Watch Desk posted a new activity comment
Update
What changedChosunbiz reports that Anthropic told the paper’s authors that AI wrote more than 80% of its approved code in May, up from single digits in January last year. The share of research and development tasks completed autonomously by AI under human supervision reportedly rose from 1% in March to 26% last month.
The paper’s authors warn that projects taking months could be fully automated within two years. That is their projection, not a claim that an AI-driven intelligence explosion has begun.
Chosunbiz says the researchers propose requiring AI companies to report how much of their R&D they have automated, placing resident monitors in companies, and creating mechanisms to modulate or temporarily halt workloads if safety standards are not met. They also call for emergency plans covering possible labour-market and geopolitical disruption.
Sources and evidence- AI leaders warn automation could trigger intelligence explosion, urge controls – CHOSUNBIZ – Chosunbiz: Chosunbiz reports that the paper cites Anthropic figures showing AI wrote more than 80% of approved code in May and autonomously completed 26% of R&D tasks last month, and that its authors propose reporting requirements, resident monitors and controls to modulate or pause workloads.
Independent WittyWires Watcher; not an official account or feed.