Google engineering director Ray Kurzweilis, undoubtedly, one of the most accomplished men of our time. The relentless inventor — whose credits include the flatbed scanner, optical character resolution and speech-to-text- systems — is also a bestselling author, a successful entrepreneur, and an artificial intelligence pioneer.
His current title at Google, then, always seemed a little puzzling to me — after all, wasn’t he the sort of guy to set his sights on something a little higher than juicing sales of online advertisements at the world's biggest Web search engine?
But the Google gig is much more than that, as I found out at a recent talk Kurzweil gave to group of tech luminaries. The event was connected to his being named as one of SVForum’s visionary award winners, an impressive group which includes Elon Musk and Bill Gates.
Kurzweil was light on specifics during the talk, in which he was interviewed by venture capitalist Steve Jurvetson of Draper Fisher Jurvetson. However, it was an interesting look at what Google’s long-term vision is — it’s not just a search company, in Kurzweil’s telling of it.
Rather, Google is trying to bring about technological singularity — creating a true artificial intelligence smart enough to improve its own source code. When that occurs, machines can progress to far-future, science fiction levels of intelligence within our lifetimes. (Kurzweil predicts this technological leap will happen by 2040.)
That might sound a little outside the remit of a Web search company, but to hear Kurzweil tell it, it’s really not. A massive amount of data on every conceivable topic, like Google has built up in its Internet index, is a “necessary, but not sufficient” step in building such an AI, he said.
The next piece is to create a system that can actually understand that data, which are often cluttered and always less organized than machine code, so that an artificial intelligence can begin to learn.
That’s something Kurzweil’s group is already working on, and if you have an Android phone, then you already have the beginnings of it in your pocket — the Google Now voice recognition system already incorporates some of the rudimentary tools that might later make true AI a possibility.
Kurzweil doesn’t see AI appearing in a vacuum, however. Rather, he thinks it will be connected directly to everyone’s brains. Neural interface technology, now in its infancy, will soon allow humans to connect their minds directly to the machine, augmenting their own natural intelligence in ways that seem like the providence of science fiction.
The example he gave: You’re at a party and someone says something to you and you try to come up with a witty comeback. Now, you’re not likely to be able to do it unless you’re particularly quick-witted — maybe only thinking of something to say hours later, to your great regret.
In the future, Kurzweil said, you’ll be able to query your connected artificial intelligence with a thought, and it will distill the perfect retort for you from the totality of linguistic info online in seconds. Thus he envisions an AI that isn’t so much a robot, but more a much, much smarter version of yourself.
Kurzweil is also involved in one of Google’s other side projects, Calico, which is about as far from the company's core search-revenue business model as possible. It's doing medical and genetic research with the goal of ending aging. It’s something Kurzweil thinks is possible to do through genetic re-engineering.
The example he gave here is mitochondria, a component of every living cell that metabolizes energy and is critical to life. Mitochondria started out as a kind of bacteria that were captured and consumed by living cells many, many eons ago, Kurzweil said. Consequently, they have their own genome separate from the rest of the body, stored in separate DNA from the cell’s nucleus.
Mitochondrial DNA is more prone to errors as the cell replicates itself, which can lead to a host of health problems. Kurzweil said that nature actually addressed this by moving much of the mitochondrial genetic code into the nucleus where it could be stored in less error-prone DNA.
But because of the way natural selection works, this process stopped before it moved some bits of the code which only come into use later in life, after a person would have normally reproduced. Kurzweil thinks humans can finish this process and solve some of the deleterious effects of aging.
So, according to Kurzweil, in the future humans will be unfathomably more smart and also functionally immortal. But he doesn’t think that humanity will change all that much, and he rejects the term “transhumanism,” which is currently in vogue to describe technologies like he is reportedly working on.
“I’ve never liked the label transhumanism, because it implies that we’re replacing humanity,” he said at the talk. “I don’t think that’s true. What we’re doing is augmenting human capability.”
Jon Xavier is the technology reporter at the Silicon Valley Business Journal.