AI Safety Poland Reading Club #11

​Who should come: Anyone interested in AI safety, machine learning research, or the broader societal impacts of AI systems.

What it's about: We'll read and discuss influential papers in AI safety research. Read the paper beforehand, show up, and talk through the ideas, questions, and disagreements it raises.

This time, we will start another new paper from Anthropic: Verbalizable Representations Form a Global Workspace in Language Models. You might also know it as the paper that introduces the J-Space concept.

​For the next meeting, please read:

• ​The accessible summary of the paper: https://www.anthropic.com/research/global-workspace

• ​Sections 1 and 2 from the paper itself: https://transformer-circuits.pub/2026/workspace/index.html

​If you are in a hurry, you can skip Section 2 from the paper for now, as we might not get there that fast. You might also want to play with the demo.