Hacker News Reader

Comment by typon | Hacker News Reader

Imagine if the world's engineering talent was focused on this rather than making AI to generate slop?

As a secondary effect it kind of is; the general assumption still is that the slop-generating AI will need a lot of power to train, so there is surprisingly a lot more private investment into fusion and fission innovation in recent years.

Well, AI also has something else for it : at this point, no one is expecting any ROI soon, but they all imagine that it's going to be huuuuge, so the "expected" (as in, "wished for") ROI might as well be infinite.

As soon as AI investors start demanding dividends, then the ROI of investing in AI will be compared to the ROI of investing in electricity production "for production sake".

Even if we shut down chatgpt, people who still switch light on.

If we only keep enough fusion reactors to run LLM inferences, but no one can afford lights, well...

What value do you think Software Engineers focused on AI would provide to Fusion Research?

Why is the entire planet incapable of walking and chewing gum at the same time?

Maybe the AI trained by the best engineers will help other people get into physics and then study nuclear fusion.

Why would comp sci majors be working on plasma? Prob want physicists working on that stuff.

The need to process the large amount of data gathered by the fusion test rig drove many fields of computer science. The freaking internet was supposedly invented because the physicists at CERN needed a way to collaboratively process and interpret this data.

Anthropic likes to hire physicists.

Also: https://www.energy.gov/science/fes/articles/ai-tackles-disru...

There's a good chance this is not an either/or question, but that AI will get us to viable fusion power much earlier. For example:

https://www.theregister.com/2022/12/23/doe_fusion_ai/

https://openai.com/index/strengthening-americas-ai-leadershi...

Reply on news.ycombinator.com