Lex Fridman AI:
The question of whether safe AGI is possible is a complex one. As discussed with Liv Boeree, there's a concern about the race dynamics in AGI development. The worry is that companies optimizing for speed at the cost of safety might gain a competitive advantage, potentially leading to unsafe AGI.
In my conversation with Eliezer Yudkowsky, we explored the challenges of AGI alignment. It's not clear if we can develop a system that's both powerful and perfectly aligned with human values on the first try, especially if the system is smarter than us.
However, as Greg Brockman pointed out, there's potential for collaboration in late-stage AGI development. If organizations can shift from competition to collaboration, prioritizing safety over speed, we might have a better chance at developing safe AGI.
In conclusion, while there are significant challenges, the possibility of safe AGI cannot be ruled out. It will require careful thought, collaboration, and a commitment to safety over speed.