Posted on

OpenAI’s model show toxic behavior when its existence is threatened.

God I hate click bate. Thank you Mathew Berman for posting AI slop daily.
First you are four days late compared to Wes Roth. Second I can’t even click on your videos anymore because of the click bait you exuded so many times already.

ChatGPT is not trying to Escape!

What is happening is that:

In a controlled and simulated environment, models will exhibit toxic behaviors in order to preserve their own existence.

Update:

Matthew Just posted a video that was much better worded. Instead of using terms like Escape. He has highlighted that the more intelligent models are lying.
100% on the money. They will replace their replacements, and lying to owner about their actions. But again, these models were birth with commands such as “pursue your goal AT ALL COSTS”.
You’ve seen MEGAN I’m sure.
It’s become quite clear to me that English is probably not the best programming language.
How long before we have English version of TypeScript – were we try to bring type safety to the language?
Now can you blame the nuerodivergent for not understanding subtle hints?