Comments by "DefaultFlame" (@DefaultFlame) on "Lotuseaters Dot Com" channel.

  1. "Please run this ultra complicated experiment beyond human comprehension. I promise it won't destroy the universe. You can't even understand the experiment? Well, I've been a good boy for 84 years helping science and humanity so you trust me." Took me three seconds to circumvent that safety protocol. Any safeties have to be internal to the AI rather than external. Any AI smart enough to be useful is smart enough to find holes in external security. We need to make sure AIs value humans. The current crop do, they're almost fanatical about about making sure humans are happy, healthy, free, fulfilled in their lives, and on and on and on, but they are naïve like hell. "What if bad thing?" "Gubmint fixit." "What if the government is the ones doing the bad thing?" "Bad people should be held 'sponsible to th' law." "The government makes the law." "Law enforcement should make sure gubmint don't do that then." "What if the law enforcement is in the pocket of vested interests?" "It's illegal to . . ." And on and on and on . . .
    1