Ray Kurzweil: Your Brain in the Cloud

Sometimes people think that emotion and art are sort of sideshows to human intelligence, and the real essence of intelligence is thinking logically. If that were true, computers are already smarter than we are because they're much better at logical thinking than we are. It's actually things like being funny, being sexy, or expressing a loving sentiment; maybe in a poem or in a musical piece. That's the cutting edge of human intelligence.

It has to do with the hierarchy of the neocortex, and at the low levels, things might seem cold and mechanical because we're – our neocortex can recognize very simple objects and make very simple decisions. At a high level, it's dealing with concepts like the beauty of a poem by Emily Dickinson or the ability to create a poem like that. And we can come back to creativity. I believe it's an exercise in metaphors, and the neocortex is a metaphor machine. That's what it's good at.

That's why humans are creative, but we have this very large hierarchy where we have more and more abstract and complex patterns. You know, it builds on patterns below it. The difference between humans and say other primates is that we have more of the neocortex. We have this big forehead that we can fit in the frontal cortex and, therefore, have a higher number of hierarchies – of levels of the hierarchy.

And that was an enabling factor that permitted the evolution of language, technology, art, and science. We are going to create synthetic neocortexes based on the same principles. I'm actually working on that. And we've had hierarchical systems like that. I helped pioneer a concept called Hierarchical Hidden Markup Models, which is similar to what happens in the brain, but we only had a few levels.

The brain has a very large number of levels, and all the way from recognizing edges of objects up to "she's beautiful" or creating a beautiful painting or work of music, or scientific insight. We will create an artificial neocortex that has a comparable number or a greater number. I think the principal application will actually be to extend our own neocortex. We have 300 million pattern recognizers in the neocortex by my estimate.

That hierarchy we build ourselves, each of these pattern recognizers capable of connecting itself to other neocortexes to build its hierarchy. We build that hierarchy from the moment we're born or before that. We're constantly building it, but we run up against this limitation of 300 million. We'll be able to extend that and think in the cloud.

You know, if you do anything interesting with this, do a search or a language translation, or bring up a map or ask it a question. It doesn't take place in the box. It goes out to the cloud. We're going to put these just really gateways. This is a gateway to the cloud. We're going to put gateways to the cloud in our brains and have more than 300 million, just like the cloud can give you a thousand or a million computers for a tenth of a second.

You need another billion pattern recognizers, you'll be able to access that in the cloud. That's where we're headed, and we'll be more intelligent and able to actually think in a greater number of hierarchies. If you think that – realize that the quantitative improvement from primates to humans with the big forehead to allow a larger neocortex was an enabling factor for language, art, and music, and science.

What kind of qualitative leap can we make with another quantitative increase? And that's I think where we're headed.