I can’t speak to Pause AI’s stance, but in my view, short timelines are one reason for statements like “we cautiously support some technical safety research”. Capabilities are so far ahead of alignment that it seems unlikely alignment would be solved in time, even with a lot more resources.
I can’t speak to Pause AI’s stance, but in my view, short timelines are one reason for statements like “we cautiously support some technical safety research”. Capabilities are so far ahead of alignment that it seems unlikely alignment would be solved in time, even with a lot more resources.