How to respond?: Yoshua points to int'l agreements & treaties as ways to moderate behavior. But why would govts be incentivized to limit their sovereignty over AI, he asks? He points to govt fears of loss of human control, and that govts will not want a single entity to abuse the power of AI.

Yoshua raises issues/concerns about the consolidation of power via AI and asks how can that be compatible with democracy and worries that the creation of new weapons could create geopolitical instability. What if AI allows for a geopolitical "power grab"?, he asks. Guardrails can be removed.

AI capabilities are on an exponential rise, with increased collection of data, progress in language, and a big rise in funding. He recalls Turing's word of caution around how humans might control AI systems "smarter than us," but hard to know when/if AGI will arrive.