3 ms·

“Safety” in the context of AI has nothing to do with actual safety. It’s about maximizing banality to avoid the slightest risk to corporate brand reputation.

• imjonse 3y ago
  That too, but dismissing AI safety entirely because big companies are cautious not to get sued if their chatbots parrot hate speech is missing a large part of the picture. In the coming years 'free' AI will no longer mean just rogue chatbots and deepfakes, but start looking a lot more like cars, weapons and heavy machinery; you can't really postpone talking about safety/ethics/reglementation.
  • 65a 3y ago
    It's a feature not a bug. Maximizing banality is being cast as safety to avoid discussing the harder problems long term. I agree that any new technology comes with risks, including serious ones, and humans should get better at managing those risks, but it's a long jump from that argument to "only we should have this new advanced technology, for your own safety", where safety is ill-defined and likely misses the very real safety risks of "only them having this new advanced technology"
    • imjonse 3y ago
      I never argued for restrictions (I don't agree with the article authors) and it's a real risk if powerful AI is only controlled by already powerful entities. My comment was about the dangers of equating AI safety with aligning chatbots and censoring diffusion models.