Never say never with machine intelligence

Humans balk at the notion of any form of machine-based intellectual superiority - but we should embrace it

As long as I can remember, robots and AI have been universally perceived as a threat, and this remains the case despite the converse being blatantly true.

In reality, societies are critically dependent on robotics and AI for food, energy, technology, products and services including all transport and utilities.

Machines design electronic circuits and products, and assemble and test all our consumer goods. Chip technologies invisibly populate all our vehicles, homes and offices, with every appliance subsuming a degree of intelligence and robotics. So where does all the fear come from?

I think we can safely assume that decades of dystopian scenarios featuring rogue machines conjured by Hollywood for cinema and TV have had a big part to play.

This is illustrated nicely by introducing a little child to a robot. They immediately see a potential friend and quickly settle down to play. Repeating this experiment with young adolescents sees a very different reaction because they have become conditioned by a general media paranoia.

Another dimension to the threat scenario involves a question of intellectual superiority and capability. It appears most humans balk at the notion of any form of machine superiority.

Never say never with technology

Strangely, people do not appear to worry about having muscle power subverted by a hammer, screwdriver, spanner or back hoe! But, a machine that beats them at chess, poker or Go is an entirely different matter!

This is manifest in the continual stream of ‘never' utterances supported by philosophically ill-informed arguments and statements going back more than 100 years, all of which have been eclipsed by reality.

Some of my favourites are: a machine will never play a game of chess/poker/Go; they will never read a book or play music; they cannot possibly debate, reason, comprehend or think; machines will never be truly intelligent…

This endless list is often augmented by: never beat a grandmaster; never be creative; never beat human intellect; or never become sentient.

And as each of these hurdles (or ideals) falls at the altar of tech progress, new ones are created, with sentience being the latest obsession scheduled to fall.

In reality, our species has reached the limit of its evolution as a general purpose biological entity. In contrast, our silicon-based counterparts are on an exponential racetrack of specialised applications far exceeding human abilities.

This ought not to be a surprise as we have been here before with analogue and digital computing. During the 1960s, all computers were task-specific - for artillery targeting, payroll processing, industrial control, banking, government data, and so on. Further, these machines were isolated with no network connections.

Fast forward to today and the vast majority of machines are general purpose and networked by the millions.

Face the future

The next phase for AI and robotics has probably been telegraphed by the AlphaGo victory over the human world champion (2016). This came about with the Alpha software moving to self-programming, and in a lesser sense, self-determination. Interestingly, AlphaFold2 similarly exceeded 90 per cent molecular-structure-to-protein folding prediction accuracy in 2020.

The implications here are profound as human progress in this field had been essentially static for 50 years. For human health alone this is the biggest breakthrough since the characterisation of DNA.

Where does this leave us? At best; a massive upside for new human-machine partnerships improving life and doing good including smart replacement organs and implants for the ailing. And at worst; AI and robots dedicated to killing us, but only because we were dumb enough to instruct them to do so!

But hey, we lived with the atomic bomb for 75 years, so why not AI?

Peter Cochrane OBE is Professor of Sentient Systems at the University of Suffolk