Thanks! Haven't seen that before. I still think it would be better to specialize on ethics issue and than apply its result on AGI sytsem developed by other (hopefully friendly) party. But It would be awesome if someone who is genuinely ethical develops AGI first. I'm really hoping that some big org which went furthest in AI research like google decides to cooperate with MIRI on that issue when they reach the critical point in AGI buildup.
This is something that I think is neglected (in part because it's not the relevant problem yet) in thinking about friendly AI. Even if we had solved all of the problems of stable goal systems, there could still be trouble, depending on who's goals are implemented. If it's a fast take-off, whoever cracks recursive self-improvement first, basically gets Godlike powers (in the form a genii that reshapes the world according to your wish). They define the whole future of the expanding visible universe. There are a lot of institutions who I do not trust to have ...
https://www.reddit.com/r/LessWrong/comments/2icm8m/if_we_knew_about_all_the_ways_an_intelligence/
I submitted this a while back to the lesswrong subreddit, but it occurs to me now that most LWers probably don't actually check the sub. So here it is again in case anyone that's interested didn't see it.