NYC

AI Street Meetup - Tonight!

Tonight, with Jordan Hauer of Amass Insights, I’m co-hosting a meetup with folks interested in the data + LLMs + investing space starting at 6:00 p.m. in Midtown Manhattan.

We have 50+ signed up between the two groups. So come by! Just fill out this form for location details. See you then!

STOCKS

AI Reveals What Investors Really Think About Stocks

Using transformer models on portfolio data, researchers uncovered relationships missed by traditional financial metrics.

Last week, we touched upon how LLMs infer meaning in ways previous machine learning couldn’t. Here’s what I wrote:

A very short recap of the evolution in AI:

The current AI boom began with a 2017 Google paper on how to better translate text. Before that, machine learning generally processed language word by word, and struggled to capture contextual relationships between words. For example, if I say to you, Paris → France → Berlin → ______ you know it’s → Germany — something earlier machine learning struggled to do.

Now imagine applying this LLM infrastructure to a dataset where you don’t know that the answer is Germany, for example, like stock market holdings data.

That’s what researchers roughly did in a recent paper titled “Asset Embeddings.” I recently wrote about their findings for the Chicago Booth Review, the research publication of the University's Business School.

They used institutional portfolio holdings data to create “embeddings” — vector representations of each stock based on who held them. The embeddings grouped stocks by themes such as growth, sector exposure, ESG preferences, and macro sensitivity.

To test their accuracy, the researchers withheld some holdings from the training data and asked the model to predict the missing stocks — and it beat traditional metrics at explaining valuations, comovement, and investor behavior under stress.

Check out the full writeup here if only to see how I managed to sneak in a LeBron James analogy into an academic publication about AI & stocks. 😎 

To understand embeddings, think about basketball star LeBron James. Traditional statistics indicate that he is a 6 ft., 9 in., 250-lb. forward. But fans know he can play point guard in some lineups, and center, shooting guard, small forward, or power forward in others. His role depends on who else is on the court.

Takeaway:

Large language models predict the next word / token, and this architecture can be applied beyond language to stock market holdings data to uncover new relationships.

SPONSORSHIPS

Reach Wall Street’s AI Decision-Makers

Advertise on AI Street to reach a highly engaged audience of decision-makers at firms including JPMorgan, Citadel, BlackRock, Skadden, McKinsey, and more. Sponsorships are reserved for companies in AI, markets, and finance. Email me ([email protected]) for more details.

AI FED

AI Models Simulate FOMC Decision-Making

Researchers at George Washington University used AI agents modeled on real Federal Reserve policymakers to simulate a July 2025 FOMC meeting. Each agent was trained on a member’s historical policy stance, biography and speeches, then fed real-time economic data and news to reach a rate decision.

The results: when the simulation was subjected to political pressure, the AI agents became more polarized and dissent increased. The authors, Sophia Kazinnik and Tara Sinclair, conclude that the Fed is “only partially insulated from politics” and that outside scrutiny can influence even rule-based decision-making.

To better understand the implications of their findings, I asked Kazinnik about how LLMs might shape future policy analysis.

“LLMs could absolutely be used for modeling different scenarios. These tools are actually really great for generating and exploring ‘what if’ scenarios. However, I think it’s important we don’t use them blindly. Simulations need to be grounded in historical data, validated against out-of-sample episodes, and paired with traditional tools like event studies, DSGE models, and market micro data.”

She emphasized that the approach could be extended to other decision-making processes:

“Absolutely; this approach can be modified and adapted to other decision processes – as long as you give agents a realistic information set and constraints.”

Adoption of LLMs in financial modeling may be just a couple of years away.

“It will take some time (1-2 years perhaps), but it's headed in that direction for specific classes of problems. LLMs are becoming a standard tool in the financial modeling toolkit, especially for problems involving language or human behavior. LLMs are great at reasoning under qualitative uncertainty and synthesizing information from multiple sources — both of which are areas where traditional quantitative models struggle.”

Study: FOMC In Silico: A Multi-Agent System for Monetary Policy Decision Modeling

Takeaway: 

Running an LLM-driven policy simulation where human behaviour is central could soon be as routine as a Monte Carlo run.

QUICK HITS

AI Agents Are Getting Ready to Handle Your Whole Financial Life

AI promises to reshape Wall Street—and individual investors—like few other tech changes in its history. (WSJ) ← covers many of the topics written about here.

DeepMind’s Demis Hassabis says calling AI PhD Intelligences is ‘Nonsense’

The DeepMind chief said AGI is still five to ten years away, pointing to missing capabilities such as continual learning and intuitive reasoning. (AIM)

Goldman Sachs Partner on the Pros and Cons of AI

Kerry Blum, a partner in Goldman’s private wealth management business, explains how its AI assistant saves employees hours each week — but warns it should not be treated as a ‘source of truth.’ (FT)

Alipay Launches Agentic AI With Luckin Coffee

Luckin Coffee has integrated with Alipay's AI agent for agentic ordering and payments via a conversational interface. (Finextra)

ADOPTION

Neudata Survey: AI Adoption Climbs

Neudata published its annual data buyer/seller survey and the results show a meaningful jump in AI adoption across the industry:

• AI use nearly doubled for key use case – Twice as many firms report using AI-processed data to optimize investment and trading strategies (14% in 2024 → 31% in 2025).

• Productivity gains are widespread – Two-thirds say they’re using AI to enhance internal workflows.

• Foundation models lead the pack – Data buyers most commonly reported using foundation models.

• Top dataset demand – Investment managers are most interested in IT, consumer discretionary, and financials datasets — likely to track the AI boom and macro volatility.

The survey, conducted between June and August, drew 171 responses globally, including 67 data buyers and 104 data providers. Quants made up the largest share of investor respondents, followed by discretionary and macro managers.

WHAT ELSE I’M READING

• US data center build hits record as AI demand surges, Bank of America Institute says (Reuters, BofA report)

• BNY launches $10M AI lab at Carnegie Mellon University (Axios)

• Trump Calls for Ending Quarterly Earnings Reports (WSJ)

• British Startup Installs New York City’s First Quantum Computer (Bloomberg)

• Hedge Funds Have a Reputation for Ruthlessness. Dmitry Balyasny Took a Different Approach. (Institutional Investor)

• Powell: ‘Unusually Large’ AI Investment Masks Soft Labor Market (Fortune via Yahoo)

• White House Rolls Out New Action Plan to Speed AI Development (Bloomberg Law)

• California lawmakers pass AI safety bill SB 53 — but Newsom could still veto (Tech Crunch)

CALENDAR

Upcoming AI + Finance Conferences

New entries bolded*

• Cornell Financial Engineering Manhattan 2025 Future of Finance & AI Conference – Sept 19, 2025 • New York (I’m attending)

  A one-day forum on AI, quantitative finance, and hedge-fund strategies, attracting leading quants and industry practitioners.

• Bloomberg-Columbia ML in Finance Conf – Sept 25, 2025 • New York

  Academic–industry event hosted by Columbia University and Bloomberg, focused on ML applications in finance including asset pricing, market forecasting, and LLM risk.

• GAIIM Conference 2025 – Sept 30, 2025 • New York

  Forum on practical applications of AI in investing, featuring tools for research, valuation, and portfolio workflows.

• Open Source in Finance Forum - Oct. 21-22 • New York

  Finance and tech leaders tackle how open source and AI can be governed, scaled, and applied in financial services.*

• AIFin Workshop at ECAI 2025 – October 26, 2025 • Bologna, Italy

  One-day academic workshop on AI/ML in finance, covering trading, risk, fraud, NLP, and regulation.

• AI in Finance 2025 – October 27–30, 2025 • Montréal

  Academic event covering ML in empirical asset pricing and risk.

• ACM ICAIF 2025 – November 15–18, 2025 • Singapore

  Top-tier academic/industry conference on AI in finance and trading.

• AI for Finance – November 24–26, 2025 • Paris

  Artefact’s AI for Finance summit, focused on generative AI, future of finance, digital sovereignty, and regulation 

• NeurIPS Workshop: Generative AI in Finance – Dec. 6/7 • San Diego One-day academic workshop at NeurIPS focused on generative AI applications in finance, organized by ML researchers.