Apr 30, 2024 · 37m · a16z
Marty Chavez (Sixth Street): Finding a Single Source of AI Truth
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of a16z's 'In The Vault' podcast, former Goldman Sachs CIO/CFO Marty Chavez discusses his career arc, the power of digital twins in modeling risk, and strategic frameworks for enterprise AI adoption and regulation.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 2.2 Guest teaching 5.1 Guest disagreement 0.6 The host pushing back 0.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 29:40 Rejecting LLM Creator Liability
Marty strongly dismisses proposed regulatory frameworks that hold LLM creators responsible for all downstream outputs, comparing it to holding Microsoft liable for crimes committed on Windows computers.
Hardest push from the host ▶ 16:15 Probing Historical Regulatory ImpactThe host presses Marty to unpack the structural drivers of market evolution, questioning whether electronification was fundamentally forced by regulatory policy or enabled by technology.
Biggest teaching moment ▶ 8:55 Mathematical Reality of Medical AIMarty educates the host on the exponential combinatorics of internal medicine diagnosis, explaining how a search space of 1000 to the 10,000 power led directly to an early AI winter.
The host holds their own ▶ 27:33 Host Outlines Enterprise AI DeploymentWhen Marty asks David for his perspective, David demonstrates strong domain knowledge by enumerating concrete back-office workflows like compliance, vendor onboarding, and risk management.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Legal Disclaimer | 2 | 3 | 0 | 0 | The host sets up the podcast with an open invitation for Marty to explain his career arc. Marty shares a personal anecdote about his father introducing him to computers in 1974 and his early work with Monte Carlo simulations. | |
| Computational Biology and the Concept of Digital Twins | 1 | 4 | 0 | 0 | Marty describes his time at Harvard working on computational biology and introduces the core concept of digital twins. The host listens attentively as Marty connects these scientific simulation methods to financial modeling at Goldman Sachs. | |
| AI Research at Stanford and Transition to Wall Street | 1 | 5 | 0 | 0 | Marty details his PhD research at Stanford, explaining why general medical diagnosis posed an intractable mathematical problem that led to an AI winter. He recounts how Goldman Sachs recruited him to build SecDB's core architecture. | |
| How SecDB Navigated the 2008 Financial Crisis | 3 | 5 | 1 | 0 | The host frames a thoughtful question about SecDB's role during the 2008 financial crisis. Marty gently reframes public misconceptions, explaining that SecDB simply provided high-fidelity present-state data across 47 Lehman entities rather than predicting the future. | |
| Impact of Regulation on Technology in Finance | 3 | 6 | 1 | 0 | The host asks whether electronic trading was driven by regulations or technological innovation. Marty provides an educational breakdown of Dodd-Frank stress tests and uses a railroad junction metaphor to explain how AI should be regulated at boundary interfaces. | |
| The Evolution of AI and Production Deployment | 2 | 6 | 1 | 0 | The host asks Marty to compare modern generative AI to his early Stanford research. Marty contrasts early Bayesian models with connectionist neural networks and explains how transformers rely on stationary data distributions. | |
| Data Foundations and the Single Source of Truth | 4 | 5 | 0 | 0 | Marty emphasizes the importance of establishing a single source of truth for enterprise data and highlights Gemini's long context windows. When Marty turns the question back to David, the host fluently lists key enterprise workflows that software will automate. | |
| Policy Frameworks and Regulatory Boundaries for AI | 2 | 6 | 2 | 0 | The host inquires about policy frameworks for AI governance. Marty forcefully argues against holding LLM creators liable for downstream user actions, making an analogy to holding Microsoft accountable for crimes committed using Windows. | |
| Transformative Impact of AI in Biotech and Drug Discovery | 2 | 6 | 0 | 0 | The host invites Marty to discuss AI's impact on life sciences and drug discovery. Marty explains how chip simulation parallels biology and illustrates the massive scale of molecular search spaces. |
Statements from this episode (14)
Chavez ran nuclear simulations on a Cray-1 supercomputer at age 16
“My very first summer job when I was 16 was at the Air Force weapons lab. In Albuquerque, the government had decided that blowing up bombs in the Nevada desert was really problematic in a lot of ways. And some scientists had this idea crazy at the time that we …”
Chavez: Digital twins enable safe counterfactual testing for complex systems
“And the amazing thing about a digital twin is you can You can do all kinds of experiments, and you can ask all kinds of questions that would be dangerous or impossible to ask or perform in reality and then you can change your actions based on the answers to th…”
Medical diagnosis requires calculating a massive joint probability distribution
“The actual problem of diagnosis in general internal medicine Is you've got about a thousand disease categories and about 10,000 various clinical findings laboratory findings or manifestations or symptoms. And the joint probability distribution that you have to…”
Goldman Sachs didn't predict the 2008 crisis, they just hedged their risk
“We didn't know that the financial crisis was coming. Of course, you know, we got in the press and elsewhere accused of all kinds of crazy things. Like they were the only ones who hedged so they must have known it was coming. We were just predictors of the pres…”
Goldman delivered 47 Lehman closeout claims within an hour of bankruptcy
“We had our courier show up, At Lehman's headquarters within an hour of its filing bankruptcy protection for the 47 entities, and we had 47 sheets of paper with our closeout claim against each of those entities rolled up from wide across all the businesses, and…”
Chavez: Post-crisis stress testing made the banking system significantly safer
“So this, that caused a massive change and made the system massively safer and sounder. We saw that in the pandemic.”
AI regulation must focus on external API boundaries, not internal model thoughts
“That's going to be an important principle for LLMs and AIs generally as they start agenting and causing change in the world. We have to care a lot about those boundaries.”
Generative AI struggles with non-stationary distributions like financial markets
“It all depends on the training set, and it also depends crucially on on a stationary distribution, right? So, you know, the reason all this works on is it a cat or not a cat Is that cats change very slowly in evolutionary time. They don't change from day to da…”
Chavez: Most enterprise companies have done a terrible job with data engineering
“Getting your single source of truth right, that data engineering problem, I think a lot of companies have done a terrible job of it.”
Million-token context windows will make hard RAG problems easy within months
“And I think over the next few months, you're going to see a lot of those changes, problems that were really hard are going to become really easy.”
Holding LLM creators liable is like holding Microsoft liable for Windows misuse
“Let's make the LLM creators liable for every bad thing that happens with an LLM. To me, that is the exact equivalent of saying, let's make Microsoft Liable for every bad thing that someone does on a Windows computer, right? They're fully general, and so these …”
99% of Dodd-Frank was unnecessary red tape, except for CCAR rules
“That was the history of Dodd-Frank. Like, we don't really know what went wrong in the financial crisis, so let's just go regulate everything, and I think 99% of it was red tape that did not make the world a better place, and some of it Such as the CCAR regulat…”