Ethan LockshinSales at NVIDIA focused on AI adoption

Quant models ruled for 50+ years. Now LLMs enter —Rogo's President explains.

Name: Quant models ruled for 50+ years. Now LLMs enter —Rogo's President explains.
Uploaded: 2025-11-10T14:34:22+00:00
Duration: 1 min 29 s
Channel: Ethan Lockshin
Description: Rogo's president explains why LLMs beat quant models that dominated for decades. LLMs handle messy real-world data like corporate filings and press…

10 Nov 2025 · 1 min

Watch on YouTube

1:29

Rogo's president explains why LLMs beat quant models that dominated for decades. LLMs handle messy real-world data like corporate filings and press releases, while reinforcement learning teaches them the kind of judgment that top deal makers use. Traditional models couldn't do either.

Chapters

What do you think large language models enable that traditional quantitative models never could? It's a great question, so I put it into two categories. Where LLMs really thrive is in wider sets of unstructured data across text from filings, for example, or research reports. Hard structured data like tables, for example, within filings or within press releases, transactions data from databases, et cetera. They can traverse across all of these and synthesize insights or summaries or patterns in a way that's coherent and flexible. So to me, that's, that's one big pillar. The other pillar is coming back to our topic about reinforcement learning. If you show an LLM what good looks like in the right reinforcement learning regime, it will naturally create connections and distill learning that is very difficult to instill programmatically in the context of traditional machine learning techniques. And so if we, we stop and think about what makes a great an A banker, for example, there's so many different rules, contingencies, heuristics, that one could spend a lifetime writing those down and still not really capture the nuances of judgment. But the power of LLMs and agents is that with the right reinforcement learning, in the same way that you can train AI to beat the world's greatest go players, for example, you can train them to think like the world's best deal makers.