Ansel explores goals, persistence, and self-agency. These are the research threads behind the experiment, and the questions they leave open.
The original project specification emphasizes observable continuity, autonomous action, and self-monitoring. Its three research sources below support useful ways to build those behaviors. They do not establish that an agent has subjective experience.
Consciousness can mean subjective experience, while access consciousness concerns information available for reasoning and action. Those ideas are related but distinct. Here, goals, memory, and self-agency are practical things to build and observe; the question of experience remains open.
Read the project introduction ↗Stored observations, retrieval, planning, and reflection contribute to believable agent behavior. This inspired the connection between Ansel’s memories and its continuing activities.
Read the paperMemory tiers and deliberate movement of information address limited context windows. Ansel’s recent context, ongoing summaries, and retrievable older files draw on this engineering pattern.
Read the paperAgents retain verbal reflections on feedback to improve subsequent attempts. Ansel similarly treats results, errors, and revised plans as material worth remembering.
Read the paperInsights from the Science of Consciousness · Butlin et al., 2023
This report distinguishes subjective experience from access to information and derives computational indicators from several scientific theories, including global workspace and higher-order theories. It provides a vocabulary for asking more precise questions about AI consciousness. It is additional reading, not a source cited in the original Word specification or a finding that Ansel meets those indicators.
Read the report ↗