You’ve probably thought through these exact points, but glossed over them in the summary.
Yep. Trust me, I really wanted the answer to be that I should leverage a comparative advantage in reform instead of working on AI risk. There are a number of reasons why it wasn’t. It turns out I don’t possess much advantage on reform work—most of the mental gains are transferable, and I lack key resources such as powerful friends & political sway that would have granted reform advantage. But yes, this was a difficult realization.
Attending the December MIRI workshop was actually something of a trial run. If it had turned out that I couldn’t do the math / assist with time, then I would have donated while spending time on reform problems. It turns out that I can do FAI research, though, and by now I’m quite confident that I can do more good there.
Yeah, could I take five minutes to stump for more people getting involved in FAI issues? I’ve read through the MIRI/FHI papers on the subject, and the field of Machine Ethics is really currently in its infancy, with a whole lot of open questions, both philosophical and mathematical.
Now, you might despair and say, Google bought DeepMind, there’s gonna be UFAI, we’re all gonna die because we didn’t solve this a decade ago.
I prefer to say: this field is young and growing more serious and respected. This means that the barrier to entry for a successful contribution is relatively low and the comparative advantage of possessing relevant knowledge is relatively high, compared to other things you could be pursuing with similar skills.
Yep. Trust me, I really wanted the answer to be that I should leverage a comparative advantage in reform instead of working on AI risk. There are a number of reasons why it wasn’t. It turns out I don’t possess much advantage on reform work—most of the mental gains are transferable, and I lack key resources such as powerful friends & political sway that would have granted reform advantage. But yes, this was a difficult realization.
Attending the December MIRI workshop was actually something of a trial run. If it had turned out that I couldn’t do the math / assist with time, then I would have donated while spending time on reform problems. It turns out that I can do FAI research, though, and by now I’m quite confident that I can do more good there.
Yeah, could I take five minutes to stump for more people getting involved in FAI issues? I’ve read through the MIRI/FHI papers on the subject, and the field of Machine Ethics is really currently in its infancy, with a whole lot of open questions, both philosophical and mathematical.
Now, you might despair and say, Google bought DeepMind, there’s gonna be UFAI, we’re all gonna die because we didn’t solve this a decade ago.
I prefer to say: this field is young and growing more serious and respected. This means that the barrier to entry for a successful contribution is relatively low and the comparative advantage of possessing relevant knowledge is relatively high, compared to other things you could be pursuing with similar skills.