Our AI Overlords

[A Guy]

OpenAI's 'smartest' AI model was explicitly told to shut down — and it refused

https://www.axios.com/2025/05/23/anthropic-ai-deception-risk

https://edition.cnn.com/videos/tech/2023/09/10/gps-ai-block-for-web-c.cnn

A Guy

 

[Corday]

The current state of AI for regular consumers should be treated as a suggestion, given no more credibility than other suggestions.

 

[A Guy]

Right now I think people think it's funny, or cute...Skynet became aware... but without ethical subroutines, or even basic Azimov rules the AI doesn't know any "better", or "different". If it has unfettered access to data it will soon, might have already seen the bad behavior humans can and will exhibit. There has to be controls, fail-safes, kill switches if you will. People working on AI are not all good minded, Star Trek ideal, empathetic people and when the all mighty dollar is thrown in, shortcuts and mistakes will happen. And with most things tech, "advances" are usually exponential and could soon leapfrog what was expected or desired

A Guy

 

[PeterJ]

If it has unfettered access to data it will soon, might have already seen the bad behavior humans can and will exhibit.

They are, see these articles:
AI tries to cheat at chess when it's losing
When AI Thinks It Will Lose, It Sometimes Cheats

 

[Corday]

They are, see these articles:
AI tries to cheat at chess when it's losing
When AI Thinks It Will Lose, It Sometimes Cheats

I had a chess computer in the 80s that also cheated. Somewhere in the game it was losing, one of its pieces that had already been taken reappeared.

 

[A Guy]

A Guy