Steve Eisman on AI: LLM Improvements Slowing & Market Outlook

The Irrelevance of the Fed

• 🎯 The Federal Reserve's current actions are considered irrelevant unless they are aggressively cutting or raising rates.
• ⚠️ Tinkering around the edges of monetary policy has no significant impact on investment decisions.

The Foundational AI Debate

• 💡 The most important discussion right now is around Artificial Intelligence (AI), not whether it's a bubble.
• 🧠 The core argument, popularized by Gary Marcus, suggests that Large Language Models (LLMs) will see diminishing returns as they scale.
• 📉 This means the rate of improvement in LLMs will slow down, contrary to the expectations of many.

Shifting Perspectives on AI Scaling

• 🗣️ Initially a lone voice, Gary Marcus's argument that LLM scaling has limits is gaining traction.
• 🚀 The perceived improvement of ChatGPT-5 over GPT-4 was noted as being less significant than previous upgrades.
• 🧩 Ilya Sutskever, a founder of OpenAI, has echoed this sentiment, stating that scaling LLMs is a dead end and a return to basic research is needed.

Market Implications and Personal Stance

• 📉 If the argument about LLM scaling limitations proves true, companies like Microsoft could reduce chip purchases.
• ⚠️ This foundational argument is compared to the assumption that housing prices couldn't fall before the great financial crisis.
• 🔒 Despite the concerns, the speaker is not selling any AI stocks they own, as they have held them for years.
• 🧐 The speaker acknowledges that this argument makes them nervous and they pay attention to it daily.