Researchers at Anthropic have identified a representational structure called 'J-space' in language models like Claude, which shows features similar to a global workspace theory of consciousness. The study suggests that while this does not prove consciousness, it may indicate access consciousness, where information is available for reporting and reasoning. The findings were discussed by experts, including Stanislas Dehaene and others.
The chart will appear after repeat observations. The current metric comes from the source.
8 source pointsReal observations only. History before source connection is not reconstructed.
This summary provides an overview of recent findings on J-space in language models and its implications for understanding consciousness theories.
Pion is an agent designed to autonomously manage any company.
4 days agoHugging Face BlogThe source material only provides a headline without any additional information or context.
3 days agoLessWrongThe text discusses collective epistemics through mathematical models, focusing on Condorcet's Jury Theorem and the aggregation of independent information. It explores how majority decisions improve with more independent participants, even with slight individual accuracy. The article also touches on collective intelligence, model selection, and antifragility, aiming to enhance community rationality.
4 hours agoShow HNThe author created a platform for engineers to share their AI workflows, focusing on the tools, agents, and methods they use for managing AI projects.
yesterdayGitHubThe official DeepSeek Harness repository describes an approach that packages components of the agent environment as plugins. The project is listed under the ai-agents topic on GitHub.
36 days agoGitHubAn open-source project for AI agents that interact with a browser. The repository connects this workflow with Python, Playwright and LLMs.
687 days ago