Yesterday a good friend of mine shared Bloomberg’s piece on the 10$ Hedge-Fund super computer and asked me “Will there even be anaylists in the future or just adminstrators who operate the software?”. Here is the answer I gave him:

The answer is layered. Technology doesn’t remove jobs, it just changes them. Excel for example didn’t put accountants out of business, but rather it improved their performance. Casebriefs doesn’t put lawyers out of business, it just helps associates be more productive in their jurisprudential research. Yes, machines, automation and AI can indeed replace human effort, much like a computer can consistently outperform top chess players. However, the best outcome is always achieved by integrating the machine with the human. Specifically, in the case of chess for example, a chess players and a computer will consistently beat just a computer. They don’t even have to be the top chess players and top computer vs a mediocre computer, because it’s about the synergy between the player and his or her machine.

In the case of financial analytical technology, even the most sophisticated algos are pretty dumb sometimes. Beyond the famous example of the Flash Crash, some algos started trading huge chunks of Berkshire Hathaway holdings after Oscan night a couple of years ago for no reason other than the fact that Anne Hathaway was there and articles about her came out….yea, pretty dumb. That’s why the best algos just make suggestions to traders, because the one thing they are much better at is parsing through huge and complex pricing data to discern patterns. But true success can only be achieved if there is a human on the other side that confirms and locks in the position.

AI is still waaaay behind in terms of having a personal assistant like Tony Stark or something like Skynet (although we are closer to Skynet in some ways). Soooo much code has to go into doing something as simple as comparing 2 images of cats, when a 3 year old can tell you almost immediately that those are 2 images of cats. The advantage the financial industry has however is that the data to be analyzed by the AI is actually pretty simple (for a computer that is) because it’s basically just numbers. AI that processes numbers, and engages in supervised or unsupervised machine learning, are still hard to make but much easier than say the AI Netflix uses to figure which original shows to produce based on the movies and tv shows their viewers watch (yep, that’s how House of Cards, Marco Polo and Orange is the New Black got made).

So in terms of quantitative trading, yea we will need progressively less humans to do stuff and the traders left will be very smart people that understand higher level math and some CS so that they can confirm trades suggested by algos. The era of Jordan Belfort and Bud Fox is long gone, because now Wall Street is a place for academically and technologically grounded professionals. In the case of value investing (whether it’s a real thing or not) and shareholder activism, the best AI can do right now is find discrepancies between accounting reports. No AI so far has been able to discover that Herbalife was a pyramid scheme (but Bill Ackman did), that Apple is worth a trillion dollars (but Ichan could), or that Lehman was going to collapse because of toxic CDOs and MBSs (but David Einhorn could) so humans will still have an important role to play in the stock market, just not as much in quantitative trading, but even there AI won’t completely replace them.