Peter Cochrane: Fish can't climb trees and AI won't eclipse humanity

AI will steer our evolution but not take over

Science doesn't care what you believe, and AI doesn't care what you think.

In 50 years, we will look back at the endless AI debates and how it doesn't think like us, is not self-aware, has no soul, no consciousness, or indeed any originality of thought and creation.

Today I am mildly amazed at the level of discussion and irrelevant debate. For the most part, AI is confined to physical boxes connected to networks with no resemblance to human physiology. And it is apparent that people do not realise that the human brain is not entirely contained in the head; indeed, if it were, we would not have survived!

A distributed brain in the form of a nervous and sensory system distributed throughout the human body is essential to affording us a perspective we define as our ‘awareness'. This will be hard to artificially replicate without making our AI mobile and similarly equipped.

See also: AI doesn't care what you think

For 50 years, I've watched the arguments and discussion progress: computers will never play a game of chess; computers will never play a good game of chess; computers will never win against a grandmaster; computers don't play a conventional game of chess and so on! This liturgy has variously been followed by computers will never play poker, GO et al.

But AI has now exceeded all these and far more in decoding more than 200 million protein folding patterns in three months. In contrast, humans have only been able to decode a handful of simple variants over several decades.

This is but one of many spheres, including drug design, where fundamental human inabilities are sidelined by AI, which is also demonstrating features that we can never realise.

Rather than seeing all this as some form of threat, or affront to humanity, we have to look at our relationship with AI (and indeed robots) in a more symbiotic fashion.

The only important questions are: what can we do well that robots and AI cannot, and what can robots and AI do that we do poorly or are incapable of doing?

At present we see robots and AI emulating and learning from us, with the interesting prospect of us learning from them! Evolution always builds bottom up and never top down, and she does so to ensure the survival of species in a given environment.

AI amplifies intellect and ultimately it will steer our evolution

So, trying to compare the relative intelligence of different species is therefore fundamentally erroneous! In the way a fish can't climb a tree or read a book, a man cannot survive alone at sea.

I therefore rail against IQ and Turing tests claiming to show the relative intelligence of species and the comparative AI progression. Projections for the eclipse of the human race are also meaningless as this is already happening a domain at a time.

For example, we cannot design leading-edge integrated circuits, nor can we assemble mobile phones, or indeed automobiles, and in many instances we can't perform medical operations with the dexterity and accuracy of our machines.

It is far more powerful to think of robots and AI, and indeed all technologies as tools, aids - and friends. I regard a screwdriver as a muscle amplifier, whilst AI amplifies intellect, and ultimately it will steer our evolution.

Instead of panicking about what the technology might do to us, it would be wiser to consider what we might do for, and with, it! In reality, we've been here many times before and experienced the same thinking and debates.

The arrival of the first task specific electronic computers saw similar panic reactions (~1950 -1970) and here we are again with relatively simple AI empowering the human race to do and achieve more.

Peter Cochrane OBE, DSc, University of Hertfordshire