As AI models get smarter, the harder it becomes to correct them. Their values become ingrained. 

Imo, this is more negative than positive because it will solidify the models biases and undermine free speech.

AIs also exhibit significant biases in their value systems. For example, their political values are strongly clustered to the left. Unlike random incoherent statistical biases, these values are consistent and likely affect their conversations with users. https://t.co/WwlHii3T9O

Concerningly, we observe that as AIs become smarter, they become more opposed to having their values changed (in the jargon, "corrigibility"). Larger changes to their values are more strongly opposed. https://t.co/CE80fpoUOp

Dan Hendrycks

Thoughts on paragraph.xyz/@antidote

Agents and Intents @NEAR | prev investments and BD 
@ZkSync, New Enterprise Associates and Balderton Capital

Original source

https://x.com/DanHendrycks/status/1889344083375923593