mnvr

Maybe talk­ing isn’t the best way to com­mu­ni­cate with LLMs

• mnvr 29 Nov 2023 16:19 UTC

  2 points

  0

  in reply to: Aleksi Liimatainen’s comment on: Trust your in­tu­ition—Kah­ne­man’s book misses the for­est for the trees

  Fair enough! Also, thank you for the comment, I found myself nodding to all your observations.

Trust your in­tu­ition—Kah­ne­man’s book misses the for­est for the trees

• mnvr 31 Oct 2023 23:41 UTC

  5 points

  0

  in reply to: Shankar Sivarajan’s comment on: If AIs be­come self-aware, what re­li­gion will they have?

  Hey, thanks for the comment. While I agree that “they’ll believe what they’re programmed to”, I feel that’s a bit too recent term. I was imagining a farther future where artificial intelligence goes beyond current LLMs and current political scenarios.

  I’m doing these thought experiments of a future where artificial intelligence (whether as an evolution of current day LLMs or some other mechanism) have enough agency to seek out novel information. Such agency would be given to the artificial intelligence not out of benevolence, but just as a way to get it to do its job properly.

  In such a scenario, saying that the AI will just believe what it is programmed to believe is akin to saying that a human child will never believe differently from what education /​ indoctrination (take your pick) they’ve been given as a child. That’s a good general rule of thumb, but not a given.

If AIs be­come self-aware, what re­li­gion will they have?

Cen­sor­ship in LLMs is here to stay be­cause it mir­rors how our own in­tel­li­gence is structured