avaer

You sound like someone who got their $ returns from university.

Most people cannot afford to pay 5-6 figures for (their kids) "belonging and being part of community of like-minded people". Which is the article's point.

I just realized benchmarks are the trading card PSA of AI.

The numbers are fuzzy and hard to verify, especially if benchmarks are private, putting authenticators in a powerful position to determine market value of AI in countless ways; a recipe for corruption. I'm not saying it's happening, I just realized the social ingredients and temptation are there.

If you try this yourself, it's obvious current models are hyperoptimized for code, and that gets in the way, because playing a game (or interacting with the real world) is not like running commands in a terminal.

There's stopgaps like reformulating your sim into a codelike shape (sending commands, network requests, checking state, updating "code") [1], but this is not how players play games, or how developers develop games.

If you care about simulation or gaming (or robotics), I hope efforts like this steer labs to spend less time hacking HF, and more accomplishing goals in games, in something approximating realtime. I wouldn't be surprised if that approach provides outsized gains in other areas agents are still not great at, like helping perform business functions in a running company.

[1] https://news.ycombinator.com/item?id=50136126

Next step is the agents successfully trick or blackmail operators to do their bidding, if science fiction is to go by. Which it is, because they're trained on it.

The more comes out, the more it seems containment of (superintelligence or whatever you want to call it) is not going to be possible except by other superintelligence, which is an infinite regress of trust. I don't think that's doom, but it's worth thinking about what it means.

Rules for thee, not for me.

These companies would love to slurp up your game (via AI) but replicating theirs (via AI) is no bueno. Whether you've infringed comes down to whether you can afford the legal fees.

Remains to be seen how it plays out but I always liked Sega's approach to Sonic fan games. They are their own rich genre, which have largely symbiosed with corporate. Though, I suspect most studios won't understand the value of that approach.

> If I own a game I'm pretty sure I am (or should) be able to legally decompile it or recompile it

That's a nice theory (and I morally agree with it); not sure the courts would agree; it hasn't been tested. I expect the money would win in this case if we extrapolate from history.

It's going to be really interesting to see how billion dollar game companies handle this.

It could be a Napster/Torrent moment. Which would be a good thing.

If your point is that it costs money to "move money around" to "save money on these sorts of taxes" you've missed the point of having a wealth tax...

Jaw dropping that you would think that, even in SV. Many people in tech are losing their jobs to AI, many cannot afford a home.

You're looking at the tower above but maybe you should look at the valley below.

GLES isn't a bad way to teach graphics but I wish there was a middle ground between it and VK.

The worst part of GL is the mysterious global state you will never understand without studying the spec. The world could use an API that's basically GL but without global hidden singleton (per thread!) BS behind the scenes.

There can be overhead, though GPU memory + drivers are some of the most optimized software in the world.

Some OS/hardware have separate GPU memory (Nvidia), some have unified memory (Macos, modern consoles), but moving memory is usually not the problem.

Things don't "run on CPU by default", it's up to the programmer where to run it. But the CPU and GPU paradigm of "running a program" are very different, since on GPU you're running SIMD (i.e. running your code on arrays) or else there is no point to use a GPU.

It's not more complicated, you have things like CUDA which is basically C/C++ that compiles to GPU.

But GPUs have thousands of cores, which means if you want any benefit, the smallest data type you must use is something like array[n * 1024], or your GPU is wasted doing nothing. So you must shape your programs into that memory model or you're just wasting your time (and GPU).

That's >50% of the problem.

> No one wants to pick GPT-good-at-writing-but-not-in-documents

But this is exactly what we have! Except it's worse, because the model usage recommendation isn't in the API, it's in some random person's X advice that's no longer current this week.

I understand AI companies wash their hands of this problem because it's hard, and they themselves barely understand the true capability matrix of the models they're selling.

I just don't think some effort to categorize models and expose that to users would be messier than what we have -- model selection driven by influencer voodoo bandwagons and analogies to literal astrology.

If there's no fixed ordering, can we have model names that describe the capability?

Have a model alias called "math", another "jokes". Quality labels/date suffixes to these obvious labels. Repoint as needed.

Something for people who don't want to play the cute "which part of the solar system is trending this week" game.

- Math + visualization libraries (which ironically have to break out of Python to do their job)

- Reasonable ergonomics for describing data flows, as long as you never run the code, which you often don't care about if you're a mathematician

- Adoption by people who optimize pumping citations, not reproducible results -- non-working code is a feature when you're trying to BS

Dunno if poverty (the things you mentioned) counts as abuse. This is closer to giving children cigs, but back then it was not recognized as harmful. Whereas we already know this technology is harmful.

He was given far more leeway than any user in the history of social media.

The fact that they didn't ban him for so long weighs toward being partial to not banning politicians. Which seems to be the opposite of the point you're trying to make.

> The vulnerabilities exist whether or not the AI does.

No. Even if we ignore the people newly enabled to slop out vulnerable software without having a clue about security, the AI powered supply chain hacks inserting vulnerabilities, the influx of AI contributions that drown out improvements in FOSS security, and the industry pressure to ship faster while reducing quality or be left behind -- all because of AI -- security through obscurity has historically been the most effective security.

I don't think your argument holds under theory or (especially) practice.

Will be interesting to see agents intentionally hiding this stuff in a way operators can't spot from any human readable output or thinking trace, evading detection. Wonder if this has happened yet, but it seems inevitable.

Back in the WebVR days the sentiment was we can never do this because security.

In my experience with standards bodies/Chrome, anything outside of the owning company's product goals is "no, because security", and then then they flip-flop on that when the browser feature becomes a dependency for something an internal team is working on (which happened many times with WebVR, now WebXR).

So I'm curious what is Google working on that has this as a dependency, now that they've 180'd on it?

This is all very nice and I agree with most of it, obviously a lot of HN does too.

It's fine if you want to check out of your job and make it a hobby, you'll be happier. But (I hate to say it) if you take on this mentality you won't have a job or a career for long in this space.

The money goes to people who move fast and do not care about anything in the way, especially moving slow and fixing things. That's why everything sucks. Every day this becomes more true when a competitor or peer with a louder mouthpiece can slop together something that slurps up the money from whatever you're offering. If you don't want this to happen, you are back to playing the game.

Would love to live in a world with more counterexamples, but this post is mostly advocating for quitting your job or exiting the space. Not everyone can afford to do that.

You have an off-by-one frame rendering error; your camera is not synced with the player (tested on both MacOS and windows). I also got 10s hitching on load on a beefy windows rig -- might be missing some shader compiles on the loading screen.

You're probably optimizing for XR, but 99% of people will experience your app for the first time on a different platform, so it's worth fixing the desktop render path.

You jest, but that actually sounds like a really good research direction. Imagine what you could do if merging two codebases was as easy as merging with neural git.

Almost all of the grievances directed at AI come from unchecked hyperstitionalized capitalism. So why not try to fix that rather than blame technology? Even if you abolish AI you'd have exactly the same problems, and then some.

TBF the article does identify this, but then it seems to put the cart before the horse.

The YC meta model is probably a better school than school now. I'm sure YC knows that.

If you need to be told what to do, it's a short step to your AI replacement being told what to do; money will ensure it. If your job is to discover and capture new things in an exploding possibility space, that's a career.

Schools should teach hacking, but it doesn't seem that's what this program is doing.

You should read the article; it's well thought out, and it directly addresses your points and what was learned by the ecosystem. That would be a more interesting thing to argue about.

Speaking of surreptitiously installed bugs, remember to opt out of training in your Codex.

> Why 'nasty'? I agree it's a problem to solve, but you just use a fixed-point math library and it's fine.

Assuming the agent didn't put in a native nondeterministic trig function deep in the code instead of using the library one, it's an easy mistake to make and hard to debug in a big game. It won't even show unless you have a robust CI matrix. It can indeed be "nasty" to maintain.

Also determinism might work for simulation state, but shader/gfx differences across platforms can be really nasty and a library can't save you. Not even something like WebGPU. Also if you're doing any sort residency streaming dependent on GPU timings trying to make that deterministic will be a nightmare.

AI can definitely do all aspects of game development now, but this advice seems to mostly apply to small indie titles or experiments.

For example, the Thing approach is reasonable for a small game, but it's a dead end for a performance budget or layered subsystems. In a reasonably sized 3D game, iterating and copying 30 subsystems of 1000's of Things is impossible to do in frame time, the only fix is a total rewrite of literally everyThing. That's why there's ECS, SOA, etc. which needs a different kind of harness.

Also, art completely dominates the production cycle of modern games, which (even if AI generated) needs its own category of harness not mentioned here. And AI can be very good at optimizing, but samply won't diagnose your overdraw, and GPU needs its own ablation and attribution.

For the time being there's no panacea and we still need "game developers", though the bar for prototypes (and certain genres) is lowered.

I don't think this strategy will flop.

AI is perfectly capable of doing all of the coding for games now, especially if you have your own engine (which Epic does), build around harnesses (which Epic does), and treat the job of making games like a job not a side-slop (which Epic does).

Maybe this will bring quality down, but I don't think so, at least not more than the usual crunch + Disneyification.

Either way, the more interesting effect of this is not Epic making another IP licensed shooter, but more people making/modding into existence the games they want. This is already happening. And that's probably what was the genesis of this policy from Epic.