AI In Hedge Funds
A friend of mine passed along a really interesting article today about how Man Group is using AI as a core part of its investment strategy. Obviously this is something of extreme interest to me, considering we’re trying to do something very similar at Apteo.
The article is a mix of prose and a surface glance at some of the tools that Man Group is using. What’s interesting is that they mention two of the same technologies that we’re currently working on: deep learning (DL) and reinforcement learning (RL). It doesn’t surprise me that these guys are starting to look at some of the more sophisticated techniques out there today.
In fact, RL seems to have found a nice home in the world of AI-managed investments, as I’ve seen a bunch several other fintech companies using it to gain an advantage.
I think it’s only natural for AI and ML to move into the world of investments. A lot of the algorithms and hardware that’s out there now is orders of magnitude better at handling data analysis and teasing out latent features and understanding patterns that humans are. In fact, I’m surprised it’s taken as long as it has for firms to start looking at these tools.
I think that a lot of the problem has been lack of talent and hardware, combined with a fear of being unable to explain why a black box is making the decisions that it is.
The issues of talent and hardware are being (or have been) addressed. And I think that as clients start to see more stable, hopefully much more positive, returns, they won’t feel as queasy with having a lack of explanation about why decisions are being made. To be clear, this is still an issue today.
In our own research, we’re finding that clients hate black boxes and want an underlying explanation of what’s going on.
Right now, this makes sense. But as time goes on, not only will this be harder and harder to do, the very request will become ludicrous.
After all, as machines get “smarter”, and hopefully, approach the general reasoning ability of the human brain (without its biases), they’ll excel at finding patterns and making decisions that humans can’t make, and they’ll do these things better than any of us can. At that point, who are we to ask an explanation of them?
In any case, it’s exciting to see that fintech is breaking into the realm of optimizing investments. To date, we’ve had linear, high-frequency trading dominating the world of millisecond-level trades, technical analysis, stat-arb, and other statistical algorithms handling more medium duration strategies (and even in the article I linked to, their holding period was from minutes to weeks), but we haven’t had too many longer-term automated investment platforms.
That’s a problem that we’re working on at Apteo, and our initial results indicate we’ve found an approach that works. I do believe at some point, a large part of decision making and analysis in the investment world will one day be done by algorithms, but I think that will also create a lot of demand for engineers, data scientists, analysts, and data cleaners in the future.