Yeah, it’s not going so well. It is in fact going so incredibly poorly that so far the whole thing is quite plausibly vastly net negative, with most funding that has gone into “AI Safety” efforts serving as de facto capabilities research that both speeds things up and divides them and is only serving to get us killed faster.
I’m pretty curious about your, Eliezer’s, or others’ opinions on when AI safety started being net negative. Was very early concern (by Minsky, I.J. Good, etc) net negative? What about Eliezer writing the parts of the Sequences that were about AI risk? What about MIRI activities 2008-2012, 2012-2016, and the broader AI safety ecosystem starting 2012ish? Was it specifically OpenAI (2015) that was a turning point?
Without having a sense of which activities were net positive or negative I’m not really sure how anyone would take productive action given a judgment of overall activities being net negative.
OpenAI was the point where the turning point became visible; obviously the actual turn must have occurred before then. Arguably it was with DeepMind, since it looks (from way outside) like Demis has basically lost the struggle for reliable/final control of it inside Google.
I’m pretty curious about your, Eliezer’s, or others’ opinions on when AI safety started being net negative. Was very early concern (by Minsky, I.J. Good, etc) net negative? What about Eliezer writing the parts of the Sequences that were about AI risk? What about MIRI activities 2008-2012, 2012-2016, and the broader AI safety ecosystem starting 2012ish? Was it specifically OpenAI (2015) that was a turning point?
Without having a sense of which activities were net positive or negative I’m not really sure how anyone would take productive action given a judgment of overall activities being net negative.
OpenAI was the point where the turning point became visible; obviously the actual turn must have occurred before then. Arguably it was with DeepMind, since it looks (from way outside) like Demis has basically lost the struggle for reliable/final control of it inside Google.