15,890 karma · joined January 7, 2022
Basically I think what we're seeing is that LLMs make everything easy, and so the standard for success is going to increase somewhat significantly simply because there are no longer any barriers whatsoever to doing things.
And in this case it gets even weirder. The alleged Chinese spy is the head of a Singaporean company. The claim is that NVidia sold chips to Singapore, and that company's later actions violated US sanctions on China. Trying to pin this on Jensen is nonsensical. So I agree with the above poster that the reason he was ousted was probably behind-the-scenes intimidation of the NVidia board, because the US can certainly make life very difficult for them, even if no criminal acts can be proven.
If you sold a GPU to a Singaporean company, suffice to say you'd be fine - even if it was at the beckoning of a cute Chinese girl.
I think there are things about 'real world' development that people from a software background tend to not have a good intuition for. An analog I'm drawn to is how annoying it can sometimes be to put together a somewhat more complex desk, chair, or whatever else from some place. In that case you have everything perfectly hand-crafted to fit together and be assembled by any old idiot, yet it can still somehow be tricky sometimes. And obviously doing 'real things' in the real world is going to be many orders of magnitude trickier. No idea if that's the issue, but it's at least my mental model for now.
So even if the government sets out to surpass you, it's quite likely to just turn into a black hole of money. See: CHIPS Act leading to many billions spent to... remain completely dependent upon Taiwan. But all of this changes when you create a scenario where domestic players are the only players in the market.
So, in other words - one can both be fully on-board with the LLM future, and also think that this IPO is a clown show.
Anthropic is nothing but LLM, so I expect much bigger fireworks here. I don't know what that means, yet - but I just refuse to believe that there's many people that are genuinely long on Anthropic. It's just such an illogical bet. If their IPO can still be cancelled, that would be my bet. If it can't then popcorn awaits.
For another example, the reason Chinese are not allowed on the 'international' ISS is because of the Wolf Amendment. [1] The idea was that if we just allowed them onto the ISS they'd steal all of our elite tech or something. Within a decade they had launched their own space station and carried out various revolutionary space activities.
And of course all of that happened while the ISS itself continues to deteriorate with no replacement in sight. And when China did an unprecedented sample return mission from the far side of the Moon and offered the whole world access to samples, US researchers had to jump through a bunch of hoops to receive them, again thanks to the same law.
It's just myopia.
LLMs will be an interesting test of the current broken systems at most large companies. You can literally fire your entire engineering team and replace them with your "really smart cousin" and Claude. And that's going to be mighty tempting when the MBA mantra seems to be that cost reductions this quarter are infinitely more important than any potential costs in the future. But of course they're not.
It's quite interesting to appreciate that these issues were observed, and written about, some 2400 years ago. Times when ostensibly everything would have been practically alien. Yet it emphasizes that the thing that's really changing is not humanity, so much as it is technology. We're still trying, and failing, to deal with the exact same issues as we were thousands of years ago.
And mind you that's just the first time it was written down. I wouldn't doubt that these observations and issues long predate even Plato. Because I think one thing that's become clear is that he wasn't necessarily unimaginably brilliant so much as he was simply capable of objectively seeing the patterns in society and recording them. Though perhaps that is some form of brilliance in its own right.
In reality, those super farms did come to pass, yet their share of the economy began to approach 0. And that was fully predictable! Because when the price of producing food plummets, the relative value of food also plummets. The exact same thing will happen with cognitive work. The winners will not be Anthropic/Amazon/etc. As the cost of cognitive work approaches $0, the relative value of cognitive work will also approach $0.
So this is going to completely revolutionize the economic equilibrium in society. Trying to translate what this means right now is like trying to predict the impact of the Industrial Revolution from somebody who spent their entire life living in agrarian society - in other words, probably impossible. But in general the winners will not be those providing cognitive work, but those building currently unimaginable things on top of it.
Churchill: Madam, would you sleep with me for $1 million?
Woman: Why Winston, yes I would.
Churchill: How about $10?
Woman: What sort of woman do you think I am?
Churchill: We've already established that. Now we're just negotiating the price.
And I think this is all a logical consequence of evolution, which is driven solely by fertility. A woman can only carry one child at a time and so it makes sense for them to seek a highly qualified/'resourceful' partner to have the most chance of her genes being successfully carried forward, whereas a man can theoretically father hundreds of children in a single year, so it makes sense for him to be more 'driven.'And beyond this you're also speaking of something different. You're talking about taking some output assembly and then modifying it in 'negligible' (as a percent of the whole) ways, as opposed to writing assembly oneself. For a program of minimal complexity, not only would it would require an unreasonable amount of effort, but you'd also invariably write inefficient assembly relative to a compiler. Compilers produce code that is better, faster, and cheaper.
---
Shifting outside the abstract, let's just do an example. Imagine you have a processor running at 3 billion cycles/s and you have a program that takes up to 1 million single cycle ops/s. And then some ultra clever human tunes a bit of the assembly and we get it down to 0.95m cycles per second. In reality you're probably not going to be able to tune your way to a generalizable 5% boost, but even if you do in this scenario - what have you achieved? Literally nothing. Of course this argument, taken to extremes, is why it now takes more processing power to run a text editor than it did to put a man on the Moon, but we're not talking about extremes.
A metaphor I like to use is somebody who's lived their entire life in an agrarian society seeing the Industrial Revolution imminently coming and trying to predict its effects. They'd be able to accurately predict 'super farms', but the relative economic value of those super farms rapidly trended towards $0 with the vacuum that the revolution created. It'll be the same thing here. The winners won't be people selling cognitive work, but the people building previously unimaginable things on top of that.
It's probably impossible to predict exactly what this means from here, in exactly the same way as somebody in our agrarian society could scarcely even begin to imagine what the Industrial Revolution would 'truly' entail. But I think a naive prediction would see a sharply increased value for stuff done in the real world, social type work, entertainment, and so forth. So one prediction would be that one big winner will be whoever can create cheap mass produced robotic shells to help bring the cognitive revolution to the real world. Even that will probably be seen as naive, but it's probably closer to the mark than expecting Amazon/Google/Anthropic/etc to end up ruling the world.
But when making a combat game there's a million subtle things going on that the game gets very right to actually make it feel like the hits have weight - the flashes, text, and other aspects all have subtle animations in them that make them feel good the hitbox triggers at exactly the right moment meaning it's a tight fit, and much more. These are things many human studios screw up. It's really hard to make combat feel good, and this game does it.
For some context, Streets of Rage 4 [1] was a very well received beat-em-up. So that's pretty much the top tier for the genre. And yeah it's obviously better than this game, but the gap is not huge, and that's borderline absurd when considering that this was a one-shot LLM generated game. And that gap could easily be closed with some refinement from a single developer - in this case our rando uncle. Interesting times ahead.
Nothing revolutionary, but a bunch of different parts all working together well that are generally implemented at a very good to excellent level. That's way better than most can say. Of course the walking animation, like the one guy mentioned is hilariously bad, but fixing that's also just a prompt away. Oddly enough I didn't even notice it when playing.
I just mention this because I think what describing something basically 100% of people experience, and it completely ruins gaming for many - or they think it's because they're an adult or whatever. When we were young and had limited access to games, everything was amazing. Then you have a Steam library with 800 hand-picked games, many of which you've never even launched, and you're like 'mehhhh, nothing to play.' It's just FOMO.
Of course this means you'll literally never get through your library for most of us, but what can you do? If our younger selves could only imagine us trying to deal with such brutal problems.
Games are subject to taste, but I think there are also objective elements. And this game just objectively feels and plays great, which is a remarkable achievement for an LLM, because it's a genre where it's easy to get things that look right but don't feel right. For instance Sacred Citadel is a game that was made by a studio with over a decade of experience, published by a publisher with decades of experience, and it's worse than this game - much worse. It got the look right, but it felt/played terrible.
The fact that a single guy can casually one-shot games dramatically better than what passed for a [relatively] big studio title a decade ago is testament to where we're at. And I think that's freaking some people out. I suppose that freak out is reasonable from one perspective, but from another it's just shaking your fist at the clouds. Even if all LLM advances stopped today, what we've already unleashed is already way more than enough to change the game forever, and things are never going back to how they were. And, of course, we're not stopping today.