I attended both days & enjoyed it quite a bit—more than I expected to. I’d be likely to fly to New York to attend next year if the speaker list looks equally interesting. There weren’t many unresolved topics in my mind because I talked a lot to other people during the event about things that occurred to me, and asked a lot of questions.
I most enjoyed Peter Norvig’s talk (largely about some recent Google AI projects) and Steven Pinker’s talk (reviewing downward-trending measures of human violence.) Those seemed pretty information-dense and covered topics that I didn’t know much about. I would recommend Jaan Tallinn’s talk to anyone who is a “light transhumanist” but hasn’t thought about the simulation argument very much—it was very well-done, but a lot of it was about ideas with which I have already made my peace.
I thought the level of professionalism and the quality of the talks reflected well on SIAI and CFAR. It was interesting to me that about half of the people I met during breaks had flown quite a way and paid quite some money, even though they weren’t tightly connected to the community and didn’t always have a ton of familiarity with all the typical transhumanist ideas—I expected that there would be less diversity in backgrounds.
I think people haven’t quite updated on the news value of what Norvig said in his talk this year. In 2007, he listed the specific components that were important in order to make progress in AGI. This year, he opened his talk by saying that substantial progress has been made towards 5⁄6 of those components. http://singularitysummit.com/the-history-and-future-of-technological-change/ is the 2007 talk.
However, some of those goals are way easier than the others. Doing stuff online with lots of data is important but something that we can predictably do after investing enough resources. Probabilistic first-order logic and hierarchical representations are the parts that require lots and lots of insight, and progress on those two is far less impressive.
Steven Pinker’s talk (reviewing downward-trending measures of human violence.) Those seemed pretty information-dense and covered topics that I didn’t know much about.
But Pinker wrote a whole (excellent, highly recommended) book about that—was there anything new?
I attended both days & enjoyed it quite a bit—more than I expected to. I’d be likely to fly to New York to attend next year if the speaker list looks equally interesting. There weren’t many unresolved topics in my mind because I talked a lot to other people during the event about things that occurred to me, and asked a lot of questions.
I most enjoyed Peter Norvig’s talk (largely about some recent Google AI projects) and Steven Pinker’s talk (reviewing downward-trending measures of human violence.) Those seemed pretty information-dense and covered topics that I didn’t know much about. I would recommend Jaan Tallinn’s talk to anyone who is a “light transhumanist” but hasn’t thought about the simulation argument very much—it was very well-done, but a lot of it was about ideas with which I have already made my peace.
I thought the level of professionalism and the quality of the talks reflected well on SIAI and CFAR. It was interesting to me that about half of the people I met during breaks had flown quite a way and paid quite some money, even though they weren’t tightly connected to the community and didn’t always have a ton of familiarity with all the typical transhumanist ideas—I expected that there would be less diversity in backgrounds.
I think people haven’t quite updated on the news value of what Norvig said in his talk this year. In 2007, he listed the specific components that were important in order to make progress in AGI. This year, he opened his talk by saying that substantial progress has been made towards 5⁄6 of those components. http://singularitysummit.com/the-history-and-future-of-technological-change/ is the 2007 talk.
However, some of those goals are way easier than the others. Doing stuff online with lots of data is important but something that we can predictably do after investing enough resources. Probabilistic first-order logic and hierarchical representations are the parts that require lots and lots of insight, and progress on those two is far less impressive.
But Pinker wrote a whole (excellent, highly recommended) book about that—was there anything new?
Sure, but I didn’t read it! If I had to guess I doubt that much of it was new.