If the post-GPT-3 acceleration in AI had occurred a year or two earlier, at a time when I had no other responsibilities, I might have completely reoriented my life around the task of making the best possible contribution to AI safety. However, it came at a time when I had already devoted myself to another task, one that requires everything that I have. Yet paradoxically, pursuing that other task has yielded innumerable insights and challenges that are potentially relevant for Ai safety, and which might never have come my way in the more straightforward scenario. So I keep fighting that other fight, and in spare moments I catch up on AI issues, and sometimes I tell myself that maybe this zigzag path is ultimately the best one after all.
If the post-GPT-3 acceleration in AI had occurred a year or two earlier, at a time when I had no other responsibilities, I might have completely reoriented my life around the task of making the best possible contribution to AI safety. However, it came at a time when I had already devoted myself to another task, one that requires everything that I have. Yet paradoxically, pursuing that other task has yielded innumerable insights and challenges that are potentially relevant for Ai safety, and which might never have come my way in the more straightforward scenario. So I keep fighting that other fight, and in spare moments I catch up on AI issues, and sometimes I tell myself that maybe this zigzag path is ultimately the best one after all.
If you are willing to share—what are you doing now?