Covering Disruptive Technology Powering Business in The Digital Age

Home > DTA news > Let’s stop calling AI intelligence.
Let’s stop calling AI intelligence.
January 19, 2017

Artificial Intelligence has captured the imagination of science fiction writers since Frankenstein’s monster woke up confused and lonely, and more recently has hit the news as an impending and inevitable cataclysm. The public have been told that their jobs are threatened, their places in society will be usurped and their lives may be snuffed out by this rising tide of intelligent super beings. Far be it from me to argue with the educated elite but I think it’s time we stopped giving ourselves so much credit. We are nowhere near creating actual artificial intelligence. What we, the great unwashed, determine as intelligence would be something self-aware of its surroundings with the ability for independent thought and creativity. What those in the field of AI R&D consider artificial intelligence is somewhat different.

Intelligence is an emergent property. It has nothing to do with the ever increasing computing power laid down by Moore’s law. Specific emergent properties cannot be fashioned, only humbly discovered through observation, and so to say we are creating intelligence is a misnomer. An individual brain cell is just a bag of chemical reactions. A hundred, thousand or a million such cells is just a larger collection of such bags akin to an ever increasing number of computer processors. The number of cells doesn’t determine the intelligence of a system, it’s how these cells interact with each other in a complex series of feedback loops that creates its emergent property. And not just any complex feedback system will develop these properties as the bulk of natural non-linear dynamical systems attests to. Only the very rare, possibly infinitesimally small, number of instances will develop such emergent properties. Only our brains have come up with a substantially complex emergent output that we can deem intelligent. There are no other instances of such complex emergence in the known universe and we think we’re going to create one?

What are the trophy’s that the AI community holds up as their proof? Watson beat some people at Jeopardy. Was this amazing? I think not. Jeopardy is a trivia based game. If I went onto that show with access to the internet I would probably win too. This is all Watson really did. It searched the internet, or its large database of previous jeopardy games plus the internet, and came back with the answers. It used some clever natural language processing to form the answers yes but ultimately it won by google searching the answers. Not intelligence. Next we have Deep Blue which beat Kasparov in a chess match. This was lauded by those in the AI field as a massive leap in algorithm design (which I agree with) and proof that they had developed an intelligence. Well, this is just plain wrong. What really happened was that Kasparov and Deep Blue played twice. Kasparov won the first one; Deep Blue won the second one. I’m no genius but this sounds like one all, a draw. The reason Kasparov lost was not because of inferior intelligence but through the human quality of making a mistake and Deep Blue capitalising on this. Deep Blue did not make a move that out-witted Kasparov, just capitalised on his mistake. Again not intelligence. A deciding match has yet to be played and despite Kasparov’s request to play again he has been refused.  There are other instances of advanced AI being held up and lauded such as in the medical field but again these are just advanced data mining, not intelligence.

With this in mind I think we should stop giving ourselves so much credit. We believe that within less than a hundred years of first developing the computer, or microprocessor, we have come so far and understand complex systems so well that we are within a lifetime of birthing a new being, a new silicon based intelligence. Let’s just calm down and take stock of what we actually have. What have we done to date is nowhere near what we claim we have done. AI, through machine learning type algorithms, is just a fancy way of regurgitating the facts and figures from a large database. Yes, an algorithm can guess at what may happen in the future based on ever increasing levels of regression and associative formulae but all they’re really doing is learning what has happened based on historical trends. This is not intelligence. There is no innovation or creativity beyond its programming here. This is just the ability to remember and guess an outcome based on this. This is great news for business and other industries as, for example, it points us in the right direction as to what customers may buy based on previous trends but intelligence is far more than this. Intelligence is creativity, innovation and more than the sum of the individual parts and until AI becomes more than the sum of its parts it will forever remain just an advanced calculator.

Niall Wynne