What Is Claude Thinking But Not Saying? with Jack Lindsey

Is AI thinking things it never says aloud? How would we know if a large language model were having a subjective experience? Is there a way to read the ‘thoughts’ of an LLM? Human brains seem to have a ‘global workspace’ — what would it mean if we discovered that AI has something similar? Can you reach inside an LLM to change one of its ‘thoughts’? Join Eagleman with Anthropic interpretability researcher Jack Lindsey to talk about what might be going on inside the mind of Claude.

Episode Audio

Episode Video

Coming Soon

More information

A global workspace in language models (includes Anthropic’s lay-explanation as well as a link to the J-space paper “Verbalizable Representations Form a Global Workspace in Language Models”)

Baars BJ. Global workspace theory of consciousness: toward a cognitive neuroscience of human experience. Progress in brain research. 2005 Jan 1;150:45-53.

Baars BJ, Geld N, Kozma R. Global workspace theory (GWT) and prefrontal cortex: recent developments. Frontiers in psychology. 2021 Nov 10;12:749868.

Mashour GA, Roelfsema P, Changeux JP, Dehaene S. Conscious processing and the global neuronal workspace hypothesis. Neuron. 2020 Mar 4;105(5):776-98.

SUBSCRIBE AND LISTEN ON:

"A popularizer of impressive gusto...[Eagleman] aims, grandly, to do for the study of the mind what Copernicus did for the study of the stars."
- New York Observer

"Eagleman has a talent for testing the untestable, for taking seemingly sophomoric notions and using them to nail down the slippery stuff of consciousness."
- The New Yorker