About this episode

Neel Nanda reviews Anthropic's global workspace paper, highlighting its insights on cognitive space in models.

The below is a public review Anthropic asked me to write for their new global workspace paper. I recommend at least skimming their paper first. TLDR: I think this is a fantastic paper - it presents compelling evidence for some kind of "cognitive space" in models, that is used as a "working memory" for intermediate variables during a forward pass, shows that J-Lens is a useful technique for accessing this space. I believe these key claims.I believe J-Lens will be a useful (but limited) to...

People in this episode

Guest: Neel Nanda

Topics covered

• cognitive science
• artificial intelligence
• working memory
• research review
• cognitive space

Keywords

• Anthropic
• global workspace
• cognitive space
• working memory
• J-Lens

Mentioned in this episode

Organizations: Anthropic

Products: J-Lens

Books & works: global workspace paper

More episodes of LessWrong (Curated & Popular)

• "Recap of bike trip/street interviews across America" by cguth7 · July 20, 2026 · 10 min
• "I don’t think Claude is misaligned in ‘Agentic Misalignment Summer 2026 - Motivated Mislabeling’" by JohnWittle · July 17, 2026 · 26 min
• "Why I Left Google DeepMind" by TurnTrout · July 15, 2026 · 1h 15m
• "The mosquito bucket of doom works" by dominicq · July 15, 2026 · 10 min
• "Our response to Séb Krier on Plan A" by MKodama, Thomas Larsen · July 14, 2026 · 31 min
• "The Whitney Biennial Should Admit That Emilie Gossiaux Wants to Fuck Their Dog" by jenn · July 14, 2026 · 17 min

Explore listener stats, chart rankings, contacts and more on the LessWrong (Curated & Popular) podcast page.