I have a few ideas:

1) It's a "basilisk", i.e. an imaginary lovecraftian threat that doesn't even make sense outside of some highly particular and probably wrong belief system. (That's not my definition of basilisk, but it is what I think of such claims.)

2) Some mundane fact about the difficulty or danger of actually trying to save the world (in the specific sense of shaping a singularity) has made his blood run cold. It could be the existence in the real world of powerful evil cliques; it could be the psychological potential of joining them, or just of becoming selfish in an ordinary sense.

3) I remember when I was 22 and realized (according to the plans I had at the time) that it might take me eight years to save the world. That was very daunting, because at the time it looked like it would be a joyless, stressful, solitary existence, for an unimaginably long period of time. And as it turned out, I didn't even get it done... Will could be fleeing the responsibilities of his "position" - I mean his existential position, which surely includes the perception that he has the potential to make a difference in a huge way.

ETA 4) He wants to create a barrier (a "credibility barrier") between himself and his former associates in SI, so as to develop his own thinking, because there's a systematic deficiency in their outlook and he must avoid the temptation of working within that paradigm.

But there does seem to be a number of people here who understand what is going on, but are refusing to offer their explanations, in spite of the fact that a lot of people are confused here.

Maybe they take the basilisk threat seriously?

Just don't be fooled by intelligence too much. Just because those people can disgorge some math that doesn't lent their extraordinary claims much credence. Most of the credence they assign is based on mutual reassurance anyway. Just like a bunch of ufologists updating on each others evidence of alien abductions.

The only danger to worry about is that this community of schizo scifi nerds is going to

Alicorn, if a "should this be moderated" poll is required anywhere in this thread this is the kind of trolling that needs to be targeted. Way across the line.

I endorse you not worrying about SI impeding AI progress on any significant scale.
I would also endorse, if you're genuinely interested in encouraging AI research, devoting more of your attention to the problems that are actually impeding AI progress.