2 comments on “Uncontrollable

  1. “The rest of this review will focus on complaints, mostly about McKee’s overconfidence.”

    It seems to me that the main claims you’re saying McKee is overconfident about are (1) that we’ll get superintelligence within 10 years of the book being published (by 2033) and (2) that Asimov’s laws are not even remotely viable.

    I’m not sure what credences you and Mckee assign to these claims, but it’s very plausible to me I agree with you more than McKee on these.

    I recall thinking that Mckee was maybe putting too much weight on 10-year ASI timelines, but maybe not. We’d have to operationalize “ASI” better. For many definitions, >50% on ASI by the end of 2033 seems quite reasonable, and maybe even ~70-80% confidence could be justified.

  2. Hmm. I don’t seem to have a clear recollection of what I meant by overconfidence.

    He’s not overconfident about when ASI will be created.

    He’s overconfident about the “remotely viable” claim, but that’s a vague enough claim that I don’t want to put too much weight on it.

    He’s overconfident about the value of the “moonshot plan”, and about neglecting alternatives such as trying to pause AI development (but enforcing a pause would be hard).

    There were probably a bunch of other little comments in the first half of the book that contributed to my impression of overconfidence, but which weren’t important enough to write about.

Leave a Reply

Your email address will not be published. Required fields are marked *