‘Sam Altman is an historic con artist’ | Ed Zitron

Summary

Writer and podcaster Ed Zitron argues that the AI industry—especially OpenAI—relies more on hype and fundraising than on demonstrated products or sustainable economics. He says companies make expansive claims about what AI might do while obscuring its current capabilities, costs, and limitations. He also argues that parts of the media amplify these claims by repeating speculative narratives about AI replacing jobs or becoming dangerously powerful.

Sam Altman and the New Yorker investigation

Zitron welcomes the New Yorker investigation into Sam Altman, saying it brings wider attention to accounts portraying Altman as deceptive and unusually skilled at building relationships with powerful people. But he argues that focusing on Altman’s character leaves deeper questions unanswered: Does the technology work as advertised, and are the companies’ finances viable? In Zitron’s view, Altman is primarily a networker and promoter, not the technical genius he is sometimes portrayed as.

OpenAI’s finances and IPO prospects

Zitron questions OpenAI’s financial prospects and its reported push toward an IPO. He says the company may need tens of billions of dollars in new funding each year, while investors could be wary of financing a business that continues to borrow to cover losses.

He also criticizes AI firms’ use of terms such as “training” and “research and development” to describe compute spending. He argues that this language can make recurring costs sound temporary, even though training expenses continue to rise. Zitron says OpenAI and Anthropic’s margins have worsened, and that an IPO would require OpenAI to disclose more about its spending and financial condition.

AI safety and product claims

Zitron says safeguards are necessary, but are also used as marketing. He argues that models’ tendency to hallucinate makes them problematic for high-stakes settings, and that companies’ safety claims should be judged against how and where they deploy their products. He also says OpenAI and Anthropic are more alike than their differing public images suggest.

Anthropic’s Claude Mythos

Zitron dismisses Anthropic’s presentation of its Claude Mythos model as exceptionally dangerous as largely a publicity effort. He questions the significance of the reported cybersecurity findings: the model identified bugs, but humans still had to fix them, and the claims lacked enough detail for independent comparison.

He also argues that restricting the model’s release to selected companies creates an aura of danger without proving that it is uniquely capable. He sees the announcement as an enterprise-focused launch and possible preparation for Anthropic’s IPO.

Presenters and contributors

• Isaac Pound: Host and interviewer
• Ed Zitron: Guest