Where Are the Customers’ Yachts... and Bots?

Fred Schwed’s book, which eponymously contributed to the title of this post, was written some eighty years ago, but that is not where this story begins. You need to go back yet another eighty or so years, to the late 1800’s, when this phrase was first turned by William Travers, a very successful lawyer and Wall Street financier. At the time, Travers was admiring the many large yachts in the Newport, RI harbor and learned that they were all owned by the Wall Street crowd, with none belonging to the clients for whom they (dutifully) served. What would be the modern-day equivalent quip from Travers, or the title of a sequel book from Schwed? Let’s ask ChatGPT as we now look out on a sea of bots.

As it turns out, ChatGPT claims it “doesn’t have access to real-time information or data about specific customers’ bots”, but further clarification was forthcoming and maybe quite telling in that “the specific location of customers’ bots would depend on the implementation and hosting choices made by the respective businesses (emphasis added).” Sounds like we have found the appropriate working title!

No doubt the speed and breadth of generative AI and its many LLM (Large Language Model) cousins should be embraced. Its ability to harness NLP (Natural Language Processing) tasks is, and will continue to be, a game changer for financial analysis, and Bloomberg has 50B (parameter) reasons as to why we should take heed. But let us not forget that genius once (spectacularly) failed when “finely tuned computer models had (theoretically) tamed the genie of risk” (ChatLOL!).

Maybe this time it’s different, and when a respected institution like JP Morgan talks, people listen (or was that EF Hutton?). Just last week, JPM announced that they are developing their own generative AI model (IndexGPT) that will actually “select investments for their customers” who are perhaps saving to buy a yacht of their own? When you read the closing quote from their global tech chief, “We’ve recognized the power and opportunity of these tools and are committed to exploring all the ways they can deliver value for the firm (emphasis added),” it is hard not to be at least a bit cynical and the bot happens to agree!

When queried, ChatGPT emphatically starts with this opening salvo: “No, a bot does not have a fiduciary duty.” It goes on to say that such a relationship is demarked where “trust and confidence are placed in one party on behalf of another” and that bots “do not possess the capacity for the type of judgement, discretion, and ethical responsibility that would be required to fulfill a fiduciary duty.” Turns out the bot is keenly self-aware, maybe to a fault, especially when the friendly regulator comes a-calling.

It also turns out that we have yet to see a bot register as a Candidate for our FDP exam, but then again, we never expected that to happen. We did not need to wait for the bot to state its limitations when it comes to the client. When we developed the FDP program several years ago, a cornerstone needed for this highly specialized course of study was the new and complicated questions related to privacy and the ethical use of unstructured data. Embracing change and understanding the use case for new technologies are part of what we must always do as professionals. But let us never forget that we are responsible for the customers’ hopes and expectations (which may include a yacht, or maybe a bot of their own), as they are the ones who have placed their assets with us, and their trust in us. Your bot and all its decisions, limitations, and outcomes must ultimately be understood and owned by your client.

Seek education, diversity of both your portfolio and people, and know your risk tolerance. Investing is for the long term.