Doctor Dark’s Diary – Episode 39
Artificial stupidity – a doctor writes...
As a doctor, I am often asked, “Doctor, is artificial intelligence possible ?” (I would
just like to apologise to Private Eye for borrowing one of their better jokes, and to
all the real doctors out there. I am not a doctor at all, so please don’t send me long,
whinging letters about your unpleasant condition, unless it is really hilarious and I
can get a laugh out of it.)
I contend that AI is possible. I would be silly if I claimed it was going to be easy, or
even soon. But it is possible. Machines can be made which will think. I nearly slipped
up and wrote “think like you or I” then, but that would have led to a trap. The way
we think has evolved over a long period, and is very complex, to say the least. The
brain is not just an electrical circuit, it also carries out chemical reactions that are
part of its processing.
As if it were not already complex enough, there is now evidence that the tiny “wires”
between the cells of the brain reconfigure themselves as part of the way the brain
works. So what was referred to in the science fiction of thirty years and more ago as
“a giant electronic brain” is unlikely to be built to work in the same way as an actual
brain. Even now, we don’t have the technology to make a machine that reconfigures
itself, is apparently all processor, and can build Nascoms for fun.
But we don’t need to build anything at all. We already have computers that can carry
out simulations of all sorts of things, including the operation of complex electronic
circuits, and chemical reactions. Even the reconfiguring trick could be reduced to
a self modifying program, or perhaps it would only be necessary to alter the data
base holding the description of what was being simulated.
Research into neuronal computer circuits has shown that learning can be carried
out by surprisingly simple machines. Simulating them in software on an ordinary
computer takes a long time because of the von Neumann architecture of these
machines, but more and more computers with many processors are appearing, and
on these, the simulations would run rather a lot faster. Basically, the above explains
why I believe that machines will one day be able to think.
In the meantime, machines with all the intellect of a stunned earwig can do a
surprisingly good imitation of thought, and apparently hold conversations that some
people find convincing. The program “Eliza”, which was also known as “Doctor”,