If everything is an AI crisis, then nothing is

LAST week, investors were given a 7,000-word jump scare by Citrini Research in the form of a dystopian memo that vividly described how agentic artificial intelligence (AI) would send the economy into a deflationary tailspin by 2028.

In this scenario, the worst things imaginable happen – businesses are gutted, stocks plunge and white-collar workers are made crushingly obsolete. This “memo”, which critics have variously deemed “market-moving science fiction” or predicated on “extreme and improbable” conditions, unfolds like a zombie apocalypse movie.

There is a terrifying headline here, an eye-popping statistic there – except it’s the economy that is the zombie and no shambling monster wants to devour our brains because AI has co-opted all the high-value cognition.

This imagined crisis was so compelling, thanks in part to the post’s unnerving specificity and emotive language. In Citrini’s memo, unemployment did not merely surge, but hit 10.2 per cent by June 2028.

Business at ServiceNow did not just slow down; instead, the software firm was envisioned to have lost 15 per cent of its contracts, its share price plunging 18 per cent by October 2026. Amid an “economic pandemic”, the social fabric was “fraying”. In the 24 hours after Citrini’s post was released, markets sold off, reacting as if the future described had already happened. The S&P 500 index shed more than 1 per cent and IBM saw its worst share-price plunge in 25 years.

Even if Citrini’s economic collapse scenario proves accurate, it might not matter. According to another report called AI 2027 released by the non-profit research group AI Futures Project last year, mankind will be annihilated by superintelligent robots by 2030, anyway. Incidentally, the AI 2027 report makes far more outlandish claims, but is backed by greater conviction than Citrini’s recent post. What is irksome about Citrini’s viral memo is not that it is completely wrong – it may well be right in broad strokes, if not in its hyperspecific details.

The problem is that its writers have cannily labelled the post a “thought exercise” – “a scenario, not a prediction” – creating wiggle room to make improbable claims about the future without having to vigorously defend them. This approach might not completely reek of intellectual dishonesty, but it would make anyone’s nose wrinkle a little.

If this sort of alarmist projection proliferates, it will do us a great deal of disservice, for we are already drowning in wildly varying extrapolations about a technology we barely understand.

No one can agree on anything: whether AI will be a credible threat to most white-collar work, whether it will create more jobs than it destroys – and do it before civilisation implodes from mass unemployment – and whether there is enough political willpower to adequately redistribute wealth should push come to shove.

There are only so many conflicting apocalyptic scenarios the human mind can hold at any given time before it simply stops trying to sort signal from noise.

The danger of this intellectual fatigue is that when someone does actually sound the alarm over a legitimate threat to our well-being from AI, nobody will hear it since we might have collectively tuned out by then.

In any case, I wouldn’t bother prepping for a 2028 doomsday economic scenario, if I were you. It’s those killer robots coming for us in 2030 that we have to worry about.