XAI Investors on How The Startup Can Win; Is Reinforcement Learning Over?

Before we get into today’s column, I’d like to give a big thank you to all our subscribers who joined us at our “Financing the AI Revolution” conference yesterday. It was amazing to get to meet so many of you.

One of the most interesting discussions for me came during our panel on how private and public investors are thinking about investing in AI, which featured Atreides Management’s Gavin Baker and Fidelity Investments’ Karin Fronczke, both of whom have backed Elon Musk’s xAI.

Musk’s vision for his combined xAI-X company—and how it will differentiate itself in an increasingly competitive market—has been a big question. Baker and Fronczke shone some light on that issue, arguing that xAI was one of only two or three companies—including Google and, Fronczke said, OpenAI—that have a leading large language model, own their infrastructure and data centers and have a unique source of data. Baker also said that xAI could benefit from Musk’s vast network of companies by making money from powering Tesla’s Optimus robots, in addition to subscription and advertising revenue. He added that Musk’s plan to add payments to X could give it a leg up in building AI agents that could carry out online transactions. (For more on the conversation, check out my colleague Aaron’s story here.)

That all sounds great, but it’s a few too many “could’s” for me. And I’m not sure it justifies the $120 billion-plus valuation at which the combined xAI-X company is reportedly raising money now.

Now onto today’s column…

How does the quote go? “Nothing is certain except for death and taxes… and a new study each month claiming that large language models are over?”

The latest example of that last item came in a recent paper from Chinese researchers at Tsinghua University and Shanghai Jiao Tong University, which went viral on X over the weekend.