A Very Brief Gen AI Operating Manual for Long/Short Investors

The small text box at the top of page responsible for most of the questions we ask Claude, GPT or whatever other stovetop you cook your artificial thinking on, provides a consistently good synopsis of topics. And what it lacks in depth, it compensates for in plain, unambiguous, low nuance speak. My AI still doesn’t get my Jersey dialect - will this be the one barrier it simply cannot break? – but if it did, it would say, “Yo! try these 5 things to find longs that go up and shorts that go down, while providing enough breadcrumbs for everyone else to follow along.”

1. Validate the data used to derive inference.

It’s easy to ask the question. And now, the response comes back within seconds of putting in the query. We can ask so many questions. We can save so much time. And we are relying on the response’s correctness with a high confidence interval. But it’s not about whether the AI is right or wrong. It’s about understanding how the AI got there, because from an investment standpoint, we are still going to be wrong a lot. And when you are wrong, it is very unlikely that your boss, PM, or fund investor is going to be satisfied with an investment process that consists largely of, “I asked Chat GPT a series of questions and was comfortable that I understood the landscape well enough to add the position.” No, that is not the explicit role of the human in the loop. You, human, are supposed to know the underlying facts as well. If you forego learning them, you will not be well-positioned to move when the trade goes wrong. You are using the data to create a forecast. Make sure the forecast’s facts are grounded in a reality that you not only understand, but that you have validated. We don’t want science without the scientific method. We can’t take outputs on faith.