More than just common sense in having man (or woman) oversee the machine as an insurance policy, investors may find more than just the need to adjust models but a broader market regime shift. And it is during these periods Cao believes, citing Oaktree Capital co-founder Howard Marks, investors will make the most of their returns.
«This seems to apply to machine learning models equally,» Cao added.
Okay for Now
For now, Cao observes that AI programs have remained resilient, citing senior industry figures on the matter. He tested, for example, trading models due to their short term nature and reliance on recent data as well as quick adjustments.
«The low latency process we favor have performed particularly well,» said Dajun Wang, managing director from State Street Bank in Boston.
«Although recent market behavior has been volatile, the features exploited by our machines learning models were not at unprecedented levels,» added Anthony Ledford, chief scientist at Man AHL in London. «In other words, our ML models did not find themselves ‘beyond the data’ they were trained on.»
«Hype to Action»
«When AlphaGo beat the best human Go player a few years ago, […] all seemed within reach,» Cao said, highlighting headline ambitions such as driverless cars, AI doctors, and robo-advisors.
