In the beginning, investment management professionals studied corporate financial statements and business strategy — fundamentals. Then analyst work evolved to include more and more technology — quantitative tools/techniques — to the extent that investment management could be done with fewer, and in some cases no, investment analysts with the work being done by an algorithm. In 2017, artificial intelligence (AI via machine learning) capabilities became such that questions began to be asked such as: Will more investment analysts begin to use AI tools, only to be ultimately replaced by a "robot?" In short, the answers are yes (to using tools) and at least not yet (to being replaced.)
Investment pros — like all people — have bias in their decisions. Given that the job is just a lot of decisions, that might be worrisome. In fact, Nobel laureate Daniel Kahneman has stated his preference for quantitative/algorithmic decision-making since the bias — the quant/algo systems are built by people — can be detected, measured and corrected more easily. Mr. Kahneman's concerns about bias are well-founded, and there's no doubt that the bias in these models is easier to detect and correct compared to trying to do that with a person.
However, in my work with investment management teams, I know that decisions can be de-biased through team structure; decision rights such as the leader vs. the majority vs. consensus decides; process design; keeping a record of decisions and rationales; and systematic structured reviews of the work product. Aside from any investment team preference, both fundamental and quant/algo techniques can work and are complementary.
Now it's 2018 and AI is even more robust, with abilities like recognizing sights and sounds as well as processing research faster than humans. Should investment professionals start to rethink their careers? Again, not yet. Let's start with a little perspective on AI and investing. AI requires two things: computational power and big data sets. AI does one thing — it organizes and contextualizes data. The computational power exists today. While the data exists always and everywhere, it is rarely clean and usable, saved and accessible, and universally available such as public company financials.
Data scientists say clean data is the exception more than the rule, and current industry professionals don't broadly know how to clean data to make it fully usable. Yes, oodles of data are saved somewhere. The ever-cheaper price of memory — one gigabyte of memory costs a few pennies today vs. $100 20 years ago — helped usher in the age of data warehousing. This happened because it was cheap to save data even though its value may not have been fully respected.
Availability is an issue. Those who have all this data today understand its value and don't share it universally or freely. To best profit from their data, they may only be able to share it on a limited basis thus making it costlier for any buyer. Note that my use of the term data is meant to be complete and include what's called "alt-data" sets, too. Alt-data is simply less generally publicly available. For example, satellite imagery of cargo ships and their buoyancy to better understand the amount being shipped. The value of this information to those who have it is real.
There are many data challenges, and those challenges are arguably too great for most investment pros. Moreover, consider how unique the last 20 years of data may be on the specifics of the 20-year period of history, the regime in which so much of the saved data was gathered. And, with the advent of the European Union's General Data Protection Regulation (and whatever the future privacy regulations will bring), the ability to gather data in the future will likely be more expensive and less robust. AI's big data requirement will likely be impinged.
The manifold data challenges make me think that instead of the old parable "water, water, everywhere but not a drop to drink," we now risk having "data, data, everywhere but too few bytes nutritious enough to eat." Furthermore, once you admit that data is not knowledge, then regardless of how AI may be more robust, it's not yet ready to broadly replace investment pros.
So, should investment pros be comfortable as a result? No. Any past informational edges they may have had to make better decisions very well might become less and less effective if only due to the costs and complexities of data. While the state of AI today could be a boon to investment analyst research specifically, the potential tool in the toolbox is not easily or universally affordable.
For the state-of-the-art data players or warehouses, it's a seller's market and great barriers of entry for new competitors exist. If Google became an investment manager, other managers would likely become concerned. Nevertheless, investment pros have always been able to strive to make better decisions on the information they glean from existing data vs. simply requiring more data. Investing is all about better decisions, and while better data can help, it is not enough. Showing evidence of good decision-making fits with well within the topic of "evidence-based investing," a worthwhile attainable goal today for investment pros.
Maybe investment pros can take solace from the words of Pablo Picasso: "Computers are useless. They can only give you answers." AI is only Almost Intelligent and not yet able to replace investment pros because it is they who imagine the questions.