dad of AI models.
00;01;17;22 - 00;01;42;07
Chris
Why do you always find the worst perspective job in the entire universe? That especially doesn't appeal to me. And then say, you know what I want you to do? I want you to jump into the coldest Fang company with a founder you don't particularly care for and take over the most dystopian project they're running in, which half the company has already revolted and has the smell of rotten eggs on it.
00;01;42;08 - 00;02;08;18
Aaron
It's so weird to me still, because, like, I mean, a they were known as like such a strong engineering culture and B, we really haven't heard much from them. And I thought that that was surprising. And I just, I wonder if this was all kind of worth it. I know some people were like musing on that, and I'm sure that they had like they overheard or maybe they had some, you know, issues that they could button up there, but it just feels it feels like that companies in general been pretty well managed, right?
00;02;08;19 - 00;02;23;16
Aaron
And I think they're like, numbers are still like I watering too. It's not like it's impacting their performance like from the the market's perspective. But yeah I find what's going on in meta like super weird. But aren't they doing some cool stuff. Did you mention that.
00;02;23;17 - 00;02;30;11
Pri
Well did you. Yeah they are. But did you hear about like the engineering culture at meta right now? Apparently it's like all time lows.
00;02;30;12 - 00;02;50;12
Aaron
Yeah, yeah. They're like they reassign people to like, label things. And there was like a story that like at a, you know, company wide meeting that they said that they're like accessing developers code bases and even like their potentially their email to use it to train stuff. That's at least my recollection. I may have gotten something wrong there.
00;02;50;13 - 00;03;09;26
Pri
Yeah. And apparently like in order to like make the culture better, they were like giving away free snacks or something like, we're going to improve. Like we're going to give you a bit larger snack budget to like, help generate better, I don't know. Cultural milieu and milieu. And it just like did not work. People like backfired tremendously.
00;03;09;27 - 00;03;12;20
Aaron
The snack budget didn't work. Oh man, that's not good.
00;03;12;21 - 00;03;13;09
Pri
Yeah.
00;03;13;09 - 00;03;32;10
Chris
We're going to take away all your agency, but we're going to give you more snacks. Yeah, I can imagine why, like grown ass adults at work are not going to respond to that. This isn't kindergarten, but I mean, it was Instagram the last great thing they did. And because that has to be over a decade old at this point.
00;03;32;11 - 00;03;35;00
Aaron
You mean on the product side? I mean, I do.
00;03;35;01 - 00;03;38;29
Chris
On the on the everything side because that wasn't the product. That was an acquisition.
00;03;39;01 - 00;03;59;14
Aaron
Yeah. I meant more like, well, I mean, they they evolved Instagram to their credit. Right. Like they were able to stave off some pretty tough competition from TikTok. Right. Like they they built reels. That wasn't, I think, in the product at the time, you know. So they did they definitely evolved it and kind of maintained its its place and kind of same thing with I guess with WhatsApp.
00;03;59;15 - 00;04;16;05
Aaron
I mean, it still seems like those ecosystems are developing products still really good even though the competition has gotten stiffer. But yeah, I mean, I mean, the metaverse stuff is always like a little bit a mixed bag for me because I do appreciate them pushing on that. I think Mark's going to be vindicated in the long run on that type of stuff.
00;04;16;05 - 00;04;30;09
Aaron
And I guess like some of the the glass stuff hasn't been as panned as like the, the competitors. Right. Like a lot of people were pretty bearish on like Snapchat's attempt at that. And obviously Google and some of the other big players haven't gotten back in the ring there.
00;04;30;10 - 00;04;45;29
Chris
Yeah, I mean I guess I just say and like they tend to run a copycat culture and this perhaps isn't the age for copycats. And you know, it's a tough they're just not doing well, you know. And so whatever Facebook was there. Certainly not.
00;04;46;01 - 00;05;10;25
Aaron
Don't you think it is the age for the copycat. So because I mean that's what people are arguing. Why like, you know, Chinese based open source models are doing so well. They're just like fast copying them like crazy. It's like, you know, like institutional, like institutionalized, like copying and distilling that's being used, I think, in part to make sure that that gap between open source and the hyperscalers is, is not too big.
00;05;10;28 - 00;05;51;07
Chris
That's that's different. Copycat. It's frontier copycat, in which, you know, there's a diffusion of best practices. Meta's copycat is seems aimless and late. You know, the idea that, oh, well, this this didn't work out and that didn't work out. And our whole companies in revolt because of data labeling, you know, at the answer is we should create a copycat prediction market like that, to me, is the losing strategy that has Facebook stuck in the muck is they copy consumer products too late and no one gives a shit.
00;05;51;09 - 00;05;52;13
Aaron
Yeah, I think that's fair.
00;05;52;15 - 00;06;14;00
Chris
And I think once you have a stink on you like that and those, you know, those consumer behaviors are already ingrained. And so what what is Zuck like counting on. Does he think he's going to switch. You know hardcore culture and Poly Market betters over to arena. Or does he think he's going to open that up to new ground.
00;06;14;00 - 00;06;30;07
Chris
And if so like who is that new ground. You know necessarily on Facebook. Like it just doesn't feel right. You know they feel tired. They feel like they're chasing things and they're just they're on a cold streak and out of stuff. I don't even know why we're talking about them.
00;06;30;09 - 00;06;30;24
Pri
Well, okay.
00;06;30;25 - 00;06;33;15
Pri
The one interesting there was interesting.
00;06;33;15 - 00;06;56;21
Pri
News out of meta this week. They like, published this paper in nature with nature. You know, obviously like the research paper magazine. And they're able to do like I mean this is actually innovative where they're able to end to end pipeline capable of real time sentencing decoding from brain signals. So they're basically able to take like brain signals and decode them into words and semantics.
00;06;56;21 - 00;07;19;05
Pri
So I mean, that's kind of getting into I feel like we talked about telepathy a while back on this pod, but like they're actually getting closer to and closer to it and like open sourcing a lot of this information, which to me is definitely getting closer to the BCI stuff. But that was like legitimate information and innovation out of a meta.
00;07;19;06 - 00;07;23;11
Pri
So that's the only reason I would bring that up. You know, the pro meta stance today.
00;07;23;12 - 00;07;28;11
Chris
Thank God. If there's anyone I want at the forefront of mind reading technology, it's meta.
00;07;28;12 - 00;07;44;14
Pri
It's meta. I mean, they basically they're probably best best suited to do that. They have like everyone's brains social media and digital twin. Now they did kind of crush with the glasses though. Like the marketing around the Kylie Jenner. I'm kind of like low key like pro meta right now.
00;07;44;15 - 00;07;52;29
Aaron
Yeah, I think I think that they're they're like kind of on the outs, but maybe we'll come back on the ends in a weird way. I don't know, I just.
00;07;53;02 - 00;07;54;20
Chris
Take the under on that take.
00;07;54;21 - 00;07;59;26
Aaron
Really. Yeah. I don't know. I feel like you can't count them out. Count them out on this one.
00;07;59;26 - 00;08;02;20
Chris
I don't see a meta renaissance on the horizon.
00;08;02;21 - 00;08;18;02
Aaron
Do you think that let's say it's like ten years in the future in the BCI stuff, or just becomes better and better? Is that like the does that finally lead to like, the interface change? Like, is that the most efficient interface when we finally can get our faces out of screens? Is that like the long term hope there?
00;08;18;03 - 00;08;23;00
Chris
I'm not particularly bullish on needing implants as an interface.
00;08;23;02 - 00;08;39;29
Aaron
Yeah, I mean, let's assume it's not implant based. I don't think that would work. I mean, maybe for some people, but that feels like many shades too far for me. But if there was something you could wear or like a sensor you could put on your head, I feel like maybe then you wouldn't need to be glued to your screen as much.
00;08;40;00 - 00;08;48;22
Chris
Well, I'm in the camp of I don't really have any network home automation stuff in my house. I certainly can't imagine giving access to my brain like that.
00;08;48;24 - 00;09;13;08
Aaron
Yeah, but we may be we may just become Boomer there. Chris. That's kind of. Yeah, I feel like the the, the the tides change a little bit kind of on that front. Well maybe we should stop bashing on, on on meta. But did you guys also see that the spin off from OpenAI Thinking Machine worked with Bridgewater, and they are now training a custom model to replicate expert investor judgment.
00;09;13;09 - 00;09;45;07
Aaron
I thought it was pretty interesting. You know, they they pretty much said that frontier models, at least as of now, are like pretty bad at doing a lot of like expert investor judgment. And they were able to kind of help improve that by focusing on data quality. And they fix the issue purportedly by training on cheap labels. So like labeling all the data that they had, checking where the model disagreed with Labelers, and then using these disputed cases to go to like real experts for review.
00;09;45;07 - 00;10;03;11
Aaron
And they used Quinn, and they claim that they were able to get much, much higher accuracy on things, which is which is pretty interesting. So I do wonder if we're going to start to see kind of more and more of this. It feels like either blowback because of mythos like concern about being able to get access to these models.
00;10;03;13 - 00;10;29;05
Aaron
You know, we heard Alex Karp yesterday from Palantir basically saying that there's a worry from lots of enterprises that the hyperscalers are just like feeding on their data, like their weights, their information. I do wonder if we're finally entering this era of of like the the fine tune model. I know a lot of folks were kind of geeked on that when like GPT 3 or 3.5 came out, but it kind of got eclipsed as the the model quality improved.
00;10;29;05 - 00;10;31;19
Aaron
But I wonder if that tide is turning a little bit.
00;10;31;20 - 00;11;10;04
Chris
This is an interesting because for me, it's a little hard to tease out, you know, what the actual benefits are here and whether or not this is really just a concerted marketing push from everyone outside of the Frontier Lab arena to to gain more control over the ecosystem and keep them from running away with everything. And, you know, I would then it's if you look at the sequencing of who who was involved in getting fable off market for a while, what followed from Satya and Viper Soft is a general positioning statement.
00;11;10;04 - 00;11;31;19
Chris
And now you're starting to see this across the board. You know, discussion around, hey, you need sovereignty. You can't hand everything over to these guys. You can't allow them to read wedding you. The answer is you let us do that instead. Right? Like that's the part where you kind of need to read between the lines a little more and then start thinking critically about it.
00;11;31;19 - 00;11;54;23
Chris
And I'm not. I'm not saying like, you shouldn't own your full stack. I'm not saying there's not benefits to fine tuning your own model. Right? It's just where is this message coming from? Why is it coming from these people? Right. You just need to think a little on this versus, you know, just being like, oh yeah, well Dario and Sam scare me, but Alex Karp I'm going to listen to Alex Karp because he's using words like own your stuff and sovereignty.
00;11;54;23 - 00;11;58;05
Chris
And it's like, well, you know, Alex Copper is also a sociopath too.
00;11;58;06 - 00;12;25;21
Aaron
Yeah. You know, but that sovereignty thing, it definitely feels like there's more energy related to it that we got the, you know, Venice closing a big round, which is run from the crypto sector thinking about, you know, hopefully protecting some of those reasonable data privacy censorship related questions. And Eric Voorhees, the the founder that started it, you know, seems to be trying to differentiate what he's he's building on those grounds.
00;12;25;21 - 00;12;34;22
Aaron
Two and at least there's some interest from the investor class to to begin to think that through a bit more. I don't know that that kind of makes some sort of sense to me.
00;12;34;23 - 00;13;05;29
Chris
I think it does make sense. It also, though, like it's posturing and positioning and the hard truth of the matter is there is no sovereignty like that's an illusion unless you're going to buy your own data center, stand up your own servers, run your own local models, have a development team whose job it is to scour open source, bring it in-house, customize it into your stack self, train on it, and you can do that at a high enough.
00;13;06;01 - 00;13;35;25
Chris
You can do all of those things to a high enough degree that you're actually going to outperform people who are not making those sacrifices. Right? Like true, true sovereignty is, I think, more of an aspirational goal, and it's more of a in a matter of checks and balances, because the actual notion of two sovereignty doesn't exist, like you're in a networked environment that has chokepoints that you can write around.
00;13;35;25 - 00;13;54;15
Chris
At the end of the day, though, like you're never truly independent because this is just the environment you're working in, but you can't have it both ways. Now, is it a whole spectrum? Yes. Are the degrees of this? Yes. Are there will we see in five years from now, like a delineation of what you should and should not do?
00;13;54;18 - 00;14;17;17
Chris
Most likely. Right. And companies will take like basic steps around owning their data, you know, customizing the tooling that accesses that data. But like we we kind of have these discussions around AWS as well. And, you know, at any time, like a major region goes down and a big consumer service falls offline. Or, you know, Cloudflare screws up like deployment in one of their routers.
00;14;17;17 - 00;14;33;04
Chris
And, you know, like a third of the world's internet traffic goes off, like we have these same conversations. But is AWS and Cloudflare like, are they still these central massive like choke points? Yes. We never got away from that. And those concerns have been around for a long, long time to.
00;14;33;05 - 00;14;52;08
Aaron
Do you think that like hyperscalers also adjust to like like as they become like quote unquote enterprise hardened? Like if they hear all these concerns, they'll probably just adjust at some point. Maybe they're just doing like a, like a, like a savage calculation here that they shouldn't do that because the data is just too valuable for them to get to the next level.
00;14;52;08 - 00;15;08;29
Aaron
But like at some point, like maybe the returns of kind of treating their customers poorly, if that actually is the case, and I'm not alleging that it is just changes. Right. And they just become a little bit more enterprise hardened. And this is just the push to kind of get them there. Like I wonder if that that may be what what's happening?
00;15;09;00 - 00;15;34;11
Chris
Yeah, I'm a very strong believer in the idea that the only thing that really exists or postures and postures are situationally dependent, right? Like you, Aaron, don't have one true consistent personality. You have reactions to a bunch of things you're presented with at any given moment in time, and you contradict yourself all the time. We all do. That's what it is to be human, right?
00;15;34;14 - 00;15;34;26
Aaron
Yeah, I know.
00;15;35;03 - 00;15;40;15
Chris
I'm sorry. You are a paradigm of virtue and consistency. You are. Thank you.
00;15;40;17 - 00;15;41;03
Aaron
I appreciate.
00;15;41;03 - 00;15;45;15
Chris
That and I however, our flip flopping, normal people, silver tech companies.
00;15;45;16 - 00;15;46;23
Pri
We're degenerates.
00;15;46;26 - 00;15;58;29
Chris
And so yeah, like, you know what what a corporation is sitting at any given moment is a reaction to the moment. And, you know, like what they're going to be saying two years from now, two quarters from now, it can be very different.
00;15;59;00 - 00;16;02;07
Aaron
Yeah. Completely. It's it's definitely super interesting.
00;16;02;08 - 00;16;30;14
Chris
But it is like more broadly that you're now starting to see, hey, we got to we've gone from AI is going to accelerate and change everything and let's get all behind it and let's push it to now a oh wait, we're we're all out here. We're all kind of, you know, it's like after that battle is over, that army now, you know, it was a bunch of tribes that came together, and now they got a kind of position amongst themselves and position for themselves.
00;16;30;14 - 00;17;04;07
Chris
And so you're starting to see disagreements or like different takes emerging. And those take some merge through, you know, interviews on CNBC through, you know, vague post Twitter articles about the importance of sovereignty, like the idea that Microsoft is saying, hey, you need to own your stuff inside our world, right? Like, that's what they're really saying, right? And so they're not particularly saying you need to become truly sovereign.
00;17;04;07 - 00;17;08;27
Chris
It's they're saying you should trust us a little more and then you should trust these guys.
00;17;08;28 - 00;17;27;10
Aaron
Yeah. I mean, that's how it plays for me. Like, yeah, we may be just as in trustworthy as you think we are, but not as bad as these guys. Right? That seems to be the positioning. I mean, I mean, it makes sense because it's also like, you see these hyperscalers kind of move into like all these different categories.
00;17;27;11 - 00;17;54;02
Aaron
Like they're aggressively like just gaining ground, right? Like, didn't OpenAI this week come up with a personal finance like, like product you're seeing? Is it OpenAI or anthropic working with Eli Lilly to begin to develop pharmaceutical technology or insights or new new research, which is which is hopefully like great for everybody. But if you're a pharmaceutical company like that must give you a ton of pause.
00;17;54;07 - 00;18;01;23
Aaron
Now you're like, you know, giving your data to somebody that you're potentially competing with on the other side. That's that's tricky.
00;18;01;24 - 00;18;22;11
Chris
Yeah. Not only are you potentially competing with them on the other side, but your internal systems and your cadence and your go to market and the whole way you're set up to operate, you know, is locked, and it's going to be hard for you to get a blockbuster drug, you know, roll out. Let's just say it's ten years.
00;18;22;11 - 00;18;49;01
Chris
I don't know what it really takes to bring a blockbuster drug to market, but let's say the whole life cycles ten years, right? If suddenly, you know, through some combination of like administrative streamlining, AI and, you know, advances in bioscience, incumbent challengers can bring blockbuster drugs to three, you know, three years. Well, you're if you're Eli Lilly, it's really hard to get down to three years without radically overhauling your company in a way that could potentially destroy it.
00;18;49;02 - 00;18;51;15
Aaron
Interesting times, interesting times, guys.
00;18;51;21 - 00;18;53;18
Chris
I wish we could talk about NFTs.
00;18;53;20 - 00;18;54;21
Pri
I mean.
00;18;54;23 - 00;19;22;05
Aaron
You know, I really and I don't know if he listens to the pod, but I appreciated back seats. Who did that Chong's project which which a lot of folks got geeked on. He had a great tweet that's kind of stuck with me. He's like, man, we should just really go back to some of the things, you know, from 2020 to 2023 that that people were really geeked on and just apply a bit more AI, you know, take advantage of the lower cost, you know, block space and have built some fun stuff.
00;19;22;09 - 00;19;33;26
Aaron
And that was just like, yeah, like I sat there and just thought that that was like, really like what I'm hoping people start doing because there's we have so much better tools now, like people did have fun with them. I don't think it was just like a moment in time.
00;19;33;27 - 00;19;59;07
Chris
Right now. It wasn't a moment in time, but it was also maybe, you know, and I know a lot of like true believers in the crypto arts face, you know, kind of threw shade at the fang, an influx that happened in that time. But maybe some of, you know, that inventiveness we got was really about like those Fang people coming in and trying to, you know, Apple fi, the NFT experience.
00;19;59;07 - 00;20;21;22
Chris
And they're probably outdoing AI work now. Right. And so like, yeah, I wonder if even that group of people, it was around and available. And the fact that we're not seeing these things speaks more to the transient nature of this one tribe of, like, fang, you know, outcast weirdos who had nothing to do during global lockdown, got hooked on this for a bit.
00;20;21;22 - 00;20;25;08
Chris
And now, you know, they they're in demand elsewhere.
00;20;25;10 - 00;20;45;08
Aaron
Yeah, that's a great point. I mean, I'm sure that that is the case. Right. And you know, I there was like blowback with those folks coming in. But I don't know if you said this or somebody else did. You know, I think knowing how to build a great product is it's a rare skill. It's like, you know, knowing how to write a great song like product is a hit driven business.
00;20;45;08 - 00;21;03;19
Aaron
And so, you know, folks that have done that successfully in the past can come in and, and kind of avoid rabbit holes or improve and experience, you know, whatever degree is necessary so that it can kind of get that inflection point. Yeah. And of course, they're working in AI now. Like how could they not. Right. It's like a super exciting time to do that.
00;21;03;20 - 00;21;16;22
Chris
And we don't quite have the liquidity. We don't have the easy money to to lure them back in. Right. That's the other part is when markets break down, it's not just one thing. It's a combination of things.
00;21;16;23 - 00;21;41;10
Aaron
Yeah, completely. And I think you I mean that's also something you we felt a little bit of just like what sits behind Silicon Valley and probably like Hollywood too. It's like this mercenary class, right, that can move from like, trend to trend and reassemble, bring their inside specialized knowledge and just crush in another subcategory. Right. Whether that's a new movie, whether that's like a, you know, an emerging tech trend.
00;21;41;12 - 00;21;45;21
Aaron
So I mean, that piece is is definitely super powerful, really hard to replicate.
00;21;45;22 - 00;22;12;06
Chris
Yeah. The interesting thing I kind of feel like lasted longer in the NFT space and is really noticeably absent right now. Is this like financial entertainment class? You know, and I think we saw it in like two different spectrums. We saw it on social fi, which is I think that group is migrated into the PCG and the card collecting and the gotcha type of stuff.
00;22;12;06 - 00;22;49;24
Chris
But then we saw a lot of like reality TV, network competition style stuff, and we haven't seen any of that at all. And I felt like that was a very promising category in a way. It was a bit like when Uniswap bought crypto. The game, right? We haven't heard a peep out of them since. Very similar to like TBN is completely falling off the planet once OpenAI bottom like, I don't know, there's these certain moments where like a thing happens and like, no one gives a shit anymore or like these things just sort of disappear.
00;22;49;25 - 00;22;59;17
Chris
I mean, I like clearly top bought TV and I thought they were like very smart to sell at that moment in time because these things didn't feel durable.
00;22;59;19 - 00;23;30;28
Aaron
Yeah. It's true. I feel like there's just like the same, you know, remember that meme with like, the scrawled out eyes? I feel like there's a little bit of that hitting the tech sector because they didn't get the full taste of that if they were playing around with digital assets. And I feel like that's happening to like how many like I feel myself like scrolling my feet and just being like, how many more like eval posts can I actually want to listen to or like some new approach, whether it's like looping or like which, you know, people, I mean, I know I was doing some of that stuff even months ago before it had like
00;23;30;29 - 00;23;34;25
Aaron
a catchy name for it. Right? So it just becomes like, exhausting.
00;23;34;25 - 00;23;55;26
Chris
Well, we're under the heat dome. I guess being exhausted is a fine position to take at the moment, but yeah, I hear you like all these emails come out and at the end of the day I'm like, I just want to know, can this follow instructions over long durations? Hold up to here and its ability to execute on complex goals?
00;23;56;00 - 00;24;05;22
Chris
Or is it an ADHD squirrel overpowered by its own model biases like none of these benchmarks, tell me a damn thing about how it actually impacts my usage of them.
00;24;05;25 - 00;24;28;01
Aaron
Yeah, and I think that's going to become more important because, you know, you see people understanding either for cost control, reasons for speed, reasons for reliability, like staying on point tasks, that the routing layer of AI is definitely maturing. And it's, I think, a really hard problem to know, like what, two route two? I think that is in part why you're seeing that fable pushback.
00;24;28;02 - 00;24;45;21
Aaron
Like they they must have like a little router that's sitting somewhere in their stack to know whether or not to to fable or, you know, a lower powered model to handle some of the more complicated tasks. So I think that it feels like it's going to get solved, Chris. But I don't know what the solution is for that.
00;24;45;22 - 00;25;04;03
Aaron
Like, none of the evals are great. You know, there is kind of like a smell test on these models. And different people like different models for different reasons. It's kind of hard to like productize that, but I feel like those, you know, some crack crack team or maybe it's an existing team. They'll figure that out over the next couple of quarters.
00;25;04;04 - 00;25;32;15
Chris
Yeah. No, I mean, maybe we'll get tunings or maybe models will just develop certain flavors. You know, I do. I was very happy with sonnet five that it's following the same pattern of sonnet for six, in that it is a steady worker. Is it slow? Sure. You know, but it's it's much better than some of these, like, quantized opus versions that just got, you know, goofy and squirrel ish.
00;25;32;15 - 00;25;49;22
Chris
And so I do think maybe, yes, we'll start to see personalities or like features that, you know, certain model brands need to take on and hold and that become the same way, like you bought a Volvo back in the day because you were concerned about safety, right?
00;25;49;23 - 00;25;54;16
Aaron
Right. Or a Honda for reliability. Right. Which they pressed at the time.
00;25;54;22 - 00;26;25;02
Chris
So I do think that's going to continue continue that to happen. It's also like you know, we talk about models and models all the time. And you know it's almost like the home run. Yeah. Well when you're in the game. But there's a lot more going on in baseball. And I do I do think a lot there's a lot of interesting stuff in the quieter reaches of this world, you know, especially as they, they start to optimize inference, for instance, or, you know, people talk a lot about commerce.
00;26;25;05 - 00;26;56;21
Chris
And I actually, you know, and I say this a lot to you in other conversations, I think that like telecom markets and information markets and communication markets are far more applicable right now where we are. And that's where a lot of the low hanging fruit is. And as we start to get into conversations about routing and blending and mixing and you know what, what the composition of your stack is or your decision making interrupts, right?
00;26;56;22 - 00;27;16;14
Chris
Like all of that stuff to me is really more like a communication market type of thing. And I think that that's like, I think very fertile ground that people aren't paying a ton of attention to either. Now, I'm not worried about it because I think it all gets filled in and it just doesn't get the attention that it should.
00;27;16;15 - 00;27;45;12
Aaron
Yeah, I think that's right. And maybe that attention kind of starts, starts kind of moving moving there. Like as the stack kind of matures a little bit. You know, somebody was flipping around like a graph or chart just looking at the AI boom versus like the railroad boom, the like internet build out boom. And if history is any guide and, you know, sometimes it's not, it feels like we have at least 3 or 4 years of kind of expiration here before the bottom falls out in some way or another.
00;27;45;13 - 00;28;01;20
Aaron
Which which I do think means maybe we start seeing, seeing like the development of the adjacent tools. Now, Chris, now that people kind of know the shape of the model, they know what they want. I mean, that's kind of like what Alex Karp was saying, whether or not he's he's a true believer in self sovereignty and all those bits are Microsoft.
00;28;01;20 - 00;28;16;14
Aaron
Right. But like, I feel like the ecosystem is getting a little bit more mature and like the opportunities and, and remaining problems are becoming a little bit more transparent, at least for now. And they probably will follow like the pattern of telecom. Like you were saying, I always find that really convincing.
00;28;16;15 - 00;28;42;23
Chris
Yeah. And I just feel like Telecom's a little more esoteric, a marketplace. Everyone understands stripe, Shopify, right? No one really understands what it is to be Verizon. And you know, so they're like just they're different beasts with less familiarity. But I think the more native way this is all going to work is like communication market based communication networks.
00;28;42;23 - 00;29;03;20
Aaron
And you see it even like on and it's going to be market based, right. There's been an increasingly steady drum of how to not only like route to models, but right at the lowest cost using some market based pricing, which does seem to make a lot of sense, especially if the spend on tokens is going to be so large, which it has been for, you know, you know, some companies, there's been reports related to that.
00;29;03;20 - 00;29;25;09
Aaron
So just being able to manage that in an effective way, just like you're managing like, well, costs if you're an airline or other commodity costs depending on your business, just feels like this is a commodity, right? And or commoditized business that's going to look and feel like a commodity. AI tokens. And it may follow these like market based approaches, which is pretty interesting.
00;29;25;13 - 00;29;50;27
Chris
Well, there are commodities that require domain expertise, and I think there's a ton of opportunity in developing that domain expertise because we, you know, it's not entirely clear for any given task. Is this something that I pay up for my model on a per token basis. But like a total unit basis, it's actually cheaper and faster to do it that way.
00;29;50;29 - 00;30;19;25
Chris
Or is this a task in which we find the cheapest, cheapest models on the planet and we let them Monje for, you know, some some longer duration, like understanding just how to size and route these sort of things like that. That in and of itself is a massive, massive opportunity. And that's the sort of like wonky, I think that, you know, market based communication network type of stuff is, is a little more esoteric.
00;30;19;25 - 00;30;29;21
Chris
And if you can get your head around it and you can start doing good jobs of that, right? That's an incredibly, you know, lucrative little niche to, to put yourself in.
00;30;29;23 - 00;30;44;11
Aaron
Yeah. And the market's so big. It's so big. Right Chris. So that those niches are going to be massive. It's going to be like you know finance on steroids in that way. Like you find your lane. You're really good at it. You're going to have a massive, massive business indeed.
00;30;44;13 - 00;31;03;22
Pri
Kind of just like something that I think people aren't giving enough. I mean, maybe there is enough like volume on this, but like the open source models are getting pretty good. And like if people are able to self host it and a lot of enterprises are already like moving over and using them, I, I don't and like Venice obviously we talked about like got some funding.
00;31;03;22 - 00;31;10;04
Pri
I just don't know where this all lands. Like why would any enterprise use a closed model.
00;31;10;05 - 00;31;13;29
Aaron
If that's the best reasoning, right? I mean that's that's the.
00;31;14;02 - 00;31;17;29
Pri
The open source models are copying it pretty like they're catching up pretty quickly.
00;31;18;00 - 00;31;25;00
Aaron
You know, if you're if you want to be on the bleeding edge, you kind of have to use the, the the top models. They still are better. Yeah.
00;31;25;00 - 00;31;50;14
Pri
But there's a natural endpoint of like, yeah, there's going to be a natural endpoint of like or like a plateau of reasoning that like the actual like delta of like whatever open source model you're using and the cost you get from that or not is not going to, like, be materially benefited by the fact that you have a three month advantage on some model, like maybe right now, but like in the year from now?
00;31;50;15 - 00;31;51;13
Pri
Probably not.
00;31;51;14 - 00;32;10;08
Aaron
Well, I think that's why you're hearing about it. I think that people are realizing that right now. And so they want to use these lower cost models. They don't want to spend, you know, tens, hundreds, millions, tens of millions of dollars on token costs. So they're looking to manage that and realizing, oh, yeah, if I'm doing like some labeling, maybe I could use this open source model.
00;32;10;09 - 00;32;36;18
Chris
It's also their way of speaking up in this conversation, right? Like earlier I said there was no true thing is like, you know, sovereignty. Okay, fine. But there's ways to control and there's ways to project. And if you were once upon a time spending nine figures to one of these Frontier Labs APIs, and you cut that off and you push it elsewhere, and enough other people are doing that, that's a very strong signal.
00;32;36;18 - 00;32;44;27
Chris
That's a very strong statement. Like, not only is it economically smart for you, but you're also projecting an opinion out into the world that they have to react to.
00;32;44;29 - 00;32;46;15
Aaron
Yeah, it's a good point, Chris.
00;32;46;17 - 00;33;12;12
Chris
I think there's just a lot of like collective, hey, we got to pull you guys back in, right? Like we needed you to be as far out as possible as you could for certain amount of time. Now you can't lap us. And I think that's the larger thing going on, but pretty like. Also, I think the pharmaceutical industry is like a good analog to what you're bringing up because we've had generic drugs forever.
00;33;12;12 - 00;33;14;24
Chris
We haven't lost like the big labs.
00;33;14;25 - 00;33;36;23
Pri
That's true. That's actually a really good comparison. But then they have IP protections and like patent protections are not patent but like whatever, you know drug protections. And they put in the work to do. They've had it. Yeah. Whatever I guess patent protections I forgot like what IP structure they're under. But like they do have I guess they do have that.
00;33;36;23 - 00;33;59;00
Pri
And like there is actually a true lag with generic drugs because they like can't produce it until, you know, the patent runs up or licensing or whatever, whatever the reason. So like there's actually in a it's also your health. So there's like an actual innovation whatever a time box innovation period. Whatever the right wording is for that, that actually is meaningful.
00;33;59;00 - 00;34;04;24
Pri
Whereas I feel like with the lab stuff that's not the same with like intelligence, if that makes sense.
00;34;04;25 - 00;34;39;06
Chris
Yeah. They're shorter, they're more compressed runtimes. The AI stuff is also able to iterate faster. So it's not a it's not a apples to apples comparison. But the shape of it, I think, is instructive enough to say you can exist in worlds where there's a premium service and there's a commodified service, and at the end of the day, like people are going to blend these things, you know, like, I'm like part of getting fable back, right, is I will heavily use it just to understand when I should and shouldn't selectively use it on an ongoing basis.
00;34;39;09 - 00;34;56;20
Aaron
Yeah. And that's why there's a debate. Like, I think it's because the answer is it's both. And I bet at some point these like hyperscalers just and there's some you know, OpenAI has a couple open source models. I don't think we've seen one yet. From anthropic. It's hard to always keep track of everything, but I think they're all going to have it right in video.
00;34;56;23 - 00;35;15;10
Aaron
You know meta. That's their strategy, whether that's effective or not. So I just think that you'll have like a whole ecosystem of these open source models along with an ecosystem, these specialized models, because it looks like you can eat more out of that, you know, frontier intelligence that way. And it's just going to be everything, like all at once.
00;35;15;10 - 00;35;26;03
Aaron
And that's why I think the router conversation is is getting some steam and why that's also interesting right now. Hard problems, but probably going to be a really, really valuable layer here.
00;35;26;05 - 00;35;30;00
Chris
Yeah. We're going from exploring to building out frontier towns.
00;35;30;00 - 00;35;53;06
Aaron
That's a great analogy. That's what I was trying to say before. Like we kind of we kind of know the shape of what this tech looks like, what it's going to be maybe to like very useful for we know roughly that we're going to get more intelligence over time. Like what the exact day or, you know, is that in a, you know, a couple quarters than a year or two year, but you kind of get a sense that this stuff's going to still continue to get more intelligent.
00;35;53;13 - 00;35;57;11
Aaron
Now you're planning it out, right? More tactically, which is pretty cool.
00;35;57;12 - 00;36;07;04
Chris
It's the Deadwood era of AI. Your your man there in Venice is the elsewhere engine of the world.
00;36;07;07 - 00;36;09;21
Aaron
That's a great analogy. Great show. I missed that show.
00;36;09;22 - 00;36;28;20
Chris
Zenith is the gem saloon. It's all the privacy. Seen that show? You got a little colorful character. Oh, it's a good one. It's a very specific mood, though. You need like, you need to be up for really grubby people talking like they're in Shakespeare using the bluest language ever.
00;36;28;22 - 00;36;36;28
Aaron
I mean, it's not a bad weekend for print. It's going to be hot. We're celebrating the fourth. I mean, that's a big part of our history. Like those frontier towns.
00;36;37;00 - 00;36;38;01
Pri
Oh my God. Yeah. We didn't.
00;36;38;01 - 00;36;39;24
Aaron
Even acknowledge the US winning the.
00;36;39;26 - 00;36;49;14
Pri
250 years, 250 years of the United States. To be honest, 250 years is like not even remotely a long time. Like was like.
00;36;49;15 - 00;36;53;18
Aaron
There's like cathedrals in Europe that took that long to build. I feel like.
00;36;53;21 - 00;37;04;27
Pri
Literally, I feel like, yeah, 250 years is like three generations, like it's like literally zero. A lot has happened, though, I will say, I got to give the US credit, like.
00;37;05;00 - 00;37;11;17
Aaron
We like to keep people on their toes. Yeah, we've done some things I think on balance, good things.
00;37;11;20 - 00;37;21;11
Chris
Yeah. Well, on that note of patriotic enthusiasm, why don't we all decamp to the the great State Fair on the national lawn?
00;37;21;13 - 00;37;24;07
Aaron
Get getting some may Chris or. No, that was last week.
00;37;24;12 - 00;37;37;07
Pri
Do you guys see that. Did you guys see. No. That was a UFC thing. But did you guys see kids like moshing at the at the the fair? Like the videos of people on Twitter like. No pits and stuff?
00;37;37;09 - 00;37;47;06
Chris
Like what? Earth, wind and fire. Who the hell is playing at the National Fair right now? The O'Jays, with no original members in it? Like, who on earth did they even manage the book?
00;37;47;09 - 00;37;54;04
Pri
No, actually, yeah. It's like, who is actually gonna go to this? It kind of looks janky, apparently.
00;37;54;07 - 00;38;13;03
Chris
Like, yeah, I don't understand why he keeps falling into this trap. Like, you know, there's so many other things he gets away with. And then he's like, you know what? I want to throw this big public event that I can't get any A-list entertainment for. That's going to, like, start to smell that. Like I still have to follow through on.
00;38;13;08 - 00;38;15;02
Pri
Yeah, he's not going to do a good job.
00;38;15;04 - 00;38;21;04
Chris
Yeah. Like you should just kind of learn at this point. Like there's things you can pull off and things you can't.
00;38;21;05 - 00;38;26;03
Pri
Yeah I agree. Like, just no one's going to hate if you don't do anything. Just don't do anything.
00;38;26;08 - 00;38;44;06
Aaron
Like you don't even need to do anything. Yeah. Like, why do you need to do that? Much like everybody's kind of it's everybody. So I mean that's what makes July 4th great, right? It's it's like inclusive if you're here or care about this place, which I know I do, then you can celebrate in so many different ways, right?
00;38;44;09 - 00;38;45;01
Aaron
I mean, I think.
00;38;45;02 - 00;38;47;29
Chris
We're getting fireworks on both rivers this year. Look at us.
00;38;48;01 - 00;38;55;25
Aaron
Right. Completely progress. Nice with our 78 degree air conditioners. Chris. Oh, yeah.
00;38;55;25 - 00;39;02;29
Pri
That was really funny. That was funny. People were really losing their shit over that. AC has been really, really the talk of.
00;39;03;01 - 00;39;12;12
Aaron
A hot topic, pun intended. Yeah, it's kind of wild, like how many AC tweets I've had to consume. It's just like it's a great technology.
00;39;12;13 - 00;39;15;26
Pri
I keep liking that. I keep liking him. So I'm getting more.
00;39;15;28 - 00;39;36;03
Aaron
I see your whole feed is just mosh pit and the AC debate, that's like not even a debate. Like, if you can, you can swing it. And I understand why. Maybe your electrical grid can't handle it. Like you should get an AC. The one I like the best was the troll one from some like Chinese pig farmer who is showing like how his his like puts it heard.
00;39;36;05 - 00;39;42;00
Chris
Like a split split unit running in like a two wall pig. Pig shed.
00;39;42;03 - 00;39;52;18
Aaron
Yeah, like like pig pen. I guess it would be called. Right? So. Yeah. And he's just like, yeah, like even our picks have air conditioning. I'm like, oh my God, this is really, really devolving faster than I can imagine.
00;39;52;19 - 00;40;12;24
Pri
No, no, I mean, how about like, the Germany like at the German hospital, people wearing ice vests to cool down, like at a hospital? I don't even have AC. Like, that's straight up like the mental illness. Just. It's so deep, like it's crazy on on AC and like, I'm just like, wow, you guys are. There's like a danger to your people.
00;40;12;26 - 00;40;38;08
Chris
There are all sorts of spectrums of sovereignty. You cannot escape the fact that it's hot out, but depending on where you are, you have remediation available to you or not available to you. Yeah, some of these stories they've run about how, you know, if you want to get AC, you can, but then you need to petition everyone in your building and only if they all approve it, then you can bring it to your local governing board for approval.
00;40;38;12 - 00;40;41;06
Chris
It's just like, wow, man, I don't know.
00;40;41;07 - 00;40;49;09
Aaron
Yeah, maybe, you know, maybe we've just been AC opt Chris in the States just ran it so much.
00;40;49;11 - 00;41;02;21
Chris
I am in a situation where for me to install splits in my house, I do have to run it through a community board. Like, I can't just, like, put external splits in my house and it's driving me nuts. Yeah.
00;41;02;27 - 00;41;03;26
Aaron
Yeah.
00;41;03;28 - 00;41;05;19
Chris
I'm in a historic district.
00;41;05;21 - 00;41;20;11
Aaron
Am I wrong? The like AC probably. Don't you think AC would probably be a top ten all time invention? It's definitely not like in the top five, but probably top ten. Like they know whole regions of the planet couldn't really, you know, develop.
00;41;20;14 - 00;41;20;21
Pri
I don't.
00;41;20;21 - 00;41;21;27
Aaron
Know of the heat.
00;41;21;28 - 00;41;22;18
Pri
Pop ten.
00;41;22;19 - 00;41;24;25
Aaron
It's Aaron there. You got ten.
00;41;24;27 - 00;41;37;20
Chris
Dude look electricity the internet like I mean I think maybe a top. So I probably I think when you get down to it, it's probably top 50 ish.
00;41;37;23 - 00;41;39;06
Aaron
No, that's too low, dude.
00;41;39;06 - 00;41;47;14
Chris
Dude. All right, time out like a mirror. I don't know, 100 years ago, the average lifespan was a third of what it is today.
00;41;47;16 - 00;41;49;11
Aaron
Yeah, completely, I get it. I'm just saying.
00;41;49;11 - 00;41;53;08
Chris
Like, there's a lot of things that are going to crowd AC out. I think AC is a top 50.
00;41;53;09 - 00;41;58;26
Aaron
Maybe. Maybe we'll we'll meet in the middle. Top 25. Top 25 I think is a is.
00;41;59;00 - 00;42;01;06
Pri
The pretty fair. What are the top.
00;42;01;07 - 00;42;04;02
Aaron
Yeah. As far as fable.
00;42;04;08 - 00;42;10;04
Chris
I think this is our next podcast. We're just going to argue until we come up with the top 25.
00;42;10;06 - 00;42;13;27
Aaron
I mean yeah, I'll give you the one argument why I'd put it.
00;42;13;27 - 00;42;16;15
Chris
Up there. I would say refrigeration over air conditioning.
00;42;16;15 - 00;42;21;04
Aaron
Well, yeah, I mean, it's kind of like this. They're kind of the same thing, right?
00;42;21;05 - 00;42;27;24
Chris
No, I mean, sure, like at some level there's a condenser and yes, I get.
00;42;27;24 - 00;42;31;22
Aaron
You there. The same family oftentimes, you know, similar.
00;42;31;23 - 00;42;35;00
Chris
Indoor plumbing. I'm going to take indoor plumbing over your conditioning.
00;42;35;01 - 00;42;45;01
Aaron
Yeah. Yeah. I'm there's no disparaging of the other ones. I'm just saying it's probably up there just in terms of like things that really, you know, power our world today.
00;42;45;02 - 00;42;50;04
Chris
I mean, I would definitely take care of conditioning over the Beatles. Right? So they've nudged the Beatles out.
00;42;50;06 - 00;42;51;04
Aaron
I think that's fair.
00;42;51;06 - 00;42;59;03
Chris
I do I do kind of want to make this list now and figure out where I'm going to place air conditioning. I'm still sticking with my press.
00;42;59;04 - 00;43;00;12
Aaron
Printing press is pretty high.
00;43;00;13 - 00;43;01;24
Pri
So do you want me to ask?
00;43;01;27 - 00;43;04;04
Aaron
There's some people put duct tape.
00;43;04;05 - 00;43;10;01
Pri
So the wheel. The wheel is number one. The wheel is number one. Two is writing.
00;43;10;02 - 00;43;11;01
Aaron
That's a good one.
00;43;11;02 - 00;43;42;05
Pri
Three. Three is a printing press. Then you got fourth electricity generation and distribution. Yep. Fifth is the internet. Six antibiotics, seven vaccines, eight steam engine nine agriculture. Slash the plow ten the transistor 11 the computer 12. Sanitation systems 13. The telephone 14. Automobile 15. Airplanes 16. Refrigeration 17. Compass 18. Optical lenses 19. Paper 20. Gunpowder 21. Light bulb 22.
00;43;42;05 - 00;44;03;06
Pri
Anesthesia 23. The Hard Bosch process 24. Semiconductor chips. Integrated circuits 25. The camera. And then I was like, well, where is Eric? Conditioning. It's. And then it. A it's a fair catch. It said fair catch air conditioning has a legitimate claim to the spot. And then it said it would give it somewhere between 18 to 25 if it had to.
00;44;03;07 - 00;44;05;19
Aaron
Probably probably. Right. We got to the answer.
00;44;05;24 - 00;44;06;24
Pri
Or the light bulb.
00;44;06;25 - 00;44;12;05
Chris
So she totally spiked Claude by asking, what about air conditioning?
00;44;12;05 - 00;44;18;19
Aaron
It's over 25. I'm sticking with it. I'm sticking with the top 25. I take a I take over a camera any day of the week.
00;44;18;20 - 00;44;20;17
Pri
I do think that it's probably more.
00;44;20;21 - 00;44;24;09
Aaron
If we had no camera, we'd be fine. Maybe we'd be in a better spot. To be honest.
00;44;24;11 - 00;44;27;10
Chris
Apparently television is getting the shaft here.
00;44;27;12 - 00;44;31;22
Aaron
Yeah, I mean, I think she's air conditioning over TV.
00;44;31;25 - 00;44;35;19
Chris
No. Are you kidding me? I'm, like, only hot, like, ten days a year.
00;44;35;20 - 00;44;38;17
Aaron
Well, you're a special guest. We know Chris, dude.
00;44;38;18 - 00;44;46;01
Pri
AC, like like not having AC and being in heat, like, lowers IQ. Like, if we didn't have AC, we might not be able to invent other things.
00;44;46;03 - 00;44;49;24
Chris
Why do you think I worked in my basement? It's cooler down here.
00;44;49;26 - 00;44;53;16
Pri
Yeah, right. Helps. You think it helps you just properly?
00;44;53;17 - 00;45;13;15
Aaron
There's a lot of when you really drill down into it. AC has, like, a lot of knock on benefits. Like, I think like just pointing at America, like, like parts of the country probably wouldn't be settled as much like a huge impact on people. Like it makes certain areas that are unlivable more livable. Right. And think about the people who are putting up all these death statistics, which are super sad.
00;45;13;16 - 00;45;17;06
Aaron
And if those are true, I don't know if they are or not. I mean, that's that's a big deal.
00;45;17;07 - 00;45;23;29
Chris
I ain't against Air-Conditioning, dude. I er, conditioning is wonderful. I'm just saying there's a lot of more wonderful things there.
00;45;24;01 - 00;45;25;25
Pri
Yeah. Fair enough. Okay.
00;45;25;27 - 00;45;29;28
Chris
Hell of an episode. Guys. This is what you get on the on the holiday weekend.
00;45;30;01 - 00;45;32;08
Aaron
As you should. Happy fourth guys.
00;45;32;09 - 00;45;48;03
Pri
I'll introduce the pod. Hey, everyone. Today it's me, Chris and Aaron talking all things crypto, AI, tech, culture and more. Just as a quick reminder, these thoughts on paying our own and not of our employer and none of this is financial advice.