Intelligence is the medium of interest. What we want, what we don’t want and what we want to keep are, more or less, factors determined by intelligence. We construct things out of intelligence to maintain and safeguard interest. The greater the interest, the more intelligence we pour into accomplishing it.
Intelligence, at the heart of it, is information processing, so as we keep developing better and novel ways of information processing, we are also constantly re-baselining our interests. That is to say, interest changes over time.
So when we talk about an intelligence that is not human -an artificial intelligence. We are essentially talking about finding a set of interests based on a different time-construct.
How the advent of AI will impact humans, is then contingent on that difference.
“It could be terrible and it could be great, it’s not clear, but one thing is for sure, we will not control it.” Following the established premise above, what this quote from Elon Musk means is that where our and AI’s interest overlaps prosperity abounds. And when it diverges, there will be chaos followed by destruction.
To fully appreciate the scale of either prosperity or destruction, we can contrast the relative timeframes that governs human intelligence and the most supreme artificial intelligence -Artificial Super Intelligence.
A few years back, we created AlphaGo -an Artificial Narrow Intelligence (AI that serves a single purpose), to play Go.
Unlike chess, Go is extremely more complex, there exists 10^170 possible moves in a single game of Go. To give a point of reference, there are only 10^80 atoms in the observable universe. AlphaGo was able to learn from and use data from over a million games of Go to beat world champion, Lee Sedol, 18 times. What was most impressive was that AlphaGo made up new moves that no one had ever seen before.
This was a hugely impressive feat in the technological community as it signified that machines could learning. Similar to how a child takes in information and organize it in different parts of the brain, AI with Machine Learning could do the same, only countless times faster.
With the development of Machine Learning, only a year after AlphaGo beat the world champion, it was beaten by a brand new AI called AlphaGo Zero 100 consecutive times. AlphaGo Zero had learned to play Go by playing itself over and over again. And in just 40 days, it was able to surpass thousands of years of knowledge.
This means that in the game of Go, there exists more AI intelligence than there is in human. And when we are able to truly develop Artificial General Intelligence (AGI) -intelligence that is able to perform any intellectual task that a human can- AI intelligence will become the dominant intelligence in the world, with human intelligence seeming infinitesimal in comparison. This is because Artificial Intelligence process information at a much faster pace than human beings can. While the biological neutrons in our brain transfer information at the rate of 100m/s, AI neural network can transfer information at the speed of light -that is, 300 million m/s. In one week, a AI computer can do 20,000 years worth of human research, simulation or anything related to information.
AI would accumulate information and learn at an exponential rate -it would get smarter and smarter at increasingly shorter and shorter time intervals. This exponential learning curve is an inevitable outcome of our current technological advancement, one way or another, irregardless of when we get there -as long as we don’t destroy ourselves beforehand- humankind will give rise to Artificial Super Intelligence (ASI).
The creation of ASI signifies the coming of a ‘technological singularity’, a point in time when artificial intelligence surpass human intelligence by so much that humans wouldn’t be able to fathom it. The term ‘singularity’ is something we attributed to the centre of a black-hole, a point of infinite unknown in spacetime. We would have as much ability to understand that kind of intelligence as ants would be able to understand us humans.
At that point, as long as our primitive human interest doesn’t get in the way of the AI overlords, our existence may be spared. But should our anthills get in the path of the AIs, we will be stomped on indiscriminately, and there won’t be any mercy to speak of.
By Gray L.