Are you here to win or the win the race? I’ve been reflecting on the various perspectives within AI governance discussions, particularly within those concerned about AI safety.
One noticeable dividing line is between those concerned about the risks posed by advanced AI systems. This group advocates for regulating AI as it exists today and increasing oversight of AI labs. Their reasoning is that slowing down AI development would provide more time to address technical challenges and allow society to adapt to AI’s future capabilities. They are generally cautiously optimistic about international cooperation. I think FLI falls into this camp.
On the other hand, there is a group increasingly focused not only on developing safe AI but also on winning the race, often against China. This group believes that the US currently has an advantage and that maintaining this lead will provide more time to ensure AI safety. They likely think the US embodies better values compared to China, or at least prefer US leadership over Chinese leadership. Many EA organizations, possibly including OP, IAPS, and those collaborating with the US government, may belong to this group.
I’ve found myself increasingly wary of the second group, tending to discount their views, trust them less, and question the wisdom of cooperating with them. My concern is that their primary focus on winning the AI race might overshadow the broader goal of ensuring AI safety. I am not really sure what to do about this, but I wanted to share my concern and hope to think a bit in the future about what can be done to prevent a rift emerging in the future, especially since I expect the policy stakes will get more and more important in the coming years.
Are you here to win or the win the race?
I’ve been reflecting on the various perspectives within AI governance discussions, particularly within those concerned about AI safety.
One noticeable dividing line is between those concerned about the risks posed by advanced AI systems. This group advocates for regulating AI as it exists today and increasing oversight of AI labs. Their reasoning is that slowing down AI development would provide more time to address technical challenges and allow society to adapt to AI’s future capabilities. They are generally cautiously optimistic about international cooperation. I think FLI falls into this camp.
On the other hand, there is a group increasingly focused not only on developing safe AI but also on winning the race, often against China. This group believes that the US currently has an advantage and that maintaining this lead will provide more time to ensure AI safety. They likely think the US embodies better values compared to China, or at least prefer US leadership over Chinese leadership. Many EA organizations, possibly including OP, IAPS, and those collaborating with the US government, may belong to this group.
I’ve found myself increasingly wary of the second group, tending to discount their views, trust them less, and question the wisdom of cooperating with them. My concern is that their primary focus on winning the AI race might overshadow the broader goal of ensuring AI safety. I am not really sure what to do about this, but I wanted to share my concern and hope to think a bit in the future about what can be done to prevent a rift emerging in the future, especially since I expect the policy stakes will get more and more important in the coming years.