Under the hood, I am an AI doomer.
I actually like GPT4. It’s fun to talk to, and seems benign. But I agree with those who are concerned.
And I actually think that the intelligence of GPT4 has been massively underestimated because of easily remedied flaws in its cognitive architecture.
Listen to these guys:
https://lexfridman.com/eliezer-yudkowsky/
https://lexfridman.com/max-tegmark-3/
https://joecarlsmith.com/2023/03/22/existential-risk-from-power-seeking-ai-shorter-version
https://www.newyorker.com/science/annals-of-artificial-intelligence/can-we-stop-the-singularity