I’m not spending the additional 34min apparently required to find out what in the world they think neural network training actually is that it could ever possibly involve strategy on the part of the network, but I’m willing to bet it’s extremely dumb.
I’m almost certain I’ve seen EY catch shit on twitter (from actual ml researchers no less) for insinuating something very similar.
[Taking the derivative of a function] oh fuck the function is conscious and plotting against us.
to be fair, assuming computers are like that because they hate all humans and want to fuck you up is basically true
Sorry the thesis is that checks reality gradient descent might be consciously trying to avoid having its nefarious goals overridden?
what if right my spellcheck dictionary got so big it TOOK OVER makes u think
It is imperative that we first build a mathematical framework for guaranteeing benevolent thesauri before we travel this path any further!
Urban Dictionary’s Basilisk
I conclude that scheming is a disturbingly plausible outcome of using baseline machine learning methods to train goal-directed AIs sophisticated enough to scheme (my subjective probability on such an outcome, given these conditions, is ~25%).
Out: vibes and guesswork
In: “subjective probability”
