I don’t think the bauhaus one is horrible, nor is the cubist one. They’re not great works of art, but if I saw one out in the wild, my reaction would not be: “bloody ai art making everything visually putrid again” nor, do I think they’re worse than the clip art and badly taken digital photograph agglomerations they would be replacing.
> Everything therefore becomes a hobby or a game. People do things because they enjoy them for their own sake, or because they are endeavours used as vehicles to socialise and enjoy others’ company, or because a shared social belief exists and is cultivated such that doing such and such a thing confers social status.
There is actually a big leap in logic you’re doing here that it seems you are naive to. The world looks the way it does today because human societies, businesses and individuals are fundamentally in competition with one another for scarce resources, and having more resources leads to more capacity to survive, so there is always an incentive to acquire more. If you can no longer compete, you can and will be exploited. You don’t have to look very hard, far or back in time to realize that this is what society looks like. If everyone is incapable of competing except those who can afford to run a datacenter, that will not end up in socialist utopia.
Quantization is a category error, the thing you care about is not in weight space, so there’s unbounded error introduced by doing it. The thing you actually want to preserve is the knowledge manifold, but that is in a different vector space. Until we have some better understanding of how to interact with that space directly, rather than inferring it through distillation of reasoning traces, I would not anticipate truly low bit models to be useful.
That seems like a pointedly naive viewpoint. You can have clearly predictive population level statistics about prevalence in family groups without clearly understanding the (very messy and unbelievably complicated) way that genetics shapes brain development.
I’m in two minds about this, in some sense they did and it’s the same way in which the C code I write is mine, but the assembly underneath is an artifact of my intent, but I owe the compiler authors for it. On the other hand the cognitive distance between my C and the assembly is likely lower, but that’s because I’ve spent time staring at what was generated in order to figure out why my code was misbehaving. Which is itself not that dissimilar to figuring out how to get better at writing code with llms.
If you ask LLM to code something for you, and you then work on the code, then yes, there is partial ownership. Just like if you had asked another, maybe junior, developer to write a piece of code, and then you helped them make it better.
I think the difference involves how many layers of stuff someone claims is "theirs" and whether that claim of credit and expertise is valid.
Imagine a coworker writes some decent C code, but then at a meeting they start taking credit for the bytecode and the default optimizations done by the compiler.
That's analogous to people who write LLM prompts and then claim the same level of authorship over the code which was all generated and edited on an indirect level.
There’s multiple databases of games in algebraic notation. You can also, very easily rl train on pitting models against one another, even without mcts.
using a 23mm auto cannon on orbit is more or less useless. There’s no way to reasonably shoot something when your relative velocities could easily be strictly greater than the bullets. Even if something is coming straight towards you it’s very difficult to aim when you have to take into account orbital mechanics. You would need a huge amount of RCS fuel in order to counteract the impulse from the gun. And even if you did shoot the gun if you miss, you now have a bunch of bullets that share your orbit and will at some point intercept your flight path again, if you don’t miss, you still have shrapnel sharing your orbital path.
reply