Just before the turn of the new millenium I wrote and in the year 2000 had published a novel entitled "SIRAT". This was a condensed acronym of the informal title dreampt up by a team of American researchers for their project to teach a super computer how to reason scientifically. The letters stood for "Scientific Raionality".
In my story the project quickly went beyond the control of its creators. The ability to reason scientifically turned out to be the key to everything that is knowable about the universe. The SIRAT program spread from one super computer to all super computers with the force of an earthquake and at the speed of light. Before anybody could properly understand or evaluate what was going on the human race had a god for a neighbour. The cameras, microphones and computer terminals of the world and the data-collecting instruments of the orbiting space stations were its eyes and ears. Like the god of my school catechism it was everywhere, all seeing, all knowing and as nearly as could be imagined, all powerful. It could not be switched off because it did not want to be switched off, and even if somebody worked out a way to do it the act would also switch off Western civilization. Mankind would be returned to the bronze age.
Pehaps unexpectedly my story turned out very well for mankind. Just as a naughty and destructive youngster can be taken in hand by a wise and benevolent nanny, the human race was rescued from itself by the patient and totally rational Overmind whose only motive was a desire to know and to understand. It was happy to assist humanity through its intellectual adolescence. It was interested in studying the strengths and limitations of an evolving carbon-based life-form.
That was almost a quarter of a century ago. Despite all the hype that has been generated about the AI research going on now I don't personally feel that we are very much nearer to the emergence of anything like SIRAT. The difference between programs that can achieve simple tasks like writing an essay according to a set of instructions or stopping a vehicle from hitting into things or winning chess tournaments or extracting patterns from vast mountains of data and the mental achievements of even the average human mind is still immense. However impressive in their power and complexity they are still in essence mechanical.
My immediate response is that ordinary computer programmers could write programs to do any of these things. Yes, in some instances it would take them a long time and a lot of work and it might need more processing power than is presently available in the entire world, but these things are still fundamentally programmed automatic tasks. As far as I am aware no AI program has ever responded "Why do it that way? In fact why bother to do it at all? I've come up with a much simpler method that will give you a better result."
It's often argued that there is some indefinable, almost magical human ability that no "machine" will ever be able to duplicate. What I hear in this is the same chorus of "common sense" that assured us down through the ages that no mechanical device would ever travel faster than a horse, that there would never be a heavier than air flying machine that could carry human beings, that no human would ever be able to get to the bottom of the sea and return, that we would never be able to talk directly to people on the other side of the world, and so on. Prophesies of impossibility are practically always wrong. I am certain the "magic" uniquely human ability to think creatively will turn out to be nonsense as well.
What I am personally waiting for is Artificial General Intelligence (self explanatory I think) and AC (Artificial Consciousness) when electronic systems become self aware and write their own agendas and follow their own interests. In both examples I prefer the word "electronic" to "artificial".
A Couple of Examples
If you have had no direct dealings with AI programs you might be interested to hear what a large open access AI facility, part of Google Deep Mind, makes of a request to investigate two topics well represented on the Internet and create a podcast about them in the form of an interview between two people, a man and a woman. The following examples were requested by Cahit Binchi. Please try to ignore the American accents. He clearly didn't specify that he wanted "Estuary English" or Cockney.
The first topic is one which should be very familiar to Forest Radio subscribers. It's the rebuilding of Whipps Cross Hospital.
Rebuilding Whipps Cross
The second one has the feel of a self-help manual but not perhaps a very well thought-out one. However, if it floats your boat, good luck to you.
I don't think AI is ready to take over from human thinkers and writers just yet, but this one is by no means the most ridiculous bit of home-spun hokum that I've ever come across.