AI is giving early 2008
Black computer screen with blurry figures and wiggly lines, suggestive of a financial performance chart Photo by energepic.com
I’ve been thinking about why I’m so cautious (maybe even a little trepidatious) about AI. It’s not very like me - I’m typically an early adopter and enthusiastic technologist.
And I’ve realised - it’s the global financial crisis vibes. Right now, AI as a sector is giving early 2008.
Let me paint you a picture:
Though everyone still associated these banks with traditional deposit-holding high street banks, they were actually international financial institutions buying and selling complex and often opaque products that no-one at the top understood (e.g. collateralised debt obligations, mortgage-backed securities). Nonetheless folks were quite complacent about not really knowing what was happening because of initially benign economic conditions and the success of the first couple of decades after financial market deregulation.
In that time there had been major industry consolidation - and the flurry of M&A activity meant that the biggest players ran fragmented legacy systems that lacked the interoperability needed to show the data relating to the assets they now held. They could not simply interrogate their data to see what was toxic and what was not.
The boards weren’t sufficiently experienced or knowledgeable about banking and banking products to ask the right questions, even if they’d been able to access the data. There was too much trust in the competence and capabilities of experts and senior management - and not enough testing and challenge. Executive leadership failed to instil a balanced risk/return culture and prioritised growth aspirations over consideration of risk.
And so the firms carried on, with utterly inadequate risk assessment and risk management. They had neither the data, expertise, desire to properly assess the risk in their own products, or the risks held by the banks they were lending to, or the risks of the businesses they were being. Never mind the major systematic risks to the global financial system.
Leaders set the tone. Risk management was seen as a constraint on the business, rather than integral to it. Risk appetite wasn’t clearly defined - and so absent clear boundaries became “growth at any cost”. No-one said no. Where risks were escalated, they didn’t lead to a change in strategy - herd mentality and optimism bias ruled. Everyone followed the pack, ignoring clear signs of danger.
And the regulators? Just like the executives, they couldn’t keep up with the complexity and pace of all that financial innovation anyway. Inadequate oversight was accepted, expected even. And regulator teams were poorly resourced, too reactive, not sufficiently focused on risk and didn’t have the teeth to do anything about what they’d have found anyway.
So:
High complexity, high opacity (black box) products that no-one really understands in enough detail to do a proper risk assessment
Complacency, optimism bias, herd mentality chasing growth at any cost. Everyone else is - so why shouldn’t we?
Boards ill-equipped to provide robust scrutiny and challenge
Executives without the data - or incentives - to properly assess and manage the risks
Systemic, societal risks largely unseen until it’s too late - and regulators simply unable to keep up with it all, and pretty much toothless anyway.
I didn’t argue for the end of the complex financial products back in 2009. And I’m not suggesting we ditch AI. We can’t put either genie back in their bottle.
But is it too much to ask that we learn from the mistakes of the past?