[1709.08568] The Consciousness Prior

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. arXiv Operational Status Get status notifications via email or slack

[1709.08568] The Consciousness Prior --> Computer Science > Machine Learning arXiv:1709.08568 (cs) [Submitted on 25 Sep 2017 ( v1 ), last revised 2 Dec 2019 (this version, v2)] Title: The Consciousness Prior Authors: Yoshua Bengio View a PDF of the paper titled The Consciousness Prior, by Yoshua Bengio View PDF Abstract: A new prior is proposed for learning representations of high-level concepts of the kind we manipulate with language. This prior can be combined with other priors in order to help disentangling abstract factors from each other. It is inspired by cognitive neuroscience theories

saved by

• David Huang

related reading

• Verbalizable Representations Form a Global Workspace in Language Modelstransformer-circuits.pub
• Cognitive Parsimony Theory — Lumen Labslumenresearch.co
• A global workspace in language models \ Anthropicanthropic.com
• Digital_Consciousness_Model.pdfrethinkpriorities.org
• Interpretability Dreamstransformer-circuits.pub
• External commentary for global workspace paper -- final finalwww-cdn.anthropic.com
• Book Summary: Consciousness and the Brain — LessWronglesswrong.com
• The Platonic Representation Hypothesisphillipi.github.io
• It’s Bayes All The Way Upslatestarcodex.com
• [2602.11399] Can We Really Learn One Representation to Optimize All Rewards?arxiv.org
• No Space Like J-Space - by Zvi Mowshowitzthezvi.substack.com
• On the Origins of Linear Representations in Large Language Modelsarxiv.org