- Bath Artificial Intelligence


Artificial Cognition through Interaction

8 Febuary 5:15pm, 1W 3.6

In 1996, I employed some fairly simple AI to develop one of the first ever adaptive autonomous trading agents suitable for real-time operation in the online versions of the global financial markets, an algorithm known as Zero-Intelligence Plus, or ZIP. In 2001, a team of researchers at IBM showed that ZIP could consistently out-perform human traders under experimental conditions. In the decade since then, I've worked with, and in, major financial institutions applying AI to the global markets. In the past five years, there has been an explosive growth in so-called High Frequency Trading (HFT), where automated adaptive systems trade at super-human speeds. On May 6th 2010, in a period lasting roughly 30 minutes, an unprecedented sequence of chaotic events took place in the New York markets: an event now widely referred to as "The Flash Crash". The role of HFT in the Flash Crash has been the topic of some debate. In this BAI Seminar, I will summarise these events, and place them in the wider context of major socio-technical systems failures in the deployment of risky technology, and will argue for the use of agent-based modelling and computational intelligence techniques in helping to identify the likelihood of similar events in future. >