- Miles Brundage, formerly of OpenAI, lays out four steps AI companies could take now toward slowing development: independent third-party audits, safety coordination with rivals, funding verification technology, and supporting oversight legislation (Alana)
- Verification tools, such as tracking the chips used to train advanced models, would make a pause checkable rather than trust-based, and answer worries that China would cheat (Alana)
- India's IT sector is being reshaped by automation, with fewer entry-level engineering roles, while Brazil puts over $400 million into AI infrastructure split between U.S. and Chinese firms, and Europe subsidizes domestic AI at a far smaller scale (Joe)
- Europe's regulatory record is criticized, but neutral parties have historically brokered agreements between rivals, and European research bodies could fund verification work (Joe)
- Pew Research Center ran 490,000 webpages through an AI detection model and found about 10% of all pages show signs of AI authorship, rising to 35% among pages published after ChatGPT's release, with caveats about over-flagging and unclear publication dates (Donald)
<- Back to digests
Daily digest
A shared responsibility
By Alana Horowitz Friedman, Joe Rogero, and Donald Gauvreau
Advice for AI companies, the world outside the race, and a deluge of AI-written webpages
Read the full dispatch on AI StopWatch
From AI StopWatch, published with their permission.
The summaries and the full Spanish translation are produced automatically. The original is always linked.