Saturday, September 3, 2011

Deeper Into the Black Box?

Stanford University is offering their class Computer Science 221: An Introduction to Artificial Intelligence online this fall. They are also making the class open to anyone and the best part is: it is all totally free. (Although there is an optional text book to buy.) The class will be team taught by Stanford professor Sebastian Thrun and Google's Peter Norvig, two leaders in the field. I am excited for this class both for its content and also for what this kind of format portends for the future of global education.


A good friend and workout buddy of mine, Olin, has in the past year become a huge evangelist for A.I. He has an excellent post on his corporate blog about why everyone needs to take this class. And I agree.  It used to be that only a small group used computers to manage funds, now nearly every successful investor incorporates some level of quant screening into either the alpha or risk process. As those screen become more sophisticated, no doubt some level of A.I. will sneak in.  Olin also makes an interesting observation about re-educating programmers to teach the computer. It reminds of my early days in the industry as a consultant working objects before they were ubiquitous. (I would argue that that they are still rarely used to their full potential.) I would spend a lot of time teaching quants to think less in a process-driven way and to program more naturally. Trust the computer handle lower-level tasks like iteration. Focus instead on tying solid economic theories to investment outcomes.

I have often said about quant investing that one does not do what the computer says, in actuality, one tells the computer what to do. Properly coded investment algorithms need not be black boxes. They can provide full transparency to the investor. At first blush, A.I. systems may apear to be a darker black box as the programmer cedes some control over how exactly each step of the program will progress. But upon further reflection I am now convinced that a properly set up A.I. system, like any good traditional quant algorithm, can shine a light on data. It should be possible to use A.I. to add a deeper understanding of data and patterns. And thus it should ultimately prove to be a better tool for transforming information into both wisdom and effective action.

Do It

This will be the next wave of data analysis. I have already enrolled. I hope you will too. The class starts in October, but you should enroll now. Let me know if you sign up and I will see you in class.

Bonus: Stanford is also offering CS229: Machine Learning and CS145: Introduction to Databases this fall. Both are also free and online.

No comments:

Post a Comment