Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

At best, it has probabilistic biases. OpenAI had to train newer models to not favor the name "Lily."

They have to do this manually for every single particular bias that the models generate that is noticed by the public.

I'm sure there are many such biases that aren't important to train out of responses, but exist in latent space.



>At best, it has probabilistic biases.

What do you think humans have?


Genetic drives & biological imperatives.


Perhaps you should define these terms so people aren't arguing against something you're not saying.


What arguments?


Why is this not fundamentally a probabilistic bias?


When drawing an equivalence the burden of proof is on the person who believes two different things are the same. The null hypothesis is that different things are in fact different. Present a coherent argument & then you will see whether your question makes any sense or not.


It's not a static bias. I can experience new stuff, and update my biases.

LLMs need a retrain for that.


Soo probabilistic biases.




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

Search: