Posted By 
By Chris Nicola
(Sponsor Content)
Wouldn’t it be nice if we could just ask a computer, “what should I invest in?” In theory, sure. But given the current state of artificial intelligence, I wouldn’t bet your money on it just yet. Simpler, tried-and-true analysis tools help us to do this job without putting our client’s money on the line.
Now, you might think AI for investing is inevitable. After all, intelligent people are coming up with new uses for AI all the time. Diagnosing eye problems. Helping corporate boards hire with diversity in mind. We can make art with it, or use it to spot fake art. With self-driving cars, we’re literally putting AI in the driver’s seat. So, why not AI for investing? In fact, it’s already happening — through the “AI Powered Equity ETF (AEIQ), which invests in a variety of U.S.-based companies and seeks to beat the returns of the S&P 500.” The numbers in this report from July show the ETF was up 8 per cent this year, while the S&P 500 has gained about 1.5 per cent YTD.
Who knows what the future could hold for AI’s capabilities? Let’s take a deeper look.
Why I’ve got AI for investing on the brain right now
I recently spoke about AI at The Summit for Asset Management and I realized that the subject of investing using AI would probably also interest many of WealthBar’s own clients. So here’s what I spoke about and some of my thoughts on the role of AI in the future of automated investing.
Who am I? I’m WealthBar’s Co-Founder and Chief Technology Officer. At WealthBar, my team and I design and build our mobile and online experiences for clients. We also decide what technology solutions to use and how to use them.
AI is really cool right now. It gets a lot of attention and I’d be lying if I said I wasn’t excited by the potential. However, I also need to ensure we’re using the right tool for each job. So far, we have not been convinced that AI is the best solution for the kinds of the problems we currently solve for our clients. More importantly, the current understanding of AI today leads to far more questions and concerns than answers. The biggest of these questions is our inability to explain exactly how an AI algorithm learns and why it makes the decisions it does.
AI-powered? Don’t believe the hype
What is AI?
Artificial intelligence can learn without being explicitly programmed.
That seems like a good definition that I would use, though what we think of when we say AI and what it really means can be two different things.
Some people will think of AI as a general intelligence: basically, a piece of software, machine or robot that thinks like a human, like in so many science fiction movies that have come out in the past few years (I, Robot, Ex Machina, Her, etc). We’re pretty far from that in real life. But for now, we’re using a definition of AI that’s more limited, precise utility — and one that could have some real applications in the investing world.
AI for investing. How does it work? And how could it work?
Right now, we use it to take on specific tasks. In theory, we could ask an AI “what should I invest in?”
Some financial firms are experimenting with this, or they say they are. But scratch the surface. You find that their process is just based around linear regression. They do some math, which has been around since long before computers. Linear regression is a quick and practical tool that the financial industry has been using for as long as there has been a financial industry.
You can use it to try to create a model, so the results fit data you’ve seen in the real world. Then you can use it to try to predict what’s going to happen next. For an investor, the power to predict with perfect accuracy would be a license to print money.
Of course, in the real world, we don’t have perfect accuracy for prediction, no matter how sophisticated the technology. A lot of that comes down to the inaccuracy of the data we feed into a computer. As the old saying goes, garbage in, garbage out.
Early experiments with AI for investing do show promise … so far
Let’s take that question I asked at the top, “what should I invest in?” How could an AI answer that?
If you just throw all the data of the stock market from the last 50 years at a so-called AI, that’s not going to work.
First, humans still need to organize the data. We need to set up a framework of how to solve those problems in the first place. Then you set up the AI to do a statistical analysis. That’s what’s happening with the few AI investing pioneers we’ve been watching.
It’s early days in terms of measuring performance, or even quite understanding how they work.
For instance, there is the Canadian Horizons Active A.I. Global Equity ETF (MIND). Horizons beckons investors to gain the opportunity to pursue “first-of-its-kind” investing where security selection is 100% chosen by an A.I. system to create a global basket of ETFs.
When it launched, its own manager said the following: “We don’t know what the computer will do.” (Read that again. That’s a big reason why we don’t do AI investing).
Maybe the manager was trying to manage expectations? If so, it was the smart move. By May, its 1.8% return was underperforming slightly against the MSCI World Index. MSCI is a broad global equities index that picks stocks from 23 different countries.
Then there is the AIEQ ETF in the USA. It invests in U.S. listed stocks and real estate investment trusts. It launched just a few weeks before the Horizons fund. This one is already showing more promise.
I was curious about AIEQ when it started out. I’ve compared the AIEQ with the index it tracks, the S&P 500. The results, at least for a few months back: AIEQ was continuing to beat the S&P 500, by a widening margin of over 6 percent.
That’s pretty encouraging, isn’t it? But here’s the big question: is it luck? Or is the AI just getting smarter? Are little differences in returns becoming big advantages for AIEQ over the long term? How is it doing this? Is it repeatable? Is it predictable?
That’s a lot of big questions right there. The biggest problem with the most advanced AI systems that I see is that when it works, we can’t actually understand why it works.
This makes it harder to trust AI. We can’t reason about its decision making ability. We simply have to accept it based on the results … and that’s not a small problem.
Some final thoughts about the future of AI portfolios
As things stand right now, AI portfolios are not on our radar to offer to clients for 2018. We will continue to follow these trends because we want to see how how they will drive the ETF market. For now, we’re taking a wait-and-see approach. Since we manage people’s money, we stay on the better side of caution. In the end, will it result in smarter investing? That’s the million-dollar question.
Chris Nicola is WealthBar’s COO and Co-Founder. He has an M.Sc in Applied Math and was managing technology for a financial firm while still a teenager. He gained experience as a programmer, web developer and CTO for a range of industries before co-founding WealthBar.