AI Is Going to Change Investing. Here’s How It Could Play Out.

I dutifully relayed what my pal, George Fellows Jr., then on the options desk at Drexel Burnham Lambert, described to me: that program trading entailed having a computer algorithm buy or sell baskets of stocks, sometimes arbitraging against an index. I remember the editors shaking their heads in dismay and disbelief—like the bankers in Mary Poppins.

Wall Street’s relationship with technology is longstanding, deep, and uneasy. Yes, computers back then were practically abacuses, but they still put markets on the brink. In fact, major technological leaps such as processing-power gains in the 1980s and consumerization of the internet in the late 1990s seem to presage stock market booms and busts. To wit: the aforementioned ’87 crash and the tech bubble, respectively. More recently, we’ve experienced smaller-scale flash crashes (May 2010 and April 2013) and high-frequency trading glitches.

And now, in this incipient age of artificial intelligence, we could well see more Wall Street tempests, big or small. The wild action in
Nvidia
stock (ticker: NVDA) and the recent momentary market dip on a fake, AI-generated photo purporting to show a fire near the Pentagon may be one-offs—or harbingers.

Broadly speaking, what are we to make of AI? Last week my colleague Randall Forsyth wrote about its impact on the economy. Another consideration, of course, is AI as an investment, not only Nvidia, but
Microsoft
(MSFT), Google parent
Alphabet
(GOOGL), and
Marvell Technology
(MRVL)—and if you can get your hands on private darlings like ChatGPT’s parent, OpenAI, or Anthropic.

Investors are right to be skeptical, though. Over the past decade, Silicon Valley has produced and shilled the likes of cryptocurrency, nonfungible tokens, virtual reality and extended reality, the metaverse, and next-to-worthless apps like the one that tracks the weight of your dog. Does AI really rise above that roster of dubious innovation? A number of veteran tech executives and investors such as Dave Roux, co-founder of Silver Lake and BayPine, say yes.

“I wouldn’t say that AI is the same as fire or electricity,” Roux says. “But it’s probably more important than—or as important as—personal computers and mobile phones. And think of how large those technologies have been.”

Like the PC and cellphones, then, AI has the capacity to rewrite how almost every business works—including investing. Though it’s still early days, I decided to delve into how AI might change things for retail and institutional investors. Spoiler alert: Bring grains of salt.

Matt Hougan, chief information officer at Bitwise Asset Management and a board member at EquBot—an AI-based investment platform which oversees the
AI Powered Equity
exchange-traded fund (AIEQ)—calls AI “part of a long-term flattening of the informational advantage between the most well-resourced institutions and the least-well-resourced retail investors that has been going on for decades.” In other words, AI could give retail investors tools that heretofore had only been available to quant hedge funds like Renaissance Technologies and D.E. Shaw. That sounds nice in theory, but practically speaking, the gap between “us” and “them” is miles wide, and AI still has a long way to go.

“I’m having a hard time figuring out how AI will change investing,” says a former chief risk officer at one of the big quant firms. AI can’t necessarily be trusted, he says, and “successful quants rely not only on smart algorithms but also on extremely efficient market execution. AI probably doesn’t help with that.” And three, he says, “AI would need to be packaged in a way that makes it usable. You can’t just tell AI: ‘Go into the market with $X of capital and make money.’ There needs to be some kind of platform or structure. Could that platform be offered to a retail investor at a price that makes sense?”

Good questions.

Nick Bostrom, a noted AI expert at Oxford University with a sometimes dystopian take, says AI will be a boon to investors who need to digest vast amounts of texts, “whether it’s a new annual report that can be read in half a second, or reading what people are saying on social media.” But Bostrom is skeptical about some of its supposed advantages. “Can AI predict better than, say, human experts would whether there will be a coup in some particular country or an expropriation of a big oil company? I’m not aware of any data on that currently,” he says.

Still, banks and broker-dealers are lurching forward, manifesting a curious dichotomy along the way. Bloomberg reports that many, including Bank of America, Citigroup, Goldman Sachs, and JPMorgan Chase, have banned ChatGPT at work—small wonder given nightmarish scenarios like at Samsung, where engineers leaked proprietary code onto ChatGPT. Oops.

Other banks, though—or in some cases, different departments at the same institutions—are cautiously experimenting with AI. In this camp are Goldman, which is using AI to help its developers write code, and J.P. Morgan, which is using AI to detect trading signals from Federal Reserve statements and speeches (as it perhaps did on Thursday when Philadelphia Fed President Patrick Harker said he supported skipping an interest-rate increase at the central bank’s June meeting).

BlackRock has set up BlackRock AI Labs (“At BlackRock, we believe that AI can help meaningfully improve financial well-being”), replete with an impressive roster of scientists, though it isn’t exactly clear what they do. A source at BlackRock says stay tuned for AI news soon.

Morgan Stanley may well be the furthest along with AI. About a year and a half ago, a Morgan Stanley team met with OpenAI, says David Wu, a managing director at Morgan Stanley Wealth Management. Today its financial advisors are being trained to use an internal tool based on GPT4 (OpenAI’s most recent version), which allows employees to have a “live chat” with all of the firm’s data and research.

“A few months ago, if you asked your financial advisor, ‘Hey, what do you think about
Apple
[AAPL],’ they might have to go through a research portal to look up that latest report,” Wu says. “Oftentimes the advisor will say, ‘Let me get back to you later in the day.’ We’re building a chat interface that allows the FA to ask in real time what’s our latest outlook on Apple, and then have the machine go through our entire corpus of internal content and synthesize an answer in two seconds.”

Is that a blessing or a curse?

What about bypassing broker-dealers and using ChatGPT directly to invest? To get at that, here’s a precis: ChatGPT, the most-prominent generative AI platform, became available on Nov. 30, 2022 (a day perhaps analogous to Aug. 9, 1995, when Netscape went public and ushered in the era of the consumer internet). If you haven’t played around with ChatGPT—and I highly recommend that you do—it’s worth noting that it doesn’t include content after September 2021. Which paradoxically means if you ask ChatGPT if it is available to the public, it essentially says no. Also, ChatGPT won’t give investment advice. If you ask it what stock to buy it spews boilerplate blah-blah:

“As an AI language model, I cannot provide personalized financial advice or make specific stock recommendations. Deciding which stocks to buy involves various factors such as your financial goals, risk tolerance, investment horizon, and individual research. It’s generally recommended to consult with a financial advisor or do thorough research on your own before making any investment decisions.”

ChatGPT can, however, answer questions about the time value of money or the difference between stocks and bonds, and so forth. Of course, there are other generative AI sites only too happy to have their algos advise you. (Yes, be afraid.)

For a reality check on all this, I called on a couple of Wall Street wise men, the Wharton School’s Jeremy Siegel and Princeton University’s Burton Malkiel, with some 120 years of market experience between them. You won’t be surprised they were in the “nothing new under the sun” camp.

First, Prof. Siegel. Does he think AI will pick our stocks and we can just go to the beach? Nope. “Don’t forget what AI does,” Siegel says. “ChatGPT scans everything that’s been written. I don’t see how that really helps pick stocks. I think AI has great possibilities, but I don’t think it has great possibilities for stock selection.”

Over to you, Prof. Malkiel. “I think there’s no question that if I am a security analyst, I am going to be able to do a better job than I had done before [because of AI],” Malkiel says. “But in no way does it affect my view that the individual investor should not be a stockpicker. If anything, with AI, the market will do a better job of having all information impounded into market prices. The paradox is that as everybody gets better, then it’s even tougher to pick individual stocks than it was before.”

Apparently it’s tough even for AI. Since its inception on Oct. 18, 2017, the AIEQ ETF is up 16.1%, compared with the S&P 500’s gain of 63.9%. Also, just this past Tuesday, ChatGPT was ridiculed for being easily fooled by this question: “Which weighs more, a pound of feathers or five pounds of lead?” (It said they weigh the same.)

Still, I went back and asked that same question two days later and got this: “Five pounds of lead weighs more than a pound of feathers. While both items are measured in pounds, five pounds is a larger weight than one pound. Therefore, in terms of weight, five pounds of lead is heavier than a pound of feathers.”

Bottom line: ChatGPT and the rest are nowhere near ready for prime time for investors. But think about this: Tonight you will climb into bed and go to sleep. And while you’re off in the land of Nod, ChatGPT will be working away, getting better.

Write to Andy Serwer at [email protected]