AI in Finance: Can AI Bankers Outperform Humans? | Richie Torres Questions Expert

The Transformative Power of Generative AI

• 💡 Generative AI is highlighted as a potentially revolutionary technology, comparable to the advent of writing or the printing press, poised to radically alter the world.
• 🚀 The discussion focuses on the new capabilities of generative AI and large language models (LLMs) in finance, distinguishing them from legacy AI.

AI Applications and ROI in Finance

• 🎯 A key capability of LLMs is their general-purpose nature, allowing them to perform a wide array of workflows through different prompting methods.
• 📈 The most mature and high-ROI application identified is developer productivity, where AI agents build applications and empower even small banks to accelerate software development.
• 💰 Compliance is also noted as a high-ROI activity due to the significant manual effort it typically involves.

AI vs. Human Bankers: Common Sense and Bias

• 🧠 A major challenge in AI outperforming human bankers is instilling common sense into AI agents, which remains an open challenge despite vast training data.
• ⚠️ Building guardrails that mimic human training is crucial but difficult, though AI companies are assisting banks with this.
• ❓ The objectivity of AI is questioned, as its data is derived from the internet and human nature, reflecting existing biases.

AI's Impact on Credit and Inclusivity

• 📊 Initial hopes that AI would lead to more loan approvals and objective credit scoring have not always materialized, with AI sometimes yielding similar or worse results than human judgment.
• 🏠 In housing appraisals, AI can still produce biased outcomes by using proxies like address and neighborhood, even if personal artifacts are removed.
• 🎯 AI has the potential to expand access to capital and credit by detecting new patterns of creditworthiness beyond traditional scoring methods.
• ⚖️ Second-look applications of AI are suggested, where declined applications are re-evaluated by cutting-edge models to potentially identify approvals and mitigate bias.

Transparency and Algorithmic Bias

• 🗣️ Public disclosure that AI is making credit decisions is essential for transparency.
• 🔍 A system is needed to evaluate the disproportionate impact or disparate impact of AI models.
• ✅ The achievable mission is not to make AI completely bias-free, but to make it less biased than the human alternative.