November 5, 2009
By Rohit Talwar
CEO
Fast Future
Rohit Talwar
LONDON, UK, Nov. 5, 2009 — Artificial intelligence continues to have ever-greater impact on our everyday lives, even though it is mostly behind the scenes performing tasks such as improving search engines, recommending music, finding the fastest route and so on.
On a good day, an AI system can make as good a medical diagnosis as a General Practitioner. Indeed, many GPs now check their own diagnosis against an on-line expert system (a fairly basic but effective form of AI).
However, doctors are still needed because their own human skills are able to obtain more valuable data from the patient by asking questions and noticing body language signals such as hesitancy, facial expressions and so on. This extra data can lead the doctor down new channels of enquiry and improve diagnosis.
Airport security, too, is starting to use AI to spot potential risks by examining human behaviour, even the ways people walk, and their responses to questions.
Such uses of AI, with appropriate precautions, are beneficial, improving our well-being while reducing costs.
However, in spite of this progress, most AI is still very bad at understanding the world and offering good solutions. We all remember Microsoft’s paper clip ‘assistant’, which was often far worse than useless. The term ‘artificial stupidity’ arose to describe such low quality AI ‘solutions’.
Now a new, worrying trend is emerging, where people are forgetting to use their own judgement along with the AI, just accepting the output without question and using it as an excuse when it all goes wrong.
This trend may in fact be due to our litigation culture: We are all becoming familiar with stories of trucks getting jammed in corners on country lanes because drivers blindly followed their Satellite Navigation (Satnavs) Systems. More serious is the trend of doctors beginning to rely too much on AI instead of their own judgement.
A friend’s personal experience a year ago highlights the pitfalls of becoming totally reliant on machine intelligence: Here is his story: “A doctor told me that the deep vein thrombosis I am certain I had (and I am no hypochondriac) couldn’t be one because there was 93 per cent chance that I shouldn’t have one. A 93 per cent chance is just that, no more. It is not certainty. But because I only had a 7 per cent chance of having a DVT, I was sent home without any treatment, or any suggestion of a possible alternative cause for my symptoms. The effects afterwards correlated very highly with it having been a DVT, one of the 7 per cent, but if I had died from it the doctor would have been seen to have followed the book and I would be just another medical statistic. My medical records still state that it was not a DVT, but that is something that was never actually checked.”
Such abuses of computer systems are becoming much more common, and it is a dangerous trend. When ‘probably is’ becomes the same as ‘is’, we have a real problem.
Blind faith in computers appears to be increasing, with growing state use of computers and AI systems, and too few checks made on accuracy. When AI makes deductions from false data, then the results will be wrong, yet the consequences are just as real, leading to anything from a poor credit rating to death. Litigation threats increase the desire to have someone or something else to blame, the computer being an ideal candidate that we all love to hate. But unless we nip this trend towards blind technological reliance in the bud now, it will become a real threat, both to the benefits that can be harnessed from properly-implemented AI and to our lifestyles.
Fast Future is a research and consulting firm which focuses on helping clients anticipate and develop innovative responses to the forces, patterns of change and ideas shaping the future.
