Man v. Machine: The Series

I recently watched a special on TV where IBM research executives & algorithm scientists created a supercomputer 'Watson' to compete against champions on the show Jeopardy! The show was an interesting playing field to pit man against machine.  The primary challenge for the programmers was getting the machine to 'understand' the questions and context (i.e. to effectively judge the intent behind the questions and decipher things humans intuitively understand). It took some of the world’s most brilliant minds ages to program ‘Watson’ and load his four terabytes of disk storage with 200 million pages of structured and unstructured content. Although I certainly would not sign up to be a contestant, ‘Watson’ made some comical errors along the way.

I didn’t want this to be a self-promoting plug about the NYSE…but I can’t help it. The hallmark of our market model marries technology with human intelligence. We maintain the only market that leverages human capital to compliment electronic trading. Why? Because nobody has figured out how to effectively program human judgment and accountability. The May 6th “Flash Crash” when the Dow tumbled 600 points in 5 minutes is just one itsy bitsy example of why these characteristics (judgment & accountability) are important in the capital markets.

Yes, computers are programmed to be fast and accurate, but no program/algorithm can be coded to consider every possible extraneous variable or assumption. Algorithmic and “high frequency” trading is certainly important in today’s environment – it contributes to, and improves, market liquidity…and simply cannot be ignored because of demand for lightening fast trade execution.  However, there are also shortcomings with “artificial” intelligence that induce volatility, reduce liquidity in times of stress, and make irrational assumptions. Hence, the need to include human oversight in the equation.