Glossary · Term

global workspace

← all terms

Definition

Plain language

A small shared space in a mind where a few thoughts get broadcast to the rest of the system, while most processing happens quietly on the side.

As stated in the literature

A theory from neuroscience holding that conscious access works by posting a limited set of representations to a broadcast channel available to many subsystems; the paper argues language models grow a functional analog that carries deliberate reasoning.

Also called: global workspace theory, workspace theory, workspace

Why it matters: It offers a way to explain how a system with many separate parts can still act on one shared idea at a time, rather than pulling in every direction at once.

For example, while reading a book your mind quietly handles breathing, posture, and background noise, but the one sentence you're focused on gets shared across your attention, memory, and speech.

Heard on the show

“Canonical system prompt, workspace files, and a nine-digit Social Security Number sitting on line nine of a user memory file.”
Episode 246 — 160 Perfect Refusals, And The Refusals Were The Leak

Mentioned in 20 episodes

  1. 246
    160 Perfect Refusals, And The Refusals Were The Leak
  2. 237
    The Model Built a Perfect Map of the Puzzle, Then Lost It
  3. 236
    Why a Printed 'OPERATOR OVERRIDE' Note Redirects Robot Planners
  4. 203
    The Thought a Model Doesn't Say — and the Lens That Reads It
  5. 190
    The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys
  6. 184
    An AI Built an Undetectable Secret Channel, And Another AI Couldn't Find It
  7. 157
    When an AI Coding Agent Drives a Phone Through the Terminal, No Screen Needed
  8. 130
    Why AI Agents Coordinate Better Through a Shared Board Than a Boss
  9. 125
    AI Coding Agents Run a Marathon, and Fewer Than One in Three Finish
  10. 113
    What If a Prompt Injection Never Left? Attacks That Wait in Agent Memory
  11. 105
    The Trojan Is Your Agent's Memory: Why Single-Step Defenses Miss Persistent Attacks
  12. 086
    Why Frozen-Weight Agents Still Get Worse Over Time
  13. 076
    Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
  14. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
  15. 066
    Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
  16. 046
    When the AI Optimizer Edits the Grade Book: Why Harnessing Evolution Needs a Wall
  17. 043
    When 'This Is False' Doesn't Stick: Why Models Learn the Lie Anyway
  18. 029
    Why Forty-Eight Percent on FrontierMath Isn't the Real Story in DeepMind's New Math Paper
  19. 004
    The Sycophancy Circuit That Survives Alignment Training
  20. 003
    How to Pick the Best of Sixteen Coding Agent Rollouts

Related concepts