Many users dismissed AI as ineffective for independent research or business tasks by stressing its inability to handle anything outside the training set and responding with hostile insults questioning the researcher's competence.
No Digg Deeper questions have been answered for this story yet.
Most Activity
@nickcammarata how good are you at intuitively modeling belief states over future possible worlds on many timescales at once? how good are language models at doing this?
@nickcammarata I’ve got some bad news about the code you haven’t been looking at
it feels a little like a tacky researcher dream, cheap fiction writing. oh no these robots are super good at doing all the lame parts of your job, but need you to do all the fun cool parts!
I’m not sure I’ve seen a line of code in 6mo and also ai is hopelessly bad at doing my research without me telling it what matters. I didn’t expect this point to happen, it’s surprisingly jagged
@nickcammarata Is it going to remain like this until another paradigmatic breakthrough or is more rollouts on bigger models going to get us there
@nickcammarata The brutal part is that the code will reflect that same lack of import, and if your research is in anyway outside of the training set it will create gates such that only the training set is validated, rather than outside it. This is the most unfortunate result of LLM architecture
@nickcammarata the ten trillion dollar question: can you engineer taste with scale and brute force
@nickcammarata what does this even mean? every time i ask codex to build even a line change i put it in planning mode & review it. then i implement it. sounds like sloppy engineering on your end 🤷🏻♀️
@nickcammarata would you say you particularly cared about the line-by-line level of code? seems prone to gellman amnesia
@Entschul1999 @nickcammarata It already isn't like this
@nickcammarata Similar with business. It has the textbook answers, which are almost exactly what you don’t need.
@nickcammarata what kind of resreach you do? and do you work for your own or working with some company?
@nickcammarata That'll be fixed when Fable is available again
@willccbb @nickcammarata Mirrorfishin mofo
@nickcammarata What sort of research are you doing such that an AI is hopelessly bad at it?
@willccbb @nickcammarata I think there are better tools than a plain language model for this
@mhousmanpr69437 @nickcammarata This is the unfortunate result of the western school curriculum, which not only does not teach logic but teaches the opposite
@nickcammarata This is the jagged edge: code generation got cheap, problem framing did not. Research still bottlenecks on taste, priors, and knowing which facts should actually change the decision.
@tactipus_ @nickcammarata Jesus Christ how do you ever get anything done?
@Johnny2Fingersz @Entschul1999 @nickcammarata Cope