“Eliminating Work & Applying Technology”

That was the way that CEO Brian Moynihan explained what Bank of American was up to after posting $8.6bn in quarterly profit cut 1,000 jobs. Together, the six biggest US banks made a combined $47bn last quarter – up 18% year on year – while slashing some 15,000 jobs. Wall Street, it appears, has pivoted from its line that AI will do all the boring work and leave the high-value work to high-value cats in their deal slides. In a particularly ironic cut, Citi is slashing jobs in its “AI Champions and Accelerators” program. The efficiencies are showing up in the quarterly numbers, investors are pleased… so what could possibly go wrong?

Superficially, replacing the quants – the number crunchers – makes sense because that is the sort of thing that AI does really well without people like me adding a lot of caveats. Since the early 1980s, when Jim Simons pioneered quantitative trading, much of whats done these days is already fully automated anyway. AI will produce the same output, and be vulnerable to the same blind spots as before. It will just achieve this faster and cheaper.

The forward looking grenade will show up in how those new black boxes will affect the entire system. As I’ve written before greater efficiencies very often lead to instabilities within a complex system. Predictive AI and quantitative trading are essentially the same thing and are very capable provided nothing changes. All the trading data is backward looking. It was Jim Simons’s good luck to pioneer quantitative trading in the dying gasps of the Cold War where the Soviet bloc and China began making moves to open up their economies to a stable global market. Since the 1980s the only change to the system has been, more or less, gentle but constant growth. I won’t describe the 2008 Financial Crisis as a hiccup – it was more full-body dry heave – but slowed growth over a decade, it didn’t reverse it.

In finance, timing is everything. It is the bad luck of AI to find its legs on Wall Street at precisely the moment that all that backward looking market data the quants trained their models on is changing. The universe of gently expanding growth that has prevailed for the last 30 or so years is collapsing, and the patterns it provided aren’t going to be of much use going forward.

And even that isn’t really the problem.

Second Order Effects

That the AI investing boom has distorted markets and we are teeing ourselves up for another financial correction is tiresome, but these things happen. The snake in the grass is that Wall Street, and the system that it underpins, is erasing its own talent pipeline in the name “eliminating work & applying technology.” A vaguely sentient human can see that the global economies are being rewired as alliances shift and commerce is weaponized. The old trading patterns won’t map as we increasingly rely on technologies that rely – entirely – on prediction rooted in old patterns. Telescope the current trend a few years and the industry is virtually assured not to have enough human managers to absorb the “noise” and chaos generated by all the AI black boxes. Money is a very, very visceral issue for humans. The knee-jerk market reaction will not fit into any predicable pattern which will amplify the amount of noise coming out of the models.

And when that happens the system will collapse.

The Field Guide:

Honestly, just stop and use your imagination. You want to use the power of AI to improve your position here, try prompting a report on whatever it is your chewing on with a conclusion and recommendations. Then generate a second report arguing the opposite conclusion. Read both with a pen in hand. Then put the pen down and – this is the hard part – sit for ten or fifteen minutes thinking about the matter with no interruption, no input. Just stay focus for ten minutes. You can’t out think AI, but that’s how you outwit it.

---

Preparing for a market expansion or product launch and want to test second-order risks?

Let's Talk