Thoughts from "Simple Economics of AI - Prediction & Judgement" Presented by Prof. Ajay Agrawal
Timothy J. W.
Director of Technology Internal Controls, Regulatory Management, and Third-Party Technology Resiliency at Scotiabank
Earlier this month, I was fortunate to attend Prof. Ajay Agrawal’s talk on “The Simple Economics of AI” at the Scotiabank iLead Digital Learning Series. The topic was very thought provoking as Prof Agrawal encouraged us to consider AI as synonymous to prediction. In order to be competitive in our marketplace, we should all be looking for opportunities where we can offload the work of prediction to machines (e.g., credit scoring, defaults, IT outages, come to my mind).
He shared with us the five different classifications of Prediction (read AI) as postulated by Graham Taylor: Classification, Regression, Clustering, Dimensionality Reduction, and Generation. Imagine using Regression to help us predict network outages – reliability index. Or using Generation to determine best next course of action for restoring service.
Prof Agrawal postulated that AI will take over the prediction function while human intervention is still required for judgement. An example would be for medical diagnosis where the test results would be interpreted by AI (which would be much more adept in parsing through the symptoms) and propose a course of treatment, while the medical doctor would review the prediction and using intuition to provide judgement.
The most shocking moment of the talk was when Prof Agrawal cited an example from MIT Technology Review (a very fine publication, BTW) that at Goldman Sachs, they replaced 600 equity traders with two traders using AI. One third of GS staff are computer engineers!
(Inserted UBS Stamford trade floor for shock and awe – perhaps it was the ’08-09 crash, perhaps it was IA, or both?!)
Photo credit: https://thisweekonwallstreet.com/the-ubs-trading-floor-before-and-after/
AI has been in existence for a long while but it is at a watershed moment because the cost of compute and the cost of data storage are low. The ones who can capitalize on it from a usage and talent perspective will win in this new machine age.