Hacker Newsnew | past | comments | ask | show | jobs | submit | 8note's commentslogin

defining the workflow such that the operation is a set of relevant questions

im not seeing it.

youd ask it to pick a location on the ast to add something from the grammar?

i dont see how this stays confined well enough? make a new output space every time? does that end up auto-regressive?


id say yes. a linear classifier that classifies between red and yellow balls will hallucinate on blue.

linear regressions hallucinate in the simpson's paradox.

the model output can be quite confident and not representative of reality


if it puts a high confidence value on a wrong answer, thats still hallucinating, no?

llm hallucinations are high probability tokens that are incorrect vs the real world


Yes, there is no magic sauce here that makes stochastic output binary if that’s what people are looking for.

Right and so maybe we should stop saying "can't hallucinate" when it can by definition.

It’s not what people are looking for, but what they wrongly claim.

if you hold people liable, then they will want liability insurance.

what's the point of liability insurance if youll never be held liable?


thats not a very convincing model. as long as training data and compute is available people are going to be able to make serviceable alternatives

If it's framed as national security being at stake why do you think the government won't regulate access to compute? Globally there are only a few firms designing competitive chips and only a handful of cutting edge fabs the world over. This could easily serve as justification to finally ramp up the war on general computing.

or rather, hack just enough and within what the user asks and not more.

I guess that's what "alignment" always comes down to? "Do what I want even if I can't tell you exactly what I want - because if I could, I could also just do the work myself"?

Here’s an easier criterion: “Do at least as good of a job at estimating what I probably want as an educated person would estimate, and do that, such that a reasonable human person who was aware of what I requested and of what you actually did, would conclude that as best as they can tell (without checking back with me), you did what I wanted.” .

This doesn’t seem like an unreasonable requirement to me. People do this all the time?

Sure, a request might not always be perfectly unambiguous. But people can generally estimate pretty well whether someone making a request is expecting the agent fulfilling the request to commit a crime in order to fulfill the request.


folding laundry is an exemplary case of a lot of household work which involves carefully collecting items through the house, putting them into a machine to run a cleaning cycle, then putting them away in their right location.

folding is a hard middle step in the laundry process, but the job is end to end


probably not quite.

intimacy over onlyfans with a guy in india pretending to be the woman is the proper killer app.


that an inssurection was stopped doesnt make it a peaceful transfer of power.

the thing about the US system is that the president doesnt do the transfer of power. they just lose power on a certain day, and congress grants it to somebody new


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: