Kurzweil: Brains will extend to the cloud
SANTA CLARA, Calif. - Human brains will someday extend into the cloud, futurist and computer pioneer Ray Kurzweil predicted at the DEMO conference here on Tuesday.
Moreover, he said, it will become possible to selectively erase pieces of our memories, while retaining some portions of them, to be able to learn new things no matter how old the person is.
"The brain doesn't grow much from a very young age," he said. Humans have, more or less, 300 million pattern-recognizing modules in the neocortex, the portion of the brain where thought occurs. "One of the reasons kids can learn new languages, or pretty much anything, so quickly is because they haven't filled up those pattern recognizers," he said. "It's virgin territory."
I think we'll be in augmented reality all the time. Raymond Kurzweil, Computer pioneer and futurist
(Well-known author, entrepreneur and futurist Ray Kurzweil talks about mapping the human brain, the continuing evolution of technology and the rise of artificial intelligence from the DEMO fall 2012 show in Santa Clara, Calif.)
"You can learn new material at any age, but there is a limited capacity. That's one of the things we will overcome by basically expanding the brain into the cloud," he said. "We need to be able to repurpose our neocortex to learn something new. People who have a rigid process and hold onto old information; they will have a hard time doing that. You need to be able to move on."
While Kurzweil did not give a timetable for these predictions, he said the notion of "brain extenders" has already begun thanks to technology including IBM's Watson supercomputer and augmented reality. "I think we'll be in augmented reality all the time," Kurzweil said. "Just to be told what people's names are [via augmented reality popups] -- that would be most helpful," he said, to laughter from the crowd. "That's a killer app."
As for Watson and its ilk, Kurzweil said, there has been "discernible progress" in artificial intelligence. But it's one thing to recall, say, anything from Wikipedia or other encyclopedias. "But understanding is way below human levels," and that's the next frontier, to help technical assistants understand the real meaning of human speech, a breakthrough he's already predicted will happen in 2029.
Advances in natural language processing will help with that, Kurzweil said. It's an area he knows about, having created the industry's first multi-font optical character recognition system, as well as text-to-speech synthesizers and other breakthroughs. His company, Kurzweil Computer Products, through a series of acquisitions and mergers ultimately became part of Nuance Communications Inc., which developed technology that was much of the basis for Apple's Siri.
"The natural language understanding in Siri is fairly weak and needs a lot of improvement," Kurzweil said. "For a version 1 product it's pretty good; usually version 1 doesn't work at all."
To really work like a human brain, artificial intelligence (AI) tools need to be built hierarchically, like a brain. "And then you have to educate the synthetic neocortex" as we do with newborn babies, he explained. That's what Watson's achievement was -- an educated brain extender, up to a point.
As with many other things, AI has the ability to be misused in the wrong hands, Kurzweil acknowledged. "Is fire a good thing? It keeps us warm and cooks our food, but it's also used to burn down our villages.
"If it turns on you, you need to get even smarter AI."
But, he said, "I think we could take comfort with how we've done with software viruses." Although these viruses have become increasingly sophisticated over time, "we have an evolving technological immune system that's kept up with it, more or less. You can have arguments about it, but I think it's working."