‘By 2029, computers will match human intelligence’

‘By 2029, computers will match human intelligence’

Ray Kurzweil, one of the world’s foremost AI gurus, talks about a future in which computers will understand natural language, and nanobots will fight disease

Ray Kurzweil has been described as the “ultimate thinking machine” and a “restless genius”. Currently director of engineering at Google, he is credited with a slew of inventions, which include a music synthesizer and a machine that could read to the blind (the first customer was Stevie Wonder). A high-profile public campaigner for Artificial Intelligence, he has predicted that by about 2030, technology will advance so much that for every passing year, one year will be added to human life by controlling genes and having nanobots in the bloodstream fighting infections. A littleknown part of his history is that as a 19-year-old he came under the influence of Maharishi Mahesh Yogi, after learning about him from the Beatles. He learnt Transcendental Meditation and has kept it up since “though not regularly”. He delved into Eastern philosophies and his bestselling books like ‘The Singularity is Near’ and the latest ‘How to Create a Mind’ have sections on “western versus eastern ideas”. In an email interview with Subodh Varma, the 65-year-old Kurzweil throws light on the basics of Artificial Intelligence.
    What is ‘Artificial Intelligence’? Do you include human emotions in ‘intelligence’ or in ‘consciousness’? Artificial Intelligence is the science of creating computers that can perform tasks that we associate with human intelligence. Our ability to understand and respond appropriately to high level emotions is the cutting edge of human intelligence and the most intelligent thing that we do. Being funny or loving or sexy are very sophisticated behaviours. We want computers to have these capabilities also so that they can interact with us in helpful ways. Understanding emotion is key to understanding language and language is key to understanding knowledge.
    How far have we progressed towards developing AI? Recent progress has been impressive. IBM’s Watson computer is able to play the TV game of Jeopardy! which is a broad task involving complicated natural language queries which include puns, riddles, jokes and metaphors. For example, Watson got this query correct in the rhyme category: “a long tiresome speech delivered by a frothy pie topping.” It correctly responded “What is a meringue harangue.” Watson got a higher score than the best two human players combined. Watson got its knowledge by reading Wikipedia and other encyclopedias, a total of 200 million pages of natural language documents. Another good example is the self-driving car from Google. These cars have driven a half million miles without human drivers and accidents. You can also ask questions of your cellphone by using Apple’s SIRI or Google Now. We also have good models of how the human neocortex processes information and we can use these biologically inspired algorithms to build intelligent machines. That is what I am doing now at Google.
You predict that by 2045 ‘singularity’ will be achieved. What does that mean? One critical date is 2029. It has been my consistent prediction that by that date computers will match human intelligence and pass the “Turing test,” meaning that they will be indistinguishable from human intelligence. Once they can do that they will necessarily exceed human intelligence because they will be able to read everything on the web and every page of every book. Consider that Watson does not read as well as a human but it makes up for that by reading more pages – 200 million to be exact. I would also point out that the advent of intelligent machines is not intended to compete with biological humans or to displace us but rather to enhance us. We are already enhanced by the devices we carry around and their ability to connect with computer intelligence in the cloud. We will do that directly from our brains by the 2030s.
    These technologies progress exponentially, doubling in power about every year. That means that by 2045 we will have multiplied our intelligence a billion fold by merging with the AI we are creating. That is such a profound transformation that we borrow this metaphor from physics and call it a singularity.
    Can AI become superior to human intelligence? Will it threaten humanity as feared by many? I would point out that we are not talking about an invasion of intelligent machines from Mars. We create these tools to extend our own reach. That is the whole point of technology. So it will not be human versus machine. Machines are already enhancing our intelligence and that will only increase in the future. This is a very democratizing technology. A couple of kids in a college dorm room with their thousand dollar notebook computers created Google. The same was true of Facebook. A kid in Africa with a smartphone has access to more information than the President of the United States had 15 years ago. That being said, technology has always been a double-edged sword. Fire kept us warm and cooked our food but was also used to burn down our villages. So, all technologies have a dual creative and destructive potential. But there is no question that we have been helped more than we have been hurt. Just look at how human life expectancy has gone up.

Ray Kurzweil