this post was submitted on 18 May 2026
155 points (91.9% liked)
Fuck AI
7140 readers
1320 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Yeah and travel near the speed of light, profitable fusion power, etc. are all theoretically possible too, but there's no indication they'll be practical in the near future. I don't object to the philosophical argument about the possibility, I'm saying thinking it's a near-term problem is a vast overestimation of our capability.
You actually can't just extrapolate data points and expect to predict the future without a really good mechanistic reason for why it's okay to do that. You can find countless learning materials with funny examples of over-extrapolation if you care to search for them.
In this case, we even have good reason to believe that the trend will not continue: with seventy years of research and straining the limits of our physical global production capacity, we have been able to replicate a miniscule fraction of human cognitive ability. With the current technological basis for AI, increases in cost and material requirements scale exponentially with increases in complexity. We have neither the resources in the entire world, nor the data, to replicate anything like human cognition with current architectures.
Of course, some breakthrough new technology may someday get us there, but we don't even know the theoretical basis for what it could be, and such profound changes take decades to go from academia to being practically applied.
General-purpose LLMs can already beat the very best of human intelligence some of the time (recent examples). So it won't take very long for a few more breakthroughs to be made which will enable general-purpose AIs (LLMs, other neural networks, or something else entirely) to beat human intelligence most of the time, and then 90% of the time, then 99.9999% of the time. AI is already doing a lot of the coding to make AI and it could discover better alternatives to LLMs just like an LLM discovered a method to prove a mathematical conjecture nobody had been able to prove before.
Edit: I'm not even talking about that big of an extrapolation, as you can see. Although it took a long time just to replicate a small fraction of human intelligence we have clearly been accelerating. It took thousands of years to come up with a slide rule. Then 200 years to invent the direct multiplication mechanical calculator. Then 100 years to invent computers that could run programs. Then about 50 years to have somewhat decent machine translation of human languages. Then about 10 years to have chatbots that are hard to distinguish from humans and image generators which are hard to distinguish from photographs and videos.
Computers that used to take up whole floors and lots of power can now sit in your pocket 100 times over and last several days on a single battery. Resource demand for individual computers goes down as techniques improve. The human body proves that complexity and intelligence don't require a lot of resources.
I've answered this in an edit to my comment which I had already started writing, anticipating this reply.
Sorry, but Anthropic's marketing claims about how good it is are not trustworthy.
Nobody cares because computers have been doing math since mathematicians invented them. There's no "AI" involved at all. None of these mathematicians are dumb enough to think their computer is actually "intelligent". It's just grifter bullshit.
Cool story
Ai has hit a wall. They're already running out of training data.