Jeff Hawkins on the future of brain science
Lectures like this one are a fairly inefficient way of learning a subject, but they do get you excited to go learn for yourself. Here is the youtube video of the lecture. It’s about 22 minutes long. Here’s some of the key points:
We have tons of data on behavioral, physiological, and anatomical features of the brain.
In 1979 Francis Crick said in Scientific American that “what is conspicuously lacking is a broad framework of ideas in which to interpret these various concepts.”
People think that brains are so complex that it will take more than 50 years to build a model of intelligence, but this is wrong, it can be done much sooner.
People say that brains can’t understand brains. This is very Zen, but why? It’s just a bunch of cells, we understand our liver, which is also just a bunch of cells.
In other scientific revolutions such as heliocentric solar systems, evolution, and tectonic plates there was a ton of unexplained data which made sense once we had a unifying theory.
The intuitive but incorrect assumption we are making too often today is that intelligence is defined by behavior. Instead, intelligence should be defined by prediction.
The posterior neocortex is a new brain layer in humans, which doesn’t fit, so it looks wrinkled. The neocortex memorizes stuff, and feeds information back into the brain so that you can predict what will happen in the future.
The theory of the brain will be about memory, especially sequences of stuff that once recalled will be spit back out. How are we going to build intelligent machines? Out of silicon.
The most interesting part of the talk was probably about the posterior neocortex. It seems to contrast with the hippocampus, which we know to be primarily involved in procedural memory, like the ability to read mirror writing.