Numenta And IBM To Build Biologically Inspired Intelligent Machines.
When we catch balls, Jeff Hawkins, cofounder of Numenta and author of “On Intelligence,” tells us we aren’t solving differential equations. A robot, on the other hand, does solve differential equations, requiring roughly 3-trillion calculations for a 1s toss (“Kinematically Optimal Catching a Flying Ball with a Hand-Arm-System,” Berthold Bauml, Thomas Wimbock and Gerd Hirzinger, Institute of Robotics and Mechatronics, 2010).
There’s a big difference between intelligence and intelligent behavior Jeff would also have you note. Deep Blue, the IBM computer that beat chess grand master Gary Kasparov back in 1996, and IBM’s Watson, which defeated two Jeopardy! champions back in 2011, displayed intelligent behavior, not intelligence. When you stop feeding these machines their specialized input, their intelligent behavior ceases. On the other hand, when you lay awake in the warm darkness of your bedroom with your eyes closed, your mind continues processing unabated, thinking, musing, possibly stumbling onto some deep aha moment. Part of the reason you keep ticking in the dark is that there are, in fact,pacemaker-like neurons always active in your brain.
This innate lack of intelligence is the problem underlying all of the current machine architectures generating all of the hype in AI. Whether it’s the work of Google’s Professor Geoff Hinton, the work of Professor Fei-Fei Li at the Stanford Vision Lab, or the work of Professor Yann LeCun at Facebook’s AI lab, the output of their machines is intelligent behavior without actually harboring any actual intelligence. The machines and their algorithms are as innately intelligent as the cursor at the end of this sentence.
“On Intelligence” makes the compelling case that a more proper way of building intelligent machines is by copying the salient features of the mammalian neocortex, an auto-associative, predictive memory pattern processing machine. The recent partnership between IBM and Numentatells us that Jeff’s ideas have traction.
See on science20.com