| Illustration by Paul Ryding. |
When I first told Angelo Calvello that the summer issue of Alpha would focus on quant funds, he responded with skepticism. Calvello, the writer of Institutional Investor’s “The Dissident” column, has never been one to mince words. Although himself a proponent of investment strategies based on data science and artificial intelligence, Calvello is no fan of the label “quant” — a term that, to him, has taken on a too-broad definition, applied to advanced machine-learning techniques and simple factor-based strategies alike.
A magazine focusing on quants, he worried, would struggle with the same problem now facing investment managers and asset allocators: how to differentiate between systematic investing, true quantitative strategies, and modern artificial intelligence when both the techniques themselves and the language used to describe them often overlap — a problem made even worse by the “black box” nature of some of the most advanced strategies, which even computer scientists can’t fully grasp.
But despite his doubts, Calvello agreed to meet me for dinner — and, perhaps, a vocabulary lesson. We met in Chicago’s Little Italy, at an unmarked eatery Calvello refers to fondly as the “family restaurant.” Owned and operated by his cousin, Joe DiBuono, Tufano’s is quintessential Chicago Italian: According to family legend, DiBuono’s grandfather Joseph Di-Buono founded the restaurant in 1930 after losing his job as a personal chef; his boss, Al Capone, had just gone to jail.
Both Calvello and his wife, Lisa, who joined us for dinner, actually worked at Tufano’s for a few years in the ’80s. Over stuffed artichokes and salad, the couple debated whether the area where we were seated existed back then, or if it was added as a part of later renovations. Everyone at the restaurant seemed to know the Calvellos, with waitresses stopping by to ask after their children and offer drinks on the house.
Eventually, after we dined on family-style dishes of pasta, chicken, and calamari, the conversation turned to quantitative investing.
“A lot of firms say they’re using machine learning, and they really don’t,” Calvello says. “A lot of firms claim to use big data, and they really don’t. There’s a lot of people just jumping on the bandwagon.”
Calvello would know the difference — his firm, investment management startup Rosetta Analytics, uses machine learning and big data to create investment strategies for asset owners, employing computer scientists to develop advanced algorithms and neural networks. For most members of the industry, however, artificial intelligence remains a murky subject — difficult to explain and still more difficult to understand.
“The problem is that people don’t understand AI,” Calvello says. “They have a view from the 1990s — they don’t understand its current state.” The machine-learning techniques pioneered in the pre-millennium years, he explains, relied heavily on human design. A computer could be programmed to play chess, but it needed a team of IBM scientists to input all the possible moves.
That is no longer the case. In recent years, there’s been “an explosion of data, an exponential increase in computer power,” Calvello explains. “We have reached a point where computers can surpass humans in certain complex skills.” Including, perhaps, investing.
“People like to say that AI will never be able to understand markets,” Calvello notes. “But they’re biased — they want to defend the existing business.” To truly embrace artificial intelligence would mean “radical change,” and the traditional active management firms now profiting off the supposedly superior skills of human portfolio managers are understandably loath to disrupt the status quo.
Calvello, however, has never been one to fear breaking with convention. My own introduction to him was at an investment forum hosted by my then-employer in late 2015, where Calvello’s presentation ranged from big data to the legalization of marijuana and included a blown-up shot of Bob Marley smoking a joint to illustrate the argument that attendees would find cannabis in their portfolios within the next 15 years.
Calvello had founded Rosetta Analytics earlier that same year, with the goal of putting into practice some of the ideas he had been thinking and writing about over the previous few years — such as partnering directly with asset owners to solve investment problems and hiring polymaths and data scientists rather than CFAs and MBA grads.
The quantitative investing techniques employed by Rosetta would not be the human-driven machine-learning methods of the ’90s and early aughts. Instead, Calvello and his team would use deep learning — a new, advanced form of artificial intelligence in which machines train themselves to perform tasks through exposure to vast amounts of data. The most prominent example of this technique is AlphaGo, an AI system created by Google’s DeepMind that earlier this year defeated the world champion at Go, an ancient Chinese board game considered to be far more complex than games like chess and checkers.
“What’s overwhelming is you’re using models with millions of parameters,” Calvello says of deep learning. “People can’t understand it because it’s a different dialect.”
This lack of understanding is a problem, both because it makes it difficult to identify which managers are actually using advanced artificial intelligence and because investors are hesitant to put money into a strategy that can’t be explained — a theme Calvello expounded on in a recent II column.
He insists investors don’t need to fully understand how artificial intelligence works: “If the results are good, you don’t need an explanation,” he argues. “You need trust.”
Still, any responsible fiduciary needs some basis for manager selection. I ask Calvello how asset owners will be able to identify those managers that are actually worthy of trust.
“For one, they need to understand what kind of AI a manager is using,” he says. “They can’t just take the words ‘quant’ and ‘big data’ at face value.” This, of course, requires education.
It also entails hiring data scientists rather than CFAs, an area in which university endowments are at an advantage — Calvello suggests these allocators can leverage the knowledge of their institutions’ faculties and students. Other asset owners could try direct investments, funding artificial intelligence startups and sharing in the resulting intellectual property. “It will be hard,” he concedes, “but the way we do things now isn’t going to be enough — we have to change.”
By this time, we’ve finished dessert and it’s time to pay the bill; Lisa, a meditation coach and yoga teacher, has an early class the next day.
On our way out, a waitress stops us to ask if I got the information I needed. “None of us really understands what Angelo’s firm does,” she jokes.
In that the Tufano’s staff are not alone.
