Why we need Tony Stark’s Suit / Coming of Assisted Machine Learning in Asset Management/ Hedge Fund Industry
Since I first the heard about the word Artificial Intelligence at the age of 15, I have been its biggest fan. It seemed obvious to me that over time, the only way forward will be teaching a machine to learn, to do things we do so badly and at the end really complement us. I am in the camp which believes yes, AI will replace a lot of what we do. But in the end, I see it complementing us, making us better — aka Jarvis of Tony Stark, and maybe some day a well intentioned Ultron (Age of Ultron anyone?).
I first got interested in the markets during the crisis of 2008, largely thanks to my brother who worked at Lehman at the time. It seemed weird to me why a machine could not take all these variables of data, crunch and tell us the signs. Years later I did learn that the Bridgewaters of the world, and anyone who has a data driven approach, did foresee signs. I have huge respect for Ray Dalio, who is founder of Bridgewater, arguably the most successful Hedge Fund Manager, and a man of principles.
Coming to that point, I joined PIMCO on their global portfolio desk dealing with Rates, FX, Swaps, Mortgages, Credit day in and day out. Again extreme respect for all the people there, and given PIMCO is one of the more data driven shops, I was surprised how still tons of money is managed based on simple Excel and mental models when the existing analytics / machine learning (ML) out there could do so much (case in point, ML can allow parsing through all possible news sources, even in local languages, to detect a current event and chart what scenarios have been triggered by similar events in the past). Thanks to the amazing computing power, easy APIs/ libraries for ML and associated systems, it is not difficult. (Thanks to Geoff Hinton, Yoshua and Andrew for their work on Deep Learning).
With this sense, I went back to the drawing board and started a company to map every news item to every event and how it would affect the markets. After two years of hard work, multiple NLP models and market musings, I realized that it was hard work. But the events in last four years have only strengthened my belief. Meeting 50+ asset managers/ hedge funds — and working again in the fund industry for my summer during Wharton, looking at Chinese markets, I found how accessing data, putting that data into models and really complementing the fundamentals with quantitative analytics/ machine learning / new data could bring a huge improvement both in the investing process and returns over time.
What I find lacking is the acceptance/ effort from so many money managers (collectively managing trillions of dollars). I wonder why they are not thinking this way. Yes the upfront cost is massive, stretching into millions, but that is the only way forward. Obviously some hedge funds/ asset managers have been doing this forever or are investing heavily into it — but its less than 2–3% of the industry. And the ones who are currenly / have previously done so are shining examples of returns — Bridgewater, Winton, RenTech, etc.
Here is my thesis on the coming quantification of the asset management industry. Coming from a markets background as a trader, a data scientist, a machine learning coder, and builder, I wish more people on both sides had the balanced view that fundamental plus machine can merge together to really capture that alpha. (Often the best ideas are out there, from tons of research — all you need is to spot them).
I took certain pieces from the recently published AI guide in Markets by JPM. Happy to hear comments/ thoughts, but one thing is certain: machines are coming. Either you embrace them soon or you will be Amazonized.
Some articles that throw more light:
Artificial Intelligence and its fast adoption in asset management — News Snippets:
- http://www.bloomberg.com/news/articles/2015-05-20/the-10-hedge-fund-supercomputer-that-s- sweeping-wall-street
PS: I love markets, macro, fundamental deep dive in companies and building machine learning data models for solving real life problems. I am a recent MBA graduate from Wharton School and an electrical engineering/ signals/ quant undergrad from IIT Delhi and have worked on Trading Desks, Space Technology and startup in ML + markets. I believe in the age of Superintelligence and Moonshots while being cognizant of Kelly Criterion where you have to live long enough to make your bets pay off in the casino of life/ markets.