Nuclear fusion can’t produce net energy, so is it really a solution to AI’s growing energy demands?
nuclear power produces long-lived radioactive waste, which needs to be stored securely. Nuclear fuels, such as the element uranium (which needs to be mined), are finite, so the technology is not considered renewable. Renewable sources of energy, such as solar and wind power suffer from “intermittency”, meaning they do not consistently produce energy at all hours of the day.
fusion technologies have yet to produce sustained net energy output (more energy than is put in to run the reactor), let alone produce energy at the scale required to meet the growing demands of AI. Fusion will require many more technological developments before it can fulfil its promise of delivering power to the grid.
Maybe AI can help us break the fusion hurdles. Oh. It's still telling people to eat rocks, just used to create waifu porn and as a mass spy application? Nothing else, really? Well shit.
I know you're being reflexively downvoted by who hate everything AI, but this is the sort of thing AI should be most useful for, which is finding patterns within large problem spaces with many variables.
Probably. Proxima fusion is using simulation-driven engineering to pave their way through the design space, no matter how you approach it it's gotta involve dimension reduction in some way and that's ML. They speak of AI but well it's a press piece.
LLMs or diffusion models? Nah, don't think so. This is actual engineers throwing statistics at a particular problem to identify what prototypes they should build, not techbros throwing shit at the wall.
It's even bad at porn. Very limited means of describing the process, forgetting that there are no bed sheets in a park, same repeating metaphors. Boring.
The most time saving for a single task that I had so far was when repairing a Ebike battery that needed a new BMS. Finding the cause and fixing it was only possible because GPT swallowed the relevant sections of the datasheet and spit out code to read/write to the chip. That would have taken dozens of hours for me to understand what does what and under which condition and then make code out of it too. I put it on GitHub so others do not need to suffer.