Prediction requires data – at least, it does if you want to have a reasonable chance of being right. Even with data it is possible to be terribly wrong. The sub-postmasters of Britain can vouch for that. The more of them that were convicted of stealing, the greater the belief by the institution’s managers that the Post Office was being robbed. The common sense absurdity of a cabal of conspiratorial sub-postmasters passed them by as they were driven to – well, we don’t actually know what, and as the police are trying to find out, better not to prejudge.
It is said that Benjamin Disraeli coined the phrase “lies, damned lies, and statistics” to highlight early examples of fraudulent persuasion. Whether so or not, it still has a ring of truth about it. The issue is brought to mind by AI. Not only because of the clever combination of data to create a model of the future but because of the capacity to absorb, analyse, digest and regurgitate almost incomprehensible quantities of data at a speed hitherto beyond the imagination of humans. We are suckers for ‘big’ and victims of ‘more’. Both are impressive; neither of itself proves anything.
Democracy and freedom depend on the quantity of people subscribing to them to survive. Both rely on majorities voting for them regardless of the thinking quality of the voters. It worked marginally well as long as voters could make up their minds by personal judgement based on seeing and hearing. Then along came social media followed by self-generative AI. Judgement, already suspended as far as products are concerned, jumped out of the window and hit the concrete below. Not only was extremism encouraged. A sort of John Wayne, Wild West bravado was presented as genius. The era of Trump had begun.
Herman Kahn taught me forecasting. His mantra was ‘forecast well ahead and change your forecast frequently’. When I asked him to define the two parameters he replied ‘about two thousand years and twice a week’! In his more serious moments Herman encouraged the belief that all forecasting was a challenge to beat the predictable. I have a lot of sympathy for this approach. Clients who come to us with problems that suggest sure-fire, hell-like consequences don’t only need to know what they are doing to themselves but, rather more importantly, what to do differently to modify it. Of course, the forecast is an essential ingredient in this process but too often it is taken as a gospel prediction instead of a cautionary warning.
AI is, to me, a challenge to become more creative and not an excuse for lazier thinking. Predictive forecasts are Roman roads of direct, not always very thoughtful, consequences. But life is a meandering river which we not only navigate as we choose but which we also design to be unique and valuable. To the extent that forecasting is a warning sign of the consequences of a pattern of behaviour if not changed it is useful. To the extent that it dictates what ‘should happen’ it is a dangerous move towards auto-pilot roboticism and the annihilation of humanity as we know it.
GIGO* is alive and well and living in, perhaps, the British Post Office?
*GIGO: Garbage In, Garbage Out
As always, your views are more than welcome at [email protected].
11 January 2024