Slashdot Mirror
Ray Kurzweil Talks Google's Big Plans For Artificial Intelligence
Nerval's Lobster writes "Ray Kurzweil, the technologist who's spent his career advocating the Singularity, discussed his current work as a director of engineering at Google with The Guardian. Google has big plans in the artificial-intelligence arena. It recently acquired DeepMind, self-billed 'cutting edge artificial intelligence company' for $400 million; that's in addition to snatching up all sorts of startups and research scientists devoted to everything from robotics to machine learning. Thanks to the massive datasets generated by the world's largest online search engine (and the infrastructure allowing that engine to run), those scientists could have enough information and computing power at their disposal to create networked devices capable of human-like thought. Kurzweil, having studied artificial intelligence for decades, is at the forefront of this in-house effort. In his interview with The Guardian, he couldn't resist throwing some jabs at other nascent artificial intelligence systems on the market, most notably IBM's Watson: 'IBM's Watson is a pretty weak reader on each page, but it read the 200m pages of Wikipedia. And basically what I'm doing at Google is to try to go beyond what Watson could do. To do it at Google scale. Which is to say to have the computer read tens of billions of pages. Watson doesn't understand the implications of what it's reading.' That sounds very practical, but at a certain point Kurzweil's predictions veer into what most people would consider science fiction. He believes, for example, that a significant portion of people alive today could end up living forever, thanks to the ministrations of ultra-intelligent computers and beyond-cutting-edge medical technology."
- out with soap!
It seems that Watson learned some bad words when IBM turned it on to the Urban Dictionary.
There goes our only chance to find out what a "holla back girl" is.
Even if your eternal existence is as a glorified chatbot doomed to bulk Google+'s userbase for unbounded time?
I'm slightly joking; but in all seriousness that's the aspect of the optimistic school of techno-rapturists that I find least plausible. Given enough time(probably more time than any 'futurist' writing today has, sorry about that...), will we achieve a variety of medical techniques that would seem nigh-miraculous today? Assuming the cheap energy doesn't run out, sure, seems reasonable enough.
However, consider diarrhea: it's an unbelievably banal disease, mostly a product of poor sanitation, and can be managed by barely-trained care staff with access to dirt cheap oral re-hydration solutions. It kills something north of two million people a year, mostly children; and nobody really gives that much of a fuck.
When people die like flies because nobody cares enough to provide them with what is basically a salt/sugar solution, how well do you think your "Brother can you spare some unobtanium medi-nanites?" appeal is going to work? Or your plea for enough CPU time to continue being conscious?
Sure, you can wave your hands and talk about 'post scarcity'; but unless some magic parameter limits the size of the singularity's AI agents, why would they accept less compute time when they could have more and be smarter still? Are you planning on staking a moral claim to your CPU time? Outwitting a superhuman AI? Dancing for the amusement of your robot overlords?
Ray Kurzweil is no doubt a brilliant thinker and an engaging writer/futurist - I've read some of his books (admittedly, not "Singularity"), and they are fun and thought-provoking. However, disciplined and realistic they are not - his main skill is in firing our imaginations rather than providing realistic interpretations of the evolution of technology.
My favorite case in point is his elevation of Moore's Law into a sort of grand unified theory of computing for all time, and using some very dubious assumptions to arrive at the idea that we'll all have merged with machines into immortal super-beings within the near to mid future. I don't need to pick apart all the reasons why this is fallacious and somewhat silly to treat as a near-term likelihood - the point is, he's basically a sci-fi writer in a lot of ways, and I read most of his statements in the same spirit as I'd read a passage out of "Snow Crash."
That said, Google has some very capable people, and can, in all likelihood, mount our best attempt at human-like intelligence to date. They'll push the envelope, and may make some good progress in working through all the challenges involved, although the notion that they'll create anything truly "human-like" is laughable in the near term.