this post was submitted on 15 Oct 2024
665 points (98.7% liked)
Technology
59627 readers
2911 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Crazy how quickly we've gone from "Nuclear is a dead technology, it can't work and its simply too expensive to build more of. Y'all have to use fossil fuels instead" to "We're building nuclear plants as quickly as our contractors can draft them, but only for doing experiments in high end algorithmic brute-forcing".
Would be nice if some of that dirt-cheap, low-emission, industrial capacity electricity was available for the rest of us.
One of the things with AI is that it's a largely constant load factor. Nuclear is really good for that.
However, I highly doubt any of these new nuclear plants are finished before the AI bubble bursts. SMRs haven't even been proven in practice yet, and this is the first good news they've had in a while. Restarting Three Mile Island isn't expected to work before 2028. The hype bubble could easily burst in the next year, and even if it doesn't, keeping it going to 2028 is highly unlikely.
So we'll probably have some new nuclear around that isn't going into AI, because those datacenters will be dead when the hype passes. Might as well use them, I guess.
You don't think there's any chance that AI as it exists today might be just the start of a huge industry?
As it exists now, no. The models are reaching their limit, and they aren't good enough. They can't absorb any more information than they have, and more training iterations aren't making them better. They'll do some useful things; a recent find of the longest black hole jet ever found was done in part from AI classification of astronomy data. It's going to get implemented into existing tools and that's about it. It won't be enough to justify the money that's already been dumped in.
Historically, the field has been very bursty. Lots of money gets dumped into it, it makes some big improvements, and then hits a wall. Funding dries up because it's not meeting goals anymore, and the whole thing goes into slumber for a decade or two. A new breakthrough eventually comes, and then money gets dumped in again. We've about maxed out what the last breakthrough can give us. I expect we'll need at least one more cycle of this before AGI works out.