I'm not sure when it comes out, but he earlier did an authors@google talk on the same subject, which has been embargoed so the TED version came out first. My comments are based on that earlier version of this talk.

Sam did a very good job of demonstrating that it would be really good if we could get morality from science. He did a good job of pointing out that science can validly inform our decision making. However he misses the fact that science is about problems that are tractable, not important. And he is unable to present his variant of Western morality with sufficient force to convince people who do not start with a morality similar to the one he already has.

In essence it is the same mistake the social sciences make. The case that something is important, and that data is a useful thing to have in studying it, does not suffice to make a compelling case that we are on the right line of research.

See http://bentilly.blogspot.com/2009/09/what-makes-it-science.h... for a more detailed explanation of my opinions about what is and is not science.