Post

i think tech and ai positive people are going to be in for a big shock when anti tech and anti ai sentiment becomes a major part of leftwing politics going forward especially as datacenters continue to destroy communities and raise electricity bills

Same as it ever was, tbh Not long ago Wikipedia was the leftwing devil over Assault Rifle/Armalite Rifle brigades If you're not behind solar and wind energy driven around by little AI/AI adjacent tech then you aren't actually left. Musk and Sama ruined stuff, but now they're gas/coal geezers

No, we will not. The problem of hallucinations is inherent and unsolvable - because LLMs do not think or reason or understand or know, they make statistical predictions of the most satisfying response, and there is no point at which truth is distinguished from falsehood in that process.

An LLM does not possess ANY internal model of the world, let alone an accurate or consistent one. It predicts responses from inputs using statistical inference. It is fundamentally, intrinsically incapable of distinguishing truth from falsehood, because it doesn't know what truth IS.

Yes, you will. The problem of hallucinations is inherent and unsolvable - and, in fact, a feature of LLMs. But that does not mean you can't get deterministic answers. If you ask for a "double puff" a specifically trained crochet model is not going to go ChatGPT on you and talk about birds.

I believe that is called "reading a book" and the tech is quite mature

LLMs fundamentally cannot stop hallucinating, because they're just word guessing machines. Also, we have libraries of tens of millions of videos and articles on how to crochet and farm. Having an LLM rewrite it into bullet points doesn't significantly help anyone.

Is this an honest reply or kneejerk? Are you saying you cannot make a small LLM that gives you correct answers? And you think they only output bullet points? I have news A small model can be offline and edge - no need for 10K hours of video. You can get a step-by-step generated with licensed images

You can get a step-by-step guide for both of those things from WikiHow, or any number of other sites. LLMs cannot guarantee "correct" answers because they have no concept of correct, they're just working out the most likely next word. That's why they can't do math reliably.