Historically, AI has found and used exploits. Before OpenAI was known for chatgpt, they did a lot of work in reinforcement learning (often deployed in game-like scenarios). One of the more mainstream training strategies (pioneered at OpenAI) played sonic and would exploit bugs in the game, for example.

The compute used for these strategies are pretty high though. Even crafting a diamond in Minecraft can require playing for hundreds of millions of steps, and even then, AI might not constantly reach their goal. Theres still interesting work in the space, but sadly LLMs have sucked up a lot of the R&D resources.

The challenge is that AI for a video game (even one fixed game) is very problem specific and there’s no generalized approach/kit for developing AI for games. So while there’s research showing AI can play games, it’s involved lots of iteration and AI expertise. Thats obviously a large barrier for any video game and that doesn’t even touch the compute requirements.

There’s also the problem of making AI players fun. Too easy and they’re boring, too hard and they’re frustrating. Expert level AI can perform at expert level, which wouldn’t be fun for the average player. Striking the right difficulty balance isn’t easy or obvious.

Late to reply but the issue was fixed.

Just sent my logs, thanks

Doesn’t work for me at all on Ubuntu, but nice to see they’re finally getting a Linux client out.

That applies to the software itself, sure, but only if you bring your own infrastructure. Large scale FOSS infrastructure services are going to be the exception not the norm.

Nah I’d disagree. Infinite growth motive doesn’t necessarily apply to private companies. To suggest there’s unbridled greed present in every company is just a falsehood.

If you’re not paying for a service, then you’re the product. I never understood the expectation that people should just provide you email and storage for free, because?

Yeah I’m sure shell and co really forced the sale of 750k F-series trucks last year, right?

I would agree with this, I use Nvidia cards for professional work on Linux and I’ve never had a problem. Yeah there’s some upfront work configuring the drivers, but I’ve never had it take more than an hour to setup.

YouTube already does that, though? In fact your premium sub helps creators much more than a non-premium viewer does. Of course the multiplier proposition would be new/different.

Not trying to shill for YouTube but that’s one of the main reasons I have a subscription is to support creators I like.

I’m not really convinced that a GPU backend is needed. Was there ever a comparison of the different CLIP model variants? Or a graph optimized / quantized ONNX version?

I think the proposed solution makes a lot of sense for the task at hand if it were integrated on the pic-rs end, but it would be worth investigating further improvements if it were on the lemmy server end.

Reprocessing already exists and it’s been done for decades. I can’t imagine reprocessing fuel for recycling the usable components is that compelling in the US and it would be more geared to waste reduction. 99% of spent fuel by mass could be reused or otherwise treated differently for disposal as it’s radioactivity is much much smaller than the portion that has been transmuted during power production.

This is really only one facet and not even the main driver in cost. MIT did a study a few years ago looking at this (https://news.mit.edu/2020/reasons-nuclear-overruns-1118). Turns out it’s complicated.

In short, in the US, lacked of skilled labor and large scale project management are big drivers also, not just regulations.

That’s a limitation of the secondary power conversion side and is true for any power generation methodology that relies on steam generation. That said, there’s alternatives to the traditional Rankine cycle that could be deployed without modifying the nuclear side of the plant.