6 Answers to 6 Questions on Artificial Intelligence Usefulness in Stock Investing
Updated: Nov 14, 2017
I had an opportunity recently to answer some questions from a US News and World Report reporter looking to do a story on Artificial Intelligence in stock investing. Good questions but my responses were probably unexpected. By the time you read this in a major publication, it will likely be too late. This is a story still waiting to be picked up.
1. How do we see AI reflected in today's investment world and how does that reflect a change from, say, five years ago?
Real AI is not reflected in the investment arena. What we have in the past five years is a move to using computers to dig through data and do standard investment analysis with a claim that some form of “learning” process is embedded in the machine. Quite the contrary, all that has happened so far is humans using machines to find an answer to a question the human raised. That is not real AI and self-evidenced by the relatively uninspiring 5-7% annualized returns that the “AI”, deep machine learning community has delivered--no better than your average active hedge fund and much worse than passive index results.
2. Looking five years ahead, how might AI change investing as we know it today?
Regardless of whether you buy into the argument of real vs fake AI, one reality to be faced by the active management community is that in the absence of any alpha generation, their job as an active manager is and should be questioned. A computer can easily deliver sub-market returns at significantly lower cost than a human.
Real AI has the ability to deliver alpha on an absolute basis at high capacity, regardless of how our $35T market place continues to evolve. It will change the dynamic of active management to where many, especially pension fund investors, will have to question their continued reliance on human-based stock picking and smart beta factors (semi-passive) as their solution to hit their return targets.
3. What are some of the things that AI can do - or could do - that would make it especially advantageous for investors?
Easy: deliver sustained above market, absolute-like returns and minimize drawdowns. All on liquid, listed asset classes and meet the suitability needs of each investor. Investors have been let down by the active management community over the past ten years or more and, even worse, by the typical hedge fund due to its high fee structure. Real AI in investing has the ability to change that perception and reality and level the playing field for all investors. The lure of low cost passive investing in a one-way up market could be a bubble waiting to burst – adopting Real AI now can change impending reckoning for investors.
4. Does the rise of AI threaten wealth managers, or does it just mean jobs and roles shift?
Depends on the wealth manager. Some wealth managers fancy themselves as investment managers; others focus on financial planning.
The ‘investment manager’ types are generally going to find themselves behind their investor’s investment performance eight ball by subjecting themselves to the risk that they will underperform in time unless they seek to replace themselves with more AI-like investment solutions. This type is directly in the bullseye of AI and systematic investment allocation processes. Investors should be on the lookout for these “cowboy traders” and steer clear in favor of more systematic forms of asset allocation and investing.
The ‘financial planning’ oriented wealth managers, if they are open to move their thinking, can benefit tremendously by being able to better substantiate the fee they want to charge (over roboadvisors) and the value-add they bring that AI cannot, like in life and event planning, while bringing their client better investment performance. Investors should seek these types of wealth managers that are more apt to deliver on those grandiose retirement plans.
On whole, jobs in standard investment management should drop due to the influx of AI and given the general sub-standard performance of the human element in investment management. It is well-documented (by others) that the average human can cost their investor at least 2% a year in performance drag before fees!
My wife refers to me as an “Anti-CFA” because what I developed renders most CFAs as largely useless in the liquid, developed asset classes like US equities. And, she is right in that I find the CFA Body of Knowledge, upon which many trillions are invested on, is turning out to be albatross in its own right. I know this, the market proves it, and the CFA Insitute is going to now add a section on AI. For most CFAs, they are unlikely to want to face this stark reality. By the way, I am a CFA Charterholder!
5. Some very wealthy people, such as Mark Cuban and Elon Musk, see pronounced dangers when AI gets smarter than humans. What are your views on this, especially when it comes to investing - such as an AI hijacking an account and calling all the shots on a portfolio?
One should divide this risk into two forms: intentional vs unintentional.
Wanton, intentional harm caused by errant machines ala movies (iRobot, Terminator and 2001: A Space Odyssey) is not likely as it requires the AI programmer to provide sufficiently soft, adaptive instruction sets to allow the machine to find a reason to cause intentional harm. Humans and our current state of machines are nowhere near this level of transcendence at a machine level. Makes for good movies though!
Unintentionally, any machine can cause excess risk without some form of programming control to identify and address those unknown situations an AI-based system may run into. Specific to investing, Black Monday was set off by program trading as were several flash crashes as computerized programs started to proliferate. But, keep in mind that the amount of money generally under the control of one computer system is very limited relative to the overall market size and the functioning of the marketplace continues irrespective of those systems that go awry. Therefore, this risk in investing is very limited and certainly no greater or costly than all the human beings making questionable investment decisions.
6. Do you have anything else to add? Statistics and personal anecdotes are especially welcome.
Humans are the real risk in the AI space, not the machine itself. I see plenty of firms with deep resources making big claims about the value of deep learning. Unfortunately, most of these people are using standardized investment analysis (like fundamental and technical) with the argument that more data leads to a better answer (read: alpha). Nothing could be further from the truth, more data ends up becoming a panacea of sorts in the investing arena that can lead to continued sub-optimal performance. Makes for good marketing though as investors get sucked into this self-rationalization.
In 2010, we were developing our investment process, we had no conception of what an “AI” oriented investment process should look like. A good thing, as here we are many years later with a system that simplistically is able to use only price on a stock and determine in its own right whether to buy or sell it. Moral of the story, AI is not all about lots of data, it is about algorithmically being able to adapt, recognize, learn and transcend standard human-driven investment ideas, biases and responses. In so doing, an AI system designed as such can beat the very market it relies on for its chief information component: price.
Note the key difference in these two points: those who think AI can be smarter than an efficient market by betting against price. And, those (like us) who believe that AI can exist inside of an efficient market by not betting against price. The latter allows one to go to bed every night knowing that you have a $35T (US Market Cap) behind your decisioning process!
A response I use often when queried about AI-based investing: Not all machines are created equally, investor beware.