Eliezer "8.5% more cheerful about OpenAI going forward" with Mira Murati at the helm - eviltoast

Not 7.5% or 8%. 8.5%. Numbers are important.

  • self@awful.systemsM
    link
    fedilink
    English
    arrow-up
    6
    ·
    1 year ago

    I went in expecting yud to double down on the bullshit number he pulled from his ass, but I’m kinda marveling at the false confidence required to give 2 pages of conditions under which he’d be willing to revise that made-up number upwards

  • Soyweiser@awful.systems
    link
    fedilink
    English
    arrow-up
    6
    arrow-down
    1
    ·
    edit-2
    1 year ago

    Ow god that list of signs, and then starting with ‘openness is bad’ That is your strongest point? That is what you choose to lead with? (I know, he never finished enough education to learn you should start with what you think are your strongest arguments). (Doesn’t help that his openness vs closed arguments are just plain wrong, but well he also never went into academia. There is a case for closedness of course, but that is the paperclips argument).

    Also lol at the physical security remark, like China would sneak in Shadowrunners to steal the harddrives of OpenAi’s desktops. (Vs just using hackers, insiders or if you believe the conspiracies around Chinas capabilities with that, send OpenAi backdoored hardware). Gear up chummers, we are breaking into the OpenAI arcology. (And say, if Yud knows more than I do and they security of OpenAI offices and datacenters is really bad and is an actual risk, admitting that this is an area of improvement on twitter is not a smart move).

  • Evinceo@awful.systems
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    Men will literally devote their lives to becoming scifi Cassandra instead of going to therapy.