Suppose there are some things that we don't understand about the universe, but if you understand human intelligence and you understand the gaps in our abilities to think about things, maybe we can engineer in a computer more advanced intelligences that can help augment our ability to think.

If you give somebody a lot of questions to answer and then they walk by a bowl of candy, they are more likely to grab the candy because they're tired out from answering questions and can't resist.

I often try to think about, What sounds like a bad idea, but if you find the right plan of attack, it's actually a really good idea? I spend a lot of time really trying to systematically tackle problems from different angles.

These disorders - schizophrenia, Alzheimer's, depression, addiction - they not only steal our time to live, they change who we are.

We don't have a "consciousness meter" that'll tell us exactly how conscious something is. I think we might get there eventually.

A lot of good ideas are actually bad ideas because, since they sound good, everybody's already doing them.

You can imagine over very long timescales, perhaps far beyond the multi-decade time scale, we might be able to ask very deep questions about why we feel the way we feel about things, or why we think of ourselves in certain ways - questions that have been in the realm of psychology and philosophy but have been very difficult to get a firm mechanistic laws-of-physics grasp on.