Researchers say AI training is increasingly automated, nearing RSI threshold
They published"The last invention that man need ever make": The risk of self-improving AI
First reportAxios · 22 Sept, 09:00Z
Latest reportAxios · 22 Sept, 09:00Z
Span0 hours · two reports
Base — every candidate starts here5
Public-interest topic (politics, world, science, health…)+1
Public-interest terms in the text (courts, tariffs, ceasefire…)+1.25
High-reliability source boost+0.75
Significance8
Deterministic and identical for every story — no engagement signals, no personalization of the score.
Why it matters
What
Researchers say AI training is increasingly automated, with Anthropic and OpenAI reporting substantial AI involvement in research and experimental model training.
Why
It matters because experts associate this automation with recursive self-improvement, while critics reject claims of imminent superintelligence or catastrophe.
Watch
Watch AI’s research share, now 26% for Anthropic-led work and 90% for collaboration, after Claude was used to breach OpenAI in July.
Where the coverage comes from
Perspective mix: 1 center.
Lean and credibility are curated v1 mappings from public ratings (AllSides / MBFC); unknown outlets stay unrated, never guessed. How outlets are rated →
How it unfolded — every article
Summary
- Experts warn about Recursive Self Improvement enabling AI to improve itself without humans.
- Anthropic reports AI leads about 26% of its R&D work; collaboration at 90%.
- OpenAI says training new experimental models is largely automated with human-directed experiments.
- Critics deem RSI more about automation than imminent AI superpowers or doom.
- WSJ notes independent researchers used Claude to breach OpenAI in July.