In a follow-up episode of Bloomberg’s “Odd Lots,” Iain Dunning, head of AI at market maker Hudson River Trading (HRT), discusses how the firm is deploying artificial intelligence and the practical constraints involved. The episode is recorded at Bloomberg’s recent live show at New York’s City Winery and comes roughly seven months after a prior conversation with Dunning about HRT’s use of AI.
The discussion covers topics including the cost and availability of computing components, with references to the “price of memory” and bottlenecks in compute capacity. It also addresses the firm’s internal spending related to tokens, describing how employees are using token-based resources in the context of AI work.
Dunning further explains considerations that could lead HRT to develop its own chips, suggesting a strategy to manage performance or cost trade-offs as AI requirements grow. The episode also touches on broader effects that can accompany large-scale AI usage, including “AI-induced delirium.” The podcast format provides a high-level view of HRT’s approach rather than new financial disclosures.
