🧵 what is the science of recursive reward modeling
Anonymous at Thu, 27 Jun 2024 20:44:10 UTC No. 16257210
does anyone know anything about this?
Anonymous at Thu, 27 Jun 2024 23:40:07 UTC No. 16257446
>>16257210
Holy shit they are trying to make ai human or lifelike.
Think of it like training a dog or animal. When you say sit and the dog sits on the floor. You give it a treat.
They want to do the same thing with ai according to this paper.
They want to give incentive based learning to a machine.
I guess with the success with adversarial ai using the techniques of biological mimicry, I assume they feel like they can apply the same techniques.
However I don’t think it will be successful mostly because neurological functions like these are something science hasn’t been able to solve.
For example how do you give an AI inspiration? How does it get inspired? What does that mean for a human? Why does a human get inspired? How is that function programmed in our genetic structure?
The are dangerous thoughts anon. Let’s get back to trolling about vaccines.
Anonymous at Fri, 28 Jun 2024 00:14:38 UTC No. 16257473
>>16257446
a punishment reward system for a machine seems pretty arbitrary.