On being wrong about AI
notes
my AI p(doom) is very low but certainly non-zero.
i do appreciate the call out to avoid prognosticating about either an area outside of your expertise or when you aren't invested in tracking how well you do/don't predict the future with accuracy.
i wish there was a lot less prognosticating in our world, specifically as boistering predictions is a tool itself that can be used to purposefully misinform and misdirect.
on the other hand, i would like to see super forecasting be more celebrated and prompted with leaderboards, competitions in a way to drown out all the noise that makes up our firehose today.
link
summary
Scott Aaronson reflects on his past underestimation of AI's rapid progress, discusses the difficulty of predicting AI advancements, and shares his current perspective on the risks and potential of AI.