9030 club reading "A global workspace in language models": #64 - ML Paper Reading Group
☕️📝 Paper Link 📝☕️
A global workspace in language models
Full Paper: https://transformer-circuits.pub/2026/workspace/index.html
Blog Post: https://www.anthropic.com/research/global-workspace
Do LLMs have something like consciously accessible thought?
In a recent paper Anthropic argues that Claude has developed a small set of internal neural patterns that play a special role: unlike most hidden processing, they appear available for reporting, control, and deliberate reasoning.
The analogy is to the brain’s distinction between unconscious processing and the thoughts we can consciously access.
Most neural activity in both brains and models is hidden machinery. The claim here is that some representations may sit in a privileged internal workspace—something the model can inspect, manipulate, and use across different tasks. If that holds up, it gives us a new way to think about both interpretability and model cognition.
Event Schedule;
7pm to 8pm --> quiet reading time, grab and snack and read! (optional)
8pm to 9pm --> open discussion about reading 📝
9pm --> we have our space for a bit longer, stay to socialize or network!
Our event is hosted within Mox SF, the gracious donors of the space we will meet.