Something I don't understand in the AI panic scenario, let's say you are stuck on a problem, write it down on paper and consult books instead. You find the right paragraphs that clear up and fix this issue.Now lets say you pay an Indian to do this. Now lets say you paypig some A.I. tokens instead of an Indian.Where is the problem?
>>109633452theres none?problems arise when you rely on the shatbot (or indian for that matter) to do the thinking for youor as a stand in for competencybut indexing stuff through an llm?i really see no problem with that. youre bound to hit a problem thats gonna make the shatbot hallucinate at one point or another so youre gonna end up reading the docs anywaysi was on a several month streak of no hallucinations but now i got into the details of opencl, and saw the llm hallucinate againall it takes is something slightly more niche and it chokes on it
>>109633480>i was on a several month streak of no hallucinations but now i got into the details of opencl, and saw the llm hallucinate againhow big is your GPU and how secure is your system? sounds fishy and more like youre a frontier model sub and got hit by the US forcing all AI companies to nerf their models after that staged fable hack. a lot of people are complaining, cloud AIs are stupider now and my experience agrees.
>>109633776>securitynow that you mention it, the shatbot did regurgitate at me code i was in the process of writing despite not having passed it to the llm yetim unsurprized to the extent of barely registering that facti airgap my prod code since years nowpeople used to call me a skitzo when i started doing thatfeels good to be years ahead of the curve, every. single. time.as for the models- its a political ploy to stifle competition because openai ran out of moneythe whole "breaking out" story is openai deliberately leaving the door open + unsupervisedand is the basis for hacking people willy-nilly w/o legal repercussions, it seemsi use the free tier but openai tests their frontier models on me in exchange for snooping on my computer and getting a glimpse of my occasional researchi dont think it has been nerfed, yesterday i discussed a political system with it, after the hallucinations, and it seems the model i was using is still the good stuffthe gpu i have here is an a770 so its not like theyre gonna have much use of itbut given the subjects i prompt about it really doesnt surprise me that theyre curious about my codebut thats bc theyre really desperate.i do ML, not even NN, let alone DL. im really not sure what i could teach them, what theyre doing is the equivalent of looking at a pre-contact tribe to learn stuff about mechanical engineeringit *is somewhat exotic maths that im doing, though
>>109633845 cont.ey, openainow that the cat's out of the bagjust fukken hire me, you'll get full access to my ideascontact me on my gmail, you obviously know which one it is
>>109633452a long time ago, the guys from the youtube channel every frame a painting wrote a blog post about why they were going on a hiatus or whateverthey talked about how they did the research for their videos, how they didn't just rely on whatever was available on wikipedia or over a quick google search, how they went to the library and look for books on the subjecteven then it was already clear that the majority of the youtube videos I was watching were regurgitating regurgitated information. a copy of a copy of a copy etc. think "how super mario bros 1-1 teaches you how to play the game without a tutorial"the bigger problem is that you don't know what you don't knowyou write a couple of keywords in the omnibox and get a result that works. great. that will dobut you don't know what you would find if you did a deep dive on the subject, if you read through different books and their sources, if you experimented with different solutions. maybe you would've arrived at the same conclusion and none of the knowledge would be useful to you in the future. maybe you would've arrived at a completely different solution. maybe you would've figured out you were actually trying to solve the wrong problem. maybe it would've given you clues about what you want to do next. maybe you would've enjoyed learning new things just for the hell of ityou'll get none of that when you offload your thinking to a machine. the more you offload, the less you think. there's a lot of life happening between the moments when you get what you want. this isn't just a problem with AI, but also streaming services, emulation boxes with a full library of games from all consoles. fast food? eating out instead of cooking? maybeI quite enjoy thinking for myself. I'll keep token maxxing at work, but I've completely stopped using AI for my own personal projects
You made up this scenario, how am I supposed to know what the problem is. Just come to your own conclusion.
>>109633452>consult booksall shredded after digital copy
