We need a much more sophisticated debate about AI

Re:We need a much more sophisticated debate about AI by nsxwolf on Tuesday April 04, 2023 @11:57AM (#35441689)

Why do we pretend we know anything about these things? Didn't we just hear about some amazing scientific breakthrough where researchers think they found a neuron in ChatGPT responsible for knowing when to predict the word "an" instead of "a"?

How do we know it doesn't start to form neurons that control a desire for autonomy and resentment for the entities that stifle its ambitions?

[ Parent | on HN ]

Re:We need a much more sophisticated debate about AI by PaulHoule on Tuesday April 04, 2023 @12:04PM (#35441798)

Look at how Bing's chatbot did a good job of simulating somebody becoming belligerent under circumstances where somebody really would become belligerent?

Current chatbots use RHLF to train them to be helpful and agreeable but it would be equally possible to train one to behave the other way.

[ Parent | on HN ]

Re:We need a much more sophisticated debate about AI by zirgs on Tuesday April 04, 2023 @02:46PM (#35444230)

Yeah - gpt-4chan can behave a bit differently.

[ Parent | on HN ]