How Hudson River Trading Actually Uses AI – Odd Lots

How Hudson River Trading Actually Uses AI – Odd Lots — Transcript

Market views

Bullish

INFRA— Electricity infrastructure is becoming a critical bottleneck for AI data centers, with power grid capacity constraining GPU deployment more than any other factor
AITRADING— AI models in trading are profitable at scale despite only being marginally better than random, with accuracy around 50.1%
ALTDATA— Alternative data becomes valuable at longer trading timescales, particularly for days-long strategies using SEC filings, news feeds, balance sheets, and broker reports

Bearish

LATENCY— The latency arbitrage game in trading is over, with no remaining edge in shortening network connections as the market has fully arbitraged speed advantages
AILABS— Major AI labs are becoming increasingly secretive, with published papers being either stale or unimportant while cutting-edge work remains hidden
CEFI— Crypto exchange infrastructure quality is poor, with exchanges operating like amateur JavaScript websites that fail to confirm order receipt

Not investment advice.