Can Artificial Intelligence Solve Everything?

Last night I went to FinTank‘s panel discussion on artificial intelligence at the CBOE.  It mostly related to trading and artificial intelligence.  It was an interesting discussion and I learned a few things.  AI is being used 24/7 in stock and commodity markets already.  It has been used for the past few years as trading operations try and find anomalies in the market.

I took a quick picture of the CBOE floor. I never traded on it. The floor at the CBOE was always smaller than the CBOT and CME. But, there were some tremendous traders there and some great companies that were started there like OptionsExpress and ThinkorSwim.  As we know, AI and technology made floors like this basically obsolete.

A post shared by Jeff Carter (@pointsnfigures) on Apr 19, 2018 at 4:51pm PDT

Does AI work better than humans?  The answer is “Yes it does in most cases.”  Can AI select superior stocks, commodities, ETFs, etc?

1. Yes.  Using AI to improve algos and learning which ones didn’t work and why.
2.  
3. Using AI to pick next trend and look at them in multiple ways, (minute, ten minute, days, months) to help pick strategies.  
4.  
5. Seeking alpha with AI

Where does it fail?

1.  When there is a lot of noise in the market.
2.  
3.  When there are two and three standard deviation events in the market.

The question becomes, is there more or less noise in the market today than there was when it was all humans and are there more two and three standard deviation events happening today than there were in the past?

I don’t think so anymore.  I do think that we saw the flash crash and we tend to anchor on that.  There are certainly things that happen every once in awhile that look pretty weird.  But, as risk management from institutions has gotten better, the frequency of outlandish events is less and less.

The amount of information that needs to be processed in order to really find alpha in a market is intimidating.  The volume of information available to us in the world doubles every five years.  Doubles.  Five years.  That is such a massive volume and short time horizon that only something like AI can process and help us make sense of it all.

Are there downsides to AI and marrying it to trading?  You bet.

First, regulators are behind the curve when it comes to understanding how to regulate it.  As their knowledge base grows they will get better, but it’s important to realize it will take time.  Private entities are more up to speed but the big institutional players that provide structure in the market are still behind.

AI can make markets a lot more efficient.  We shouldn’t turn away from AI just because regulators are behind.  The trade off between efficiency and the risk of markets going awry weighs heavily in favor of efficiency over the long term.  AI can be trained to root out deceptive trade practices like spoofing.  AI can detect patterns and find bad behavior before it ruins a market.

One of the fears out there today are trading algorithms that are self-learning.  What if they learn the wrong stuff and do something illegal in the market?

Ironically, these sorts of things will create job opportunities for humans.  While humans may do different things and perform different roles they won’t be absent from marketplaces.  They will need different skillsets.  The rags to riches “didn’t graduate from high school” to wealthy trader story is probably done.

Another risk is data.  AI runs on data.  A lot of the data is inside private funds and it’s proprietary.  Some of it might be sensitive since they are mining all kinds of data from everywhere.  If it gets hacked it could be damaging.  The way a private firm might recombine data that is disparate might be damaging if it got out into the public as well.

AI is being utilized more and more today for things like risk management.  That’s important given that today the Federal Reserve caved to the big banks and changed capital requirements.  Leverage is a double edged sword.

Where is there a gigantic opportunity when you marry AI and capital markets?

Pre-trade.  AI is being used to do a lot of post trade analysis.  It’s not being used for pre-trade risk analysis and other pre-trade operations.

Just having AI is not a panacea. If you don’t know what you are doing or if you input shitty data you will get shitty results.  There is a lot of debugging that goes on with AI.  You know where the bugs come from? Programming.