• 0 Posts
  • 355 Comments
Joined 2 years ago
cake
Cake day: July 7th, 2023

help-circle



  • Yep, thats the key issue that so many people fail to understand. They want AI to be deterministic but it simply isnt. Its like expecting a human to get the right answer to any possible question, its just not going to happen. The only thing we can do is bring error rates with ai lower than a human doing the same task, and it will be at that point that the ai becomes useful. But even at that point there will always be the alignment issue and nondeterminism, meaning ai will never behave exactly the way we want or expect it to.





  • When discussing things like this, i find it best to view it through the lens of ubi. AI is happening, theres no amount of online debating thats going to change that. So if you dont create an economic system where people dont have to work to survive, people arent going to survive in your system. Hence, they have a point. People that WANT to get into music always have that option, untied to economic success. People that only make music as a cash grab now no longer have that incentive, because they dont need to.

    People having to work to survive is barbaric









  • Yeah, thats kinda what im getting at. Currently, behind fossil fuels, the most polluting/environmentally unfriendly industry im aware of is ai development.

    Full controversial disclosure: ive been studying actual AI since the early 00s. I know what the tech is capable of and im very well versed in it. Llms are a nichely useful tool, but thats it. That is not even close to the max potential for AI. Most of the things the tech bros promise are theoretically possible, but theyre so wildly misleading about it as to basically say that they are just lying. The current AI bubble WILL pop before we see too much cooler stuff, but I am confident that the tech will still be looked into, just at a regular scale instead of being massively blown out of proportion. Eventually it will be like it used to be, with small incremental inprovements in multiple different fields and the occasional breakthrough where someone combines the fields. And this should all be happening without consumers being aware of it, basically. Not that is shiuld be hidden, but they shouldnt be writing news articles about it because real progress is slow. LLMs absorb a shitload of energy and require extreme cooling, but we know thats not a requirement for human level thought because… you dont require any of that.

    If you want to actually keep up with the real progress in actual AI instead of tech bro crap or llm crap, follow Two Minute Papers on youtube. Hes far from the only source ive looked into over the years, but he has some distinct advantages over most other ai sources. He does all the paper scouring for you so you can stay up to date on the latest ai progression(again, NOT LLMS, THIS TECH HAS 0 TO DO WITH LLMS. THEY DO NOT SHARE THE SAME ARCHITECTURE, THEY DO NOT RUN ON THE SAME HARDWARE. IT IS ENTIRELY DIFFERENT.) and he breaks it down succinctly, simply, and without overselling it or any of the other propaganda. He reviews papers that are currently released and checks out the AI behind it. He also handles algorithms which are super cool!