The Event Horizon Problem of Future-casting

At what point is the future completely unpredictable?

Ahh the future. But did the artist of this image really get it so wrong?

Today's AI models are probability machines. They predict what word will appear next, based on a trillion data-points. It's a form of future-casting in miniature.

The results are impressive. We find ourselves reading articles that have been fully composed by a machine that is 100% non-sentient, and yet which can accurately mimic, and often outperform, a smart, sentient human. In fact a recent ethically highly dubious study, using unaware participants on Reddit, showed that AI can be significantly more persuasive than many humans. (This was backed up by a far less ethically dubious study, this time on fully aware participants, using the academic research platform Prolific, with similar results.)

So a future-casting machine gets it 'kind-of' right. It semi-accurately predicts the possible-future of a text-based document, by guessing what word, or even letter, might come next. So what?

AI future-casting creep

Let's be aware that AI will increasingly do this across more and more areas of human life, from traffic analysis, to storm prediction weather reporting, to analysing shadows on radiography x-rays, to predicting the outcome of sporting events, to insurance risk, to playing the stockmarket. Indeed, it will even attempt to predict your own death.

And as it gets better at doing it, so the near future as a whole will come into greater degrees of clarity, for better, or worse.

But chaos and complexity theories suggests there will be hard limits to this capacity for prediction.

Like a fractal that is highly-predictable at first, but then eventually becomes impossibly variable, so the probabilities of future events become ever more unpredictable the further away they get from any single point in time.

Only in a 4D-cubed universe, where we are all essentially holograms, is it possible to predict with 100% accuracy what will happen in the future. And of course, we could be just one of an infinite number of 4D cubes, with fractal-like tendencies to split off from one another (like the multiverse itself).

The question is, how far out is the 'event horizon' of our prediction abilities? When do we veer away from the highly possible, to the impossible to know?

The future-casting event horizon

Weather prediction has this exact same issue, with 10 days being seen as the absolute limit of accurancy by most meteoroligists (when Accuweather tried for 90 day forecasts it seriously annoyed the entire profession).

As Edward Lorenz said, 'One flap of an seagull's wings could change the course of weather forever.' Let that sink in.

This famous idea is now known as 'the butterfly effect', and seems to be based on the late Ray Bradbury's excellent science fiction story 'A Sound of Thunder'. In this prescient story a single golden butterfly trampled by a time traveller in the Late Cretaceous period leads to a switch in future timelines. This 'domino-effects' his present, transforming it so that now everything is spelled phonetically, there's a fascist govenment, and he gets shot stone-cold dead.

So much for accuracy. But of course it gets worse, because 90 days is a far shorter timeline than most future-casting aims to hope for. We want to predict ten years out. Sometimes more, because that's the horizon that feels important to people living within the confines of our limited human lifespans.

Future-casting makes a come-back

But on the flipside, human beings and the world change far more slowly than water molecules, so at least we stand a chance. Right?

Consider this little gem. From 1899 - 1910, Jean-Marc Côté and some of his artist buddies had a crack at creating images that predict the future:

The results are hypnotic (click the link above). And almost ludicrous (my fave is the blue-whale powered sub). But are they truly wrong?

If you take your literal goggles off, and instead put your brain into thesaurus mode you begin to see that the predictions are actually pretty accurate. Brain machines? Ask Elon. Flying cars? Check. Robot barbers? Coming soon apparently. Airships? Defo on the come back.

So perhaps we are better at predicting the future than we thought. And perhaps as technological change increases in speed, so our past predictions become strangely more accurate, as they get swept up by the great tide of ingenuity that new technologies combining in new ways often produces.

From a Now Now Labs perspective the key learning here is that almost every prediction of future technology will eventually come true. It's just a matter of timing and, importantly, choice.

A values-led approach

Which brings us full-circle. If all possibilities of technology will someday be made (even if not quite in the form we envisaged) then the event horizon is perhaps not the real problem.

The real problem is which of these powerful technological spells will we decide as a society to invoke now? Future-casting, in this light, becomes a political, social and values based game.

By turning our eye towards the values of the society we currently live in, we can begin to more accurately predict the technological leanings that we will prefer tomorrow.

At Now Now Labs we help bring this unique combination of present values and potential future states into sharp-relief.

Clarifying the values of today, helps us understand the collective visions of tomorrow that we want to bring about. So we can go ahead, and actively choose them now, as willing and enabled participants in the creation of our own futures.