Ultra-realistic artist robot
Ultra-realistic artist robot ‘Ai-Da’ is on display at the world’s largest gathering of humanoid AI robots as part of International Telecommunication Union AI for Good Global Summit in Geneva on July 5 2023 © Fabrice Coffrini/AFP/Getty Images

Some people have expressed fears that artificial intelligence might supplant humans altogether (“Apple’s belated AI gambit”, FT View, June 13; and “AI should not be a black box”, FT View, May 31). I assume that “intelligence” is a definite quantity that entails understanding the external world; otherwise, these fears are vacuous. They are anyway misplaced.

There is a general principle that no entity can design a device more intelligent than itself. The proof is that supposing otherwise leads to a paradox. The entity both would and would not understand some aspect of the world understood by the device: would, because it could design the device, but would not, because it is less intelligent than the device.

This is an absolute limit to the abilities of AI. In practice, the limit is likely to be more stringent unless the designers of AI are the most intelligent people in the world. If not, there will be many people more intelligent than the most intelligent device yet constructed.

A device might be faster than a brain, because electrical impulses are transmitted much faster than nerve impulses. It might also have a greater capacity for memory, because it can be made indefinitely large. Computers are often larger and faster than human brains, but they are never more intelligent.

A device more intelligent than its creator might arise by chance, as a couple might have a child more intelligent than either of them through the accidents of genetic recombination. It might also arise through natural selection acting on random variation (which is how our brain has evolved), but this is not design in the normal sense of the word and no such process seems to have been generally used in the development of AI.

It may well be feared that AI will be misused by any organisation rich or powerful enough to deploy it. It need not be feared that AI will ever be more intelligent than us.

Graham Bell
Emeritus Professor, McGill University, QC, Canada

Letter in response to this letter:

Tech needn’t be cleverer than we are to sup­plant us/ From DA McM Wilson, Boston, MA, US

Copyright The Financial Times Limited 2024. All rights reserved.
Reuse this content (opens in new window) CommentsJump to comments section

Comments