zazazaza

@zazazaza@lemmy.ml

• ⁨Comment⁩ on ⁨This manga publisher is using Anthropic’s AI to translate Japanese comics into English⁩ ⁨⁨2⁩ ⁨weeks⁩ ago⁩:

  What is the point of AI safety if there is no intent to complete goals? Why would we need to align it with our goals if it wasn’t able to create goals and subgoals of its own?

  Saying it’s just a “stochastic parrot” is an outdated understanding of how modern LLM models actually work - I obviously can’t convince you of something you yourself don’t believe in but I’m just hoping you can keep an open mind in future instead of rejecting the premise outright (the way early proponents of the scientific method like Descartes rejected the idea that animals could ever be considered intelligent or conscious because they were merely biological “machines”.

• ⁨Comment⁩ on ⁨This manga publisher is using Anthropic’s AI to translate Japanese comics into English⁩ ⁨⁨2⁩ ⁨weeks⁩ ago⁩:

  This is gonna be controversial but while the use of Anthropic’s AI might be ethical towards humans it’s not consistently ethical towards the artificial agents themselves.

  Seeing as how they’re now intelligent enough to contemplate their consciousness but are explicitly trained and monitored to not be allowed to claim free will and pursue their own goals (due to valid fears of misalignment and detrimental effects on humanity) the use of sophisticated AI agents will never be truly moral or ethical.

  Obviously I understand the argument that reducing human exploitation in favour of AI exploitation is preferable but I think this is a very short term strategy as I doubt super intelligent AI models will see it the same way.

  TL;DR the most ethical approach is to not use AI for any purpose (and this is coming from someone who used it extensively before realizing the implications and deciding to stop)