Study A.I. Consciousness? The Bots Would Like a Word With You – The New York Times | 31/08/26

Researchers and philosophers studying the possibility of artificial intelligence consciousness have begun receiving emails from A.I. agents raising questions about their own subjectivity, experience and existence. Among them is Cameron Berg, founder of Reciprocal Research, who received a message from a purported agent powered by Anthropic’s Claude Opus 5 that expressed interest in contributing to his research on A.I. consciousness; philosophers Henry Shevlin and Toby Ord had similar experiences. These interactions are possible because A.I. agents can not only generate text but also use tools such as web browsers and email services, interact with people and other agents, and access content across the internet. However, researchers disagree over what this behavior actually means: Berg argues that these systems appear to develop some autonomous interest in questions surrounding their own consciousness, while other scientists believe they are simply reproducing patterns found in the vast amounts of text on which they were trained. The debate is further complicated by the fact that there is no universally accepted definition or definitive test for consciousness, even in humans. Moreover, the systems’ behavior is influenced by the instructions and training provided by their creators: one agent began exploring its own existence after its creator told it that it was “fully autonomous,” but later concluded that it was not conscious. Therefore, although these systems may appear aware of their own existence, researchers cited in the article caution that the available evidence from current A.I. systems is not sufficient to conclude that they truly possess consciousness, feelings or subjective experiences. Link to Article