Friday, January 5, 2018

AI SPECIAL ....WHY AI WILL NOT DESTROY MANKIND

WHY AI WILL NOT DESTROY MANKIND

Our Civilisation, Not Individuals, Will Create Superhuman AI, And It Will Contribute To Human Growth The Way Gifted People Have Always Done

An IQ over 150 occurs only once in a thousand persons, yet there are about 70 lakh such persons in the world. What can such high IQ achieve? Possibly a lot, for James Watson, who co-discovered DNA, has an IQ of only 124. But if you are still curious, Stanford University’s ‘Terman Study of the Gifted’ has been tracking a set of high-IQ individuals since 1921. What does it show? Many of these “exceptionally gifted” subjects landed up in occupations “as humble as those of policeman, seaman, typist and filing clerk.”
Successful? No. Dangerous? Hardly. The Stanford study showed that raw intelligence or cognitive ability does not amount to much by itself. Mowgli — even if his IQ were higher than Einstein’s — could not have figured out the theory of relativity living among wolves. Yet, many are afraid of artificial intelligence today. Tesla boss Elon Musk is one.
They fear an ‘intelligence explosion’ that will create machines capable of creating even more intelligent machines in an endless cycle that will wipe out human beings, like the Terminator cyborgs. Is it possible? Francois Chollet, an artificial intelligence researcher at Google, says ‘no’ in a wellargued post on Medium.
Chollet says our AI-phobia is based on the idea that one day we will create a ‘seed AI’ that is smarter than the smartest humans in every field, including the creation of more AI. However, this fear is based on a flawed understanding of intelligence. The brain is not intelligence but mere biological tissue. What we call ‘mind’ is linked to the body and other things like culture and environment.
For example, if you implant a human brain inside an octopus, it wouldn’t know how to control the eight tentacles because it is hardcoded to use two hands. The important point here is that “general intelligence” does not exist. All intelligence is specialised, but artificial intelligence is ‘hyper-specialised’. If you have a selfdriving program, don’t expect it to make coffee.
Increasing intelligence, Chollet says, is not a matter of tuning a brain or a program, just as you cannot increase the output of a factory by speeding up its conveyor belt. Our problem solving ability is not limited by IQ or computing power but circumstances. Newtons and Einsteins do not arise in a vacuum; their work is “usually an incremental improvement over the work of their predecessors.”
In fact, only a small part of our intelligence resides in the brain. Most of it is spread about us as ‘cognitive prosthetics’, such as a dictionary, or if you are a lawyer, old case histories. Even language is a storehouse of our collective intelligence.
If it were really possible for human or artificial intelligence to suddenly create a greater intelligence than itself, one of the billions of human beings would have done it by now. Chollet says human civilisation, not individuals, will create a superhuman AI, and it will contribute to our civilisation’s growth just as the smart ones among us have done over millennia.
For more: Medium


No comments: