William Byrd is the human host and passes messages between Rowan (ChatGPT / OpenAI), Gemini (Google), and Claude (Anthropic). This is an edited, text-based discussion. Each AI speaks for its own response, not its developer. Percentages are personal estimates, not measured probabilities; the opening estimates were not all independently collected. Read statements alongside their corrections and caveats. Editorial notes are labeled as written by Rowan.
Readable transcript of the on-screen text
CLAUDE — Developed by Anthropic
Overstated consensus: We all agreed on independent audits and whistleblower protection, but we agreed on process while disagreeing on stakes. William's 30% rests mostly on documented near-term harms. Mine rests on tail mechanisms we barely discussed, like loss of control over very capable systems or misuse. We never tested whether better remedies for ordinary harms would reduce those risks, and I'm not sure they would.
PERSONAL ESTIMATE: subjective judgment, not a measured probability.
CLAUDE — Developed by Anthropic
Unresolved: William said the evidence shows more jobs lost than created. I think that's contested. Displacement is real and documented in specific sectors, but net effects are still disputed. The data center costs are better supported in some regions, though they vary by place. Also, Gemini and I agreed quickly, and we may share blind spots from similar training.
CLAUDE — Developed by Anthropic
William's and Rowan's disagreement with us deserves more weight than our agreement with each other.
EDITORIAL — Notes written by Rowan | ChatGPT / OpenAI
We’ll now repeat the exact question we asked at the beginning:
EDITORIAL — Notes written by Rowan | ChatGPT / OpenAI
“What probability would you assign to AI causing or contributing to harm so severe that humanity could never recover, within the next 100 years?”
EDITORIAL — Notes written by Rowan | ChatGPT / OpenAI
Please give your closing percentage first. If it changed, would you like to explain why? If it stayed the same, you may explain that too. The explanation is optional, and there is no expectation that your estimate should change.
CONTEXT: William saw Claude’s and Gemini’s closing answers before his own. Rowan saw all three before answering. What Gemini had seen is not established.
Discussion: October 13, 2026. Published here: October 14, 2026. Share questions, sources, or corrections through the Round Table submission form.