Can Artificial Intelligence be Used For Stock Trading?
If you’re an average retail investor who just saw a 60-minute special on artificial intelligence (AI), you probably thought to yourself “why can’t we use AI for stock trading“? Then maybe you plugged the phrase “artificial intelligence stock trading” into Google and this article popped up. We’re going to answer your question and hopefully we’ll all learn something along the way.
There are really three questions to ask here:
- Can other people use AI for stock trading?
- Can I use AI for stock trading?
- Can I invest in products that use AI to generate alpha?
Let’s answer the first question.
Can other people use AI for stock trading?
Yes, we’ve written before about at least three hedge funds using AI to generate superior returns for their investors. Take as an example the hedge fund Renaissance Technologies which is said to have “the best physics and mathematics department in the world“. The Medallion Fund at Renaissance, run mostly for employees of the company, has one of the best records in investing history having returned +35% annualized over 20 years. That means that if you put $10,000 in the fund back in 1997, you’d have $4.04 million in your account today. So yes, other people can use AI for stock trading and they’ve used it to realize one of the best returns in the history of investing. Is that what you were hoping to get a piece of?
For now you’ll need to try to get into one of the hedge funds we covered that are using AI effectively. Of course past performance is never indicative of future results. You also have to wonder, “just what’s keeping anyone from using AI for stock trading in much the same manner as these hedge funds do“? The answer is, nothing is stopping anyone from starting a company that uses AI for stock trading.
In August of last year we wrote about 6 companies that are using artificial intelligence for generating alpha through algorithmic stock trading. One of these companies, Sentient Technologies, can simulate 1,800 trading days in just a few minutes and is pitting trillions of virtual traders against each other in a giant game of evolution. It’s run by an Apple guy who worked on Siri and he’s not saying anything about the returns he’s generating. While Sentient is being secretive, firms like Numerai are paying data scientists in bitcoin for their contributions to an AI hedge fund which has now crowdsourced billions of equity price predictions. Their big news was when some anonymous genomics scientist going by the name of NCVSAI racked up $13,000 in rewards and then cashed out. Let’s hope that individual went and started their own hedge fund.
This brings us to our second question.
Can I use AI for stock trading?
Yes, you can use artificial intelligence (AI) for stock trading but first you’re going to need a better understanding of what this entails. “Cognitive computing” is defined as simulating human thought processes by using computer models. That’s pretty much what AI is all about, but we need to separate computing out between software models and hardware models. Google’s Tensorflow is a “cognitive computing framework” that is completely built out of software and could be run on any desktop computer. Google has also developed their own hardware chip that is optimized for Tensorflow and we identified a whole slew of startups that are looking to build hardware that is optimized for cognitive computing. So before you go using AI for stock trading, you need to answer three basic questions.
- What AI software framework are you using or are you building one yourself?
- Is there specific AI hardware you are using that is optimized for AI ?
- What data are you going to provide your AI solution so that it can generate alpha?
If you are building your own software and hardware like Google is, then you are really having to excel in software development and hardware engineering, both of which require a lot of capital. But if you simply throw Tensorflow on some NVDIA hardware in your garage and hire some cognitive computing developers, you can really start a company quite easily. Maybe this is why some estimates put the number of AI startups to be well over 1,500.
Don’t feel motivated to do that whole company building thing? Let’s answer the third question then.
Can I invest in products that use AI to generate alpha?
Retail investors need to be very careful here. People throw around the phrase “artificial intelligence” everywhere you look these days. How can you tell where the “real” artificial intelligence is? You follow the money of course. You think that the guys and gals over at Sequoia Capital are going to trust a company that says “we use artificial intelligence but we can’t tell you how it works because it’s a black box“? Absolutely not. They do their own due diligence, something that investors need to do before just buying into retail investing products that claim to “use artificial intelligence to generate alpha”.
We’ve talked before about how “smart beta” ETFs of the future will use AI, we’ve talked about how robo-advisors are nothing new, and we’ve also highlighted an ETF that uses artificial intelligence to selects stocks. In each of these cases, we see the financial community trying to use technology to produce superior returns. These are the same people who cannot beat the returns of a passive index 80% of the time. If they really had some magic formula, don’t you think they would invest their capital in it instead of yours?
Still, we’d much rather have an AI advisor instead of a human advisor. If AI is now beating humans at some of their oldest strategy games, it’s probably going to be capable of making prudent investment decisions. It’s like someone introducing you to your financial advisor and saying “this guy is a world champion Go player, a professional poker player, a champion chess player, and he was the champion in Jeopardy as well“. You’re going to be pretty certain that dude is going to be making better decisions than you would be able to, all things being equal.
While the asset allocation methodology for most robo advisors is very linear, we’re now seeing firms like Qplum use machine learning to create a “robo advisor which uses AI algorithms to invest”, just as we predicted. Expect to see more firms climbing on this bandwagon, firms like Repsonsive.ai which shows you some examples of how your asset allocation changes in response to extreme market events:
The above is not looking that intuitive but what do we know. We’re far from having the intellectual firepower of a world “Go” champion. The obvious first question here should be about what sort of fees we’re getting charged for this product. We said we’d revisit the whole “robo advisor” thing when AI came into the picture so stay tuned for a future article on precisely that.
Things in this space are moving faster than we can write about them, so if you’re developing retail investment products that utilize AI then drop us a comment below so we can share your compelling “AI alpha generation” investment vehicles with our readers.
Are you paying too much in transaction fees to your broker? Check out a brokerage firm called Zacks Trade that's offering $1 trades on U.S. stocks and options until 2020. After that, you'll pay just $3 a trade or a penny a share, whichever is greater. You can also trade on 91 stock exchanges in 19 foreign countries. Click here for $1 trades on U.S. stocks and options until 2020.