OpenAI's Head of Financial Services on Building ChatGPT for Finance
Latent Space · youtube · 2026-10-31
Latent Space interviews Joseph Kim, who leads ChatGPT for Financial Services at OpenAI, on building vertical products atop a general-purpose model. Key points:
- Financial reasoning in the model: OpenAI trained financial reasoning into a shared model rather than relying solely on tooling.
- Why bundle data: despite existing market-data connectors, OpenAI bundled financial data directly to lower friction and ensure quality.
- Startup opportunities: startups can build on the same plugin primitives for vertical scenarios instead of competing head-on.
- Taste is hard to measure: expert judgment resists accuracy metrics; OpenAI explores human taste as a training signal, even using code quality as a taste benchmark.
- Safety and review chains: banking workflows require human review and compliance, designed in from the start.
- Evals remain hard: financial evaluations are an open problem; Codex adoption inside OpenAI also came up.
The finance team at OpenAI is hiring.
More from Companies & People
- Vapi CEO: 'Promise something impossible, then deliver it' — voice agent insights — schwentker · 2026-11-01
- Modal x Gray Area debut Scaling Intelligence show with 13 artists exploring AI as a creative medium — charles_irl · 2026-11-01
- 10 people, $3.1M monthly tokens, 77k changes: fact-checking the Spawn AI startup story — sujingshen · 2026-11-01
- Company mandates AI agent adoption but can't explain why to paying customers — KaleChipGurl1 · 2026-11-01
- Adobe Foundation Model Team Recruiting 2027 Research Interns: Wants Creative Hackers — yining_hong · 2026-11-01
- davidsenra teases AppLovin interview with Rafael Vivas, out tomorrow — davidsenra · 2026-11-01