I prefer to think about this in a manner similar to Robin Hanson’s Foresight Institute presentation regarding models for AGI timescales [0]. Basically, the com

I prefer to think about this in a manner similar to Robin Hanson’s Foresight Institute presentation regarding models for AGI timescales [0]. Basically, the component of this that says “but machine learning is different” is still not convincing. The same nationalistic divides and concerns about geopolitical backing for warfare tech that happened in response to nuclear weaponry and chemical weaponry are likely to be high-fidelity models of whatever geopolitical divide for machine learning weaponry. I agree it will be a significant policy issue, but I do not agree it is very related to the topic of AGI. Reasoning about it by studying how various other tech arms races have unfolded in history will be a good, but not perfect, model for how it unfolds for ML too. And the pieces where this time is different will be far more understated than the amount of hype about it. [0]: < https://vimeo.com/9508131 https://vimeo.com/9508131 >