SEC Chair Gary Gensler, a significant figure in U.S. regulatory affairs, asserts that AI might ignite future financial crises. Moreover, he believes regulators could find it challenging to preempt such threats.
AI’s Role in Potential Financial Downturns
The Risk of AI-Powered Trading Algorithms
A leading concern is the unpredictable nature of AI-driven trading algorithms. There’s the potential for these algorithms to simultaneously sell the same assets, precipitating a market downturn.
Gensler points out a worrying trend: a limited pool of professionals trained to develop these models. Many come from similar educational backgrounds, leading to what he terms the “apprentice effect.” This commonality might increase the risk of “model homogeneity.”
Moreover, if regulators dictate AI operational parameters, it might inadvertently steer everyone towards similar strategies, amplifying the risk. There’s a fear that companies might then gravitate towards AI services provided by a select few industry giants.
For regulators, the opaque workings of these algorithms make it almost impossible to anticipate or prevent a potential market crash. As Gensler emphasizes, the unpredictability that makes deep learning powerful is also what makes it so unmanageable.
Beyond Trading: The Broader AI Risks
The influence of AI is not restricted to trading alone. Several AI tools assess creditworthiness. Given their complex and ever-evolving nature, these tools might sometimes inadvertently discriminate against certain individuals. The ongoing evolution of these systems means that an AI tool’s behavior today might differ drastically tomorrow, creating regulatory nightmares.
According to Gensler, as the adoption of deep learning in finance grows, so will regulatory gaps and systemic risks. His solution? Financial institutions should be mandated to hold more capital if they rely heavily on AI. Additionally, results generated by AI tools should be validated against more transparent, conventional models.
However, even these measures, Gensler admits, might not be enough to tackle the impending challenges.
The Data Challenge
Data is the lifeblood of AI. But Gensler warns of the perils of all AI models relying on the same vast data sources, such as Common Crawl. This could lead to all models inheriting the same vulnerabilities, resulting in correlated predictions that might exacerbate market volatility.
This hunger for vast data has also seen certain companies gain monopolistic control. Gensler cites the Intercontinental Exchange’s dominance in the mortgage-data industry, following its acquisitions of MERS, Ellie Mae, and Simplifile, as an example.
These monopolies present “single points of failure,” reminiscent of the repercussions following the Lehman Brothers debacle. Furthermore, even the most expansive datasets are incomplete. Many don’t cover an entire financial cycle, a limitation that played a role in the last financial crisis.
Herding, Crowding, and the Global Impact
Algorithmic trading, already prone to “herding and crowding,” has been linked to flash crashes. This risk might be accentuated as traders increasingly rely on AI. Companies in emerging markets might use AI tools that aren’t calibrated with local data, raising the stakes even higher.
The rise of AI presents unprecedented challenges in the financial world. As these systems become integral to trading and other financial operations, the risks they pose grow in tandem. As Gensler succinctly puts it, the greatest danger might lie in what AI systems are unaware of.
Related Articles
