Ray Kurzweil, author of The Age of Spiritual Machines and a pioneer of artificial intelligence software, has always been one of the most provocative thinkers on technology and its future. When he spoke at the Demo conference over in Santa Clara last week, it was no surprise that he covered everything from why computers will continue to get better at an exponential pace, to how we will be able to expand our brains into the cloud with the help of biological devices in our bloodstream.
Promoting his forthcoming book, How to Create A Mind, Kurzweil spoke about technology but soon moved to a discussion of the brain, how he thinks it works, and how we will be able to enhance it.
Artificial intelligence is making discernible improvements in many areas, Kurzweil said. IBM's Watson used its "total recall" of 200 million pages of natural language documents to win the Jeopardy! challenge. It wasn't as accurate as people in understanding individual pages, but was successful because it could digest so many more documents.
Speech recognition is working quite well, he said, noting that people are critical of things like Siri, but that it's pretty amazing that people are talking to their computers at all. Google's self-driving cars are also doing very well, he added.
He defended his theory of exponential growth (popularised in The Age of Spiritual Machines) saying that compute capabilities have actually been following this path since the 1980s. While critics say silicon scaling (known as Moore's Law) can't continue forever, that's actually the fifth paradigm to bring exponential growth in computing. He expects self-organising 3D transistors will be the sixth paradigm.
He was most animated, though, when talking about progress as "reverse engineering the brain." Improvements in technology such as MRI spatial resolution have led to a better understanding of how the brain works. He espoused a thesis about the uniform structure of the neocortex, saying it is made up of 300 million undifferentiated "pattern recognisers" in a hierarchical structure. The difference in the amount of pattern recognisers compared with other animals is exponential, giving humans enough capability to create art, science, and literature.
Our brains develop those 300 million modules at a very young age. One reason kids can learn language or music so easily is that they haven't filled up the modules, he said, but by the time we're 20, we've filled them up. Therefore, we need to be able to remove things intelligently. People who are rigid will have difficulty doing that, he said, but you can learn new material at any age if you are able to move on and forget other things.
Expanding brain power via the cloud
Still, Kurzweil said we have a limited capacity, but he is optimistic we can overcome this by "expanding our brains into the cloud." Techniques that evolved in the biological brain are the same that are used for things such as speech and character recognition, and they will be used to expand our brain power.
Search engines already act in this way for many people, he said. As a result, we are now smarter and both individuals and work groups can do more.
Kurzweil also likes the potential of Google Glass (pictured right) to do things like identify people you meet, give you directions, and basically listen to your conversations to give you information that "overlays" the real world.
"You'll just get used to an assistant helping you," he said, calling such things "mind expanders."
In the long run, he doesn't think we'll have hardware implants in our brains, but rather biological devices that live in our bloodstream will give us more capabilities. These will eventually be able to functionally recreate the pattern recognition ability of the brain. This, he believes, will lead to a qualitative leap in understanding, similar to the jump between apes and humans with the expansion of the neocortex.
This might be 25 years off, he said, but already there are a number of medical devices that can be put into the blood. In the meantime, we'll all have more intelligence from our devices, even if they are not physically inside our bodies.
Asked by Demo host Matt Marshall about the potential downsides, Kurzweil said technology has always been a double-edged sword. Fire has been used for good and bad, he said; artificial intelligence can be as well. He noted that AI is now very widely distributed. "A kid with a smartphone in Africa has access to more information than the president of the United States did 15 years ago," he said.
If you have an AI that is smarter than you and it turns on you, you're in trouble, Kurzweil said, unless you get an AI on your side that is even more intelligent. This isn't an issue in AI today, but is an issue in biotechnology. Software viruses have gotten more and more sophisticated, but they haven't shut down the Internet or stopped people from using computers. Instead, we have an evolving technological immune system that is "more or less working," he said.
Of course, ideas like this have been played out in science fiction for years, but many more of them may soon become reality. Kurzweil may be wrong in some specifics, but he's always entertaining to listen to.
Michael J. Miller is Chief Information Officer at Ziff Brothers Investments, a private investment firm. Mr. Miller, who was editor-in-chief at PC Magazine from 1991-2005, authors this blog for PC Magazine to share his thoughts on PC-related products. No investment advice is offered in this blog. All duties are disclaimed. Mr. Miller works separately for a private investment firm which may at any time invest in companies whose products are discussed in this blog, and no disclosure of securities transactions will be made.