How Artificial Intelligence Could Impact Investing

This analyst blog is part of our coverage of the 2018 Morningstar Investment Conference.

Kevin Franklin of BlackRock joked that the definition he most wanted to address during a panel at the Morningstar Investment Conference was "what is artificial intelligence." In his mind, this was the term that was most fun to define from the smorgasbord of topics addressed. Similarly amusing was his response, which he confessed was from a Google query (particularly since he admitted he didn't move past the first hit). The answer he shared was that when a machine process achieves the level of an intelligence of a human, then you can call it AI. What was impressive was the boundaries of what AI is capable of, moving forward. The past, as elucidated by the panel, should serve as a reminder. In the financial field, if one were to return 30 years ago, no one would have thought to allow a machine to develop the risk model for a portfolio. Today, it would be unheard of not to systematize this process.

Other impressive developments include "word mining" through text data sets, like earnings call transcripts, for example, and the process of improving inputs. Machine learning, to a neophyte, is a dynamic process. The inputs are constantly changing to improve outcomes. Today, AI is capable of reading through an earnings call and learning and differentiating what is "bullish" versus what is "bearish." Moreover, like a child through early stage development, it appears that that knowledge grows exponentially over time. Machines can now identify what word patterns may or may not lead analysts to make rating upgrades or downgrades. The panelists confessed that the process is not 100% full-artificial intelligence as current machines are about 90% as good as a person. Even so, Franklin affirmed that machines are infinitely cheaper and more scalable.

The next frontier, it appears, is what some experts call "strong AI." While AI is general in nature, the aspiration is that strong AI extracts not just patterns in data but can also be creative. Strong AI, in theory, should be able to ask, fundamentally, "why did this happen?" and draw conclusions. Other exciting innovations include examining a video data set. Today, experts analyze truthfulness of statements, but tomorrow, machines could be capable of examining vascular dilation to ascertain whether someone is lying.

Some particular examples of data sets companies are examining while they seek to build and expand their moats include pulling massive amounts of trading data for every security in the world. With this development, firms can understand who bought and who sold which security. The advantage of this development, for example, allows firms' quant strategies to move away from crowded positions in the marketplace.

Returning to the topic of dynamism, panelists talked about the difference between their prior and current tool kits. One critical insight that was relayed was that these tools are going to be valuable provided one admits that past work is not always useful. In other words, firms have been continuously improving the models along their way. Furthermore, different firms can interpret the same data, like satellite imagery, differently. These insights, however, are not just relegated to short-term pricing dynamics, but are often extended to multiyear frameworks. Before, a quant fund may have determined that stocks with lots of equity issuance should be avoided because that was a signal of a management that allocates capital poorly. Today's marketplace has far more sophisticated methods to rate stewardship at a firm.

There was speculation that the skill sets valued in quantitative analysis haven't really changed in 15 years and machines will never fully replace humans in our lifetimes. One panelist said, "We don't use a spoon to dig after the advent of the shovel, but we do use a different skill set to achieve the same desired result."