While I don’t like Mechanize’s post fully, and have some reservations about the level of technological determinism expressed in the article, I do think that I’m substantially more skeptical of shifting the paradigm of AIs in ways that don’t boost capabilities nearly as much as Sahil thinks, and I’m much more skeptical of the claim that automating away humans was a contingent goal, and tend towards more technological determinism than Sahil:
While I don’t like Mechanize’s post fully, and have some reservations about the level of technological determinism expressed in the article, I do think that I’m substantially more skeptical of shifting the paradigm of AIs in ways that don’t boost capabilities nearly as much as Sahil thinks, and I’m much more skeptical of the claim that automating away humans was a contingent goal, and tend towards more technological determinism than Sahil:
The Future of AI is Already Written
(I already talked about why AIs are easy to shut down using a simpler hypothesis than Sahil did, this comment is more about how it’s way more difficult to steer technological development than people appreciate, and also incorrectly overestimate the level of control humanity does have over things).