I joined AI expert Aziz Lookman, Ph.D. and Capital One’s director of machine learning? Sonu Chopra-Khullar on a panel organized by moderator Nicole M. Allen from the LSEG (London Stock Exchange Group) to talk about how hedge funds are using large language models.
My main takeaway was that the fascination on Wall Street with how the AI models work is giving way to a focus on what can be done with them.
Also, as much as there is widespread agreement that generative AI is mind-bendingly revolutionary, larger companies are struggling to adopt it.
Sonu made excellent points, notably to remind the audience not to think about Gen AI too narrowly. The technology will be applicable in arenas we aren’t even thinking about now. Aziz challenged the group to allocate some small amount of money to everyone at their firm to experiment with AI solutions.
I talked about four startups that illustrate how AI is being used. Broadly, they are focused on summarization, text generation, model creation and text analysis. All address the practical issue of how to reduce time spent by analysts compiling and assessing information.
DoTadda, Inc., a start up run by Andrew Meister, started out as a way to summarize and query transcripts, especially earnings conference calls. The company has added historical documents to enable analysts to ask more complex questions.
I asked DoTadda this question: "What geographic area is Exxon showing the most interest in over the last 12 quarters and how is it changing?"
The answer:? Guyana, The Permian Basin, Mozambique and Qatar.
Pricing Culture, a startup run by Bhargav Shivarthy, is taking arcane, proprietary data sets to generate text, thereby making the information more accessible.?
I used the system to tap into data assembled by Carbon Arc and generate a story about fluctuations in discretionary spending based on credit cards data.
The big idea here is that most data sets only make sense to a small number of experts. Text, especially when context is included, makes it useful to a larger population of analysts.
SigTech is a startup funded by Alan Howard that taps into AI to provide answers to complex scenarios, many of which would have previously required quants to answer.
Yesterday, I asked the system: "What jobless claim number would justify a 50 basis point cut?"? It replied: "The answer is 450,000 or above combined with other eco weakness like jobless over 4.5%." It also provided historical context for the basis of making that determination.
Markets EQ, co-founded by Sean Austin, captures earnings conference calls and other audio files to generate transcripts, summaries and analytics .
Tone is important because studies have shown it accounts for as much as 38% of the information we communicate when we speak.
The analytics include the speed of speech as well as? tone, which can indicate sentiment and stress at odds with the words someone is literally saying.?
