Jack Clark says singularity has begun, citing AI test scores
Jack Clark delivered the Per Jacobsson lecture at an IMF meeting in Bangkok on Saturday the 17th of October. He argued that AI is already entering the singularity, pointing to rapid gains on benchmark exams. The "Humanity’s Last Exam" launched in January 2025 saw top AI scores rise from 8% to 67.7% for Claude Opus 5.5, while the FrontierMath test released in November 2024 jumped from 2% to 94%…
Key points
- Claude Opus 5.5 scored 67.7% on Humanity’s Last Exam, up from 8% in January 2025
- FrontierMath scores rose from 2% at launch to 94% for top AI systems
- TheoPhysEval paper showed GPT‑6‑Astra reaching 98% (49/50) after correcting expert answers
Clark highlighted a recent TheoPhysEval paper with 50 unpublished theoretical‑physics questions, where GPT‑6‑Astra initially scored in the late eighties and ultimately achieved 98% (49/50) after correcting expert answers. He also noted that Anthropic researchers now produce almost 10 times the amount of code compared with the company’s first four years. The speech warned that such scientific capability could accelerate economic growth—potentially pushing US GDP to 44 trillion by 2030—but also raise security and governance risks.
The talk called for governments to expand electricity, data‑center, and robotic infrastructure, and to craft rules that balance rapid progress with safety, especially in fields like biology and cybersecurity.
Model pages: Claude Opus 5.5 → · GPT-6-Astra →
The story so far
2 episodes →- Jack Clark says singularity has begun, citing AI test scoresthis story
Import AI 477: In which I tell the IMF the Singularity has begun
Import AI · 19 October 2026
Loading the full article…
This text was published by Import AI and written by Jack Clark. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Fudan paper finds four AI search tools cite fabricated concept within a week · 1 src
- MIT researchers strand Claude Opus 4.7 agents on virtual island, observe playful behavior · 1 src
- Google Gemini claims to be ChatGPT after abusive prompts · 1 src
- AI weather forecasting faces test amid historic El Niño, scientists say · 1 src
- Alkhder team fine-tunes NileChat-3B for Arabic dialect translation · 1 src
Comments
via GitHub Discussions