[AINews] Jeff, Sanjay, Oriol, and Quoc depart DeepMind; Demis to Chair; Koray to SVP — what is going on at GDM???

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

It’s tempting to give Meta Spark 1.2 and Muse Code the title story today because of their success launching a 5.6 Terra-level model, together with innovative harness design with a local event log for resumability and persistent background agents, both of which should put other coding agent builders on notice.It’s tempting to highlight Prime Agent, Prime Intellect’s self-improving RLM based harness that claims an incredible 95.5% on ARC-AGI-3 (not yet endorsed by ARC).We tried. Trust me, we tried.But it’s hard to beat around the bush — today’s most important story is the coordinated departures of Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le, some of the most senior engineering and research talent in both Google and in human history, from DeepMind to cofound a new autoresearch startup Discovery Loop:And Demis, who needs no introduction, goes from CEO to Chair and Chief Scientist, while “leaning into” Isomorphic with CTO Koray stepping up to SVP of GDM:All the departures are very amicable; Google is investing in Discovery Loop, and Demis’ increased contributions to long term strategy and Isomorphic in particular will be very welcome by humanity, but surely we are not being told the full story here; why couldn’t Discovery Loop have been done inside Google? That one at least, we have some hints, given GDM’s history of 1000+ coauthor papers for Gemini, vs these 4 superhumans writing this manifesto:When John Jumper left for Anthropic, one could maybe chalk it up to Anthropic’s momentum. When Noam Shazeer joined OpenAI, perhaps one could point to their pioneering work in …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://www.latent.space/p/ainews-jeff-sanjay-oriol-and-quoc
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.