Why Artificial Intelligence Isn't Picking Stocks. Yet.

Some of the resistance is psychological, experiential - and generational. Most humans - especially those with the most assets to invest - are still not comfortable with a machine making potentially life-altering decisions. Millennials appear more enthusiastic about AI-investing, but they dont have much money to invest.

But the other source of resistance comes from wealth advisors and financial services employees who believe their skills, knowledge and experience constitute a 'secret sauce' that makes them invaluably superior to computers: just like every other profession confronted by threat of machine-driven obsolescence. The problem is that as machines learn more they become more accurate in their assessments, which is not necessarily true for humans. JL


Larry Light reports in CIO magazine:

AI’s utility is to find patterns that will deliver insights and risk assessments after plowing through data troves. In the quant world, people to tell the computer what to look for. Machine learning allows computers to take a more free-form approach, to identify patterns in data without being given specific guidance (such as) examining wording executives use to see how optimistic they are. One problem AI can combat is false positives in back testing. The data AI peruses is from the past. Investing involves human emotions that may not be predictable.
Artificial intelligence (AI) is inexorably spreading its influence in financial services, as professional traders and money managers explore new ways to make it handy. Most expect AI to be a useful and beneficent sidekick like Data in “Star Trek: The Next Generation,” rather than an all-controlling demon like HAL in “2001: A Space Odyssey.”
What no one is expecting, at least for the foreseeable future, is that AI will choose stocks. For now, this digital tool will complement instead of replace flesh-and-blood money managers.
“The day where a computer picks stocks is far away,” said David Meier, portfolio manager of Motley Fool Asset Management, whose strategy is 20% quant-based. “What’s needed is human intervention to put all the pieces together” that AI unearths.
AI’s utility is to find patterns that will aid in portfolio management, delivering insights and risk assessments after plowing through enormous data troves.
An example: By analyzing satellite images of parking lots outside stores, AI gets a better picture of consumer preferences. Or to understand a company’s future direction, by examining wording that executives use on earnings calls to see how optimistic or pessimistic they are about the future. Or to ferret out possible weaknesses by sifting through filings to see if the risk disclosures sections have added any new risks. Such additions can be a warning sign.
One chronic problem AI can combat is false positives in back testing. Here, after numerous hypothetical dry runs using historical data, someone discovers a new investment strategy. Trouble is, in the real world, going forward, the idea doesn’t work.
A big reason is how the testing is conducted, looking at multiple strategies and then choosing the best, according to Marcos Lopez de Prado, chief executive of True Positive Technologies, a registered investment advisor with an expertise in machine learning. “It is the equivalent to buying all the lottery tickets, and claiming afterwards that we knew what lottery ticket would win.”
While the full extent to AI’s potential is unknown, it surely will be big. That’s why Vanguard, Fidelity, and BlackRock have all set up huge AI facilities and are on the leading edge of this revolution.
Numerous hedge funds are using AI to aid in their research and trading. And the concept will become increasingly important, according to Ray Dalio, founder of the world’s largest hedge fund, Bridgewater Associates, who told Recode Decode that “it’s just the tip of the iceberg.”
In the quant world, the standard procedure has been for people to tell the computer what to look for. A human might hypothesize that there’s a positive relationship between stock valuation and performance—and create an algorithm to burrow into the data to see if that is so. With AI, the machine looks into the question on its own, and determines what the truth is, and also ropes in other variables that may influence performance in conjunction or in opposition to the valuation factor.
With AI, the machines learn from experience. As an inexhaustible helper eyeing enormous amounts of information and in search of new patterns and relationships, AI is the ultimate quant strategy.
Thus, “machine learning allows computers to take a more free-form approach, aiming to identify predictable patterns in price data without being given specific guidance about what underlying relationships may look like,” wrote Graham Robertson, head of client portfolio management for Man AHL, the quantitative investment manager, in a report.
One function AI can help with, he noted, is the direction of price data—not simply gauging whether prices rose or fell, but the subtleties of the movement. To Robertson, “it is not only important that prices went up a certain amount over the last year, it is important the path they took getting there.”
Another task is to understand how “slippage” occurred in trade execution, the gap between a trade’s expected price and what it actually closed at. The goal here, Robertson maintained, is to find “the lowest cost of market access while causing the minimum of market impact.”
To be sure, digital tools are ubiquitous throughout financial services. BlackRock, for instance, created software three decades ago that screens portfolios and analyzes their weak spots. The world’s largest asset manager has long used the program, called Aladdin, for internal operations and has bolstered it with AI techniques.
Now BlackRock is rolling Aladdin out to financial advisors. In March, BlackRock unveiled seven exchange-traded funds that will use AI to peruse company public filings to determine asset weightings.
To many money managers, AI is an intriguing notion, with too many question marks attached. Tom Plumb, chief investment officer of Plumb Funds, who doesn’t use AI, thinks it could be a valuable tool in enhancing efficiency and “has the potential for disruption” in traditional portfolio management.
But what if new data emerges that upends an AI insight? And only a living, breathing money manager could see the impact? The data AI peruses is from the past, and investing involves human emotions that may not be predictable when confronted with obstacles and opportunities in the future.
“AI could help in the short term,” Plumb contended, “but for the long term, you need a human.”