Vector Database Ideas To Steal From Wall Street Quants (Vol. 8)

Wall Street quants are wicked smart. Here's what they're doing with generative AI and vectors and some ideas you can steal.

What Wall Street traders are doing with generative AI

"Machine learning can find patterns based on economic models,” says Stefan Zohren , principal quant in Man Group’s trading division. “But it can also find many other ones that are potentially not so intuitive,” he says, “which is an advantage because fewer people might have found them. However, it won’t always be clear how reliable the new patterns are.”

Wall Street traders are experimenting with hybrid time-series and vector databases to explore the edge cases of trading, not to find “answers.” They use the output of generative AI as food for thought and ideation. For spit-balling. ??

Read Can AI Beat the Market? Wall Street Is Desperate To Try to learn more about how Wall Street traders are experimenting with generative AI and vector databases for trading strategy ideation (Bloomberg subscription required).

What AlphaGo teaches us about creativity and how it relates to using GenAI for trading strategy ideation

“AlphaGo was stuck. It had no coach, no human intervention, and no lesson based on an expert’s experience. It didn’t accept the narrative of how to play the game properly. It wasn’t held back by limiting beliefs. With a clean slate, AlphaGo was able to innovate. This is the beginner’s mind – one of the most difficult states of being to dwell in, precisely because it involves letting go of what our experience has taught us.”

Generative AI, when it encounters a scenario it hasn’t seen in training, often “makes up” an answer. Some consider its hallucinations as mistakes. Others, like legendary music producer Rick Rubin, see “made-up answers” as a form of creativity.

This is the essence of how some quants use generative AI as an ideation co-pilot that, when it gets stuck, can make surprising, hallucinogenic suggestions that can be inspirational

Listen to Rubin’s reaction to the AlphaGo story with Krista Tippett on the On Being podcast (Magic, Everyday Mystery, and Getting Creative).

Vector databases that support prompting break down expression barriers

When developers use natural language to write and troubleshoot code, it can help break down expression barriers. Language makes it easier to express concepts and generate ideas.?For example, vector databases that feature a?natural language interface help developers explore data, develop hypotheses, and evaluate data quality more directly than with SQL, Python, R, PHP, or C#.

Hybrid time-series vector databases help analysts compare “as-if” scenarios of different time dimensions to simulate how trading strategies work in various market conditions. For example, during extreme market volatility, some trade execution strategies will do better than others, so finding similar market conditions to current conditions helps traders and quants compare trading scenarios and select the best fit.

Read about vector databases and prompting in The Implications of Prompt Interfaces (Vol. 7).

{The, Weekly, Vector} is a weekly newsletter of {Three, Curated, Perspectives} on generative AI and vector databases. It is a free newsletter from KX.