Jeff Hawkins: What Happens When Computers Get Really Smart
I went to see Palm and Treo creator Jeff Hawkins speak at Rensselaer Polytechnic Institute today (trivia note: RPI claims to have awarded the first engineering degree in America).
Hawkins is both smart and humble and would be worth listening to just to hear him talk about starting a company. (In fact, check out this terrific 2002 entrepreneurship lecture he gave at Stanford).
Today, he talked about his idea of Hierarchical Temporal Memory, which he spelled out in the 2004 book On Intelligence. A thumbnail description of HTM is that human intelligence is built around the neocortex's ability to recognize patterns and make predictions about them, which he argues is an algorithm that can be used to create intelligent machines (the argument is far more elegant than this thumbnail).
Since he published On Intelligence, Hawkins has started a company, Numenta, to build systems based on this idea. It has an early version of a system, the Numenta Platform for Intelligent Computing (NuPIC). The company's site says:
NuPIC implements a hierarchical temporal memory system (HTM) patterned after the human neocortex. We expect NuPIC to be used on problems that, generally speaking, involve identifying patterns in complex data. The ultimate applications likely will include vision systems, robotics, data mining and analysis, and failure analysis and prediction.In other words, it's taking aim at fundamental and long-standing problems in computing, like how to get computers to reliably recognize images. It's also developing a way to train systems (and perhaps people) more effectively.
Hawkins cannot be accused of hyping his idea â€" he stressed repeatedly that it is in the very early stages. He's also careful to note that Numenta will not replace current computers. But if he's right in his approach, HTM will become an important tool for any company with complex data needs.
As rudimentary as it is, Hawkins said there are eight companies working with the NuPIC system right now, one of which, Electronic Arts, has sponsored the Numenta Challenge, which will pay people to develop a game based on a component of the system.
You can see it for yourself at Numenta's home page, where it has made a 'research release' available for download. Hawkins thinks the work happening at Numenta will be more important than anything he did in mobile computing. That remains to be seen, of course, but there's no question that data is getting far more complex, and if he's got a way to parse it more effectively, it's well worth a look now.