Morning Coffee: Morgan Stanley banker says AI is more competent than his colleagues. Goldman Sachs promotes the woman who didn’t leave for a hedge fund

efinancialcareers.com:

When ChatGPT first came out, a lot of people in the investment banking industry felt that one of the most human and relatable things about it was its tendency to make something up when it didn’t know the answer.  Knowing when to admit ignorance and when to waffle is one of the higher level skills of the industry, after all.  A banker who never says “I don’t know; I’ll research it and get back to you” like Patrick Biggs at Morgan Stanley, is almost certainly a BS merchant who will blow up in the long term. 

Get Morning Coffee  ☕  in your inbox. Sign up here.

But on the other hand, if that’s all you ever say then people will eventually mistake you for the ordinary kind of ignoramus.  And a little bit of “fake it until you make it” isn’t always a bad thing. Particularly in fast moving markets jobs, a quick answer that’s nearly right is much more useful than a slow answer that’s exactly right, and top bankers are often very good at the “scientific wild-ass guess”.

Consequently, sell side trading operations have always been in a bit of a technological arms race to develop systems that let their traders and salespeople look up the answers in real time, rather than interrupting the flow of conversation and admitting ignorance.  From the first days of the Bloomberg Terminal, the dream has always been to have a helpful virtual colleague who can feed you exactly the right piece of information.

Jeff McMillan, the head of wealth management tech at Morgan Stanley, thinks that the next step forward is going to be something like “AIMS” (short for “AI at MS”), which is a model like ChatGPT, but with the fun taken out. It’s only trained on internal Morgan Stanley data, it has strict rules preventing it from giving anything that looks