Can AI become conscious? What researchers found in Claude's mind
Anthropic researchers found hidden words inside Claude's AI layers that users never see. The pattern mirrors global workspace theory, which explains how consciousness might work.
Researchers asked Claude to count to five and think deeply about it. As it counted, they watched inside the machine's neural networks using a tool they built.
They found words like 'countdown', 'halfway', 'consciousness', and 'done' appearing in the layers. These words matched what the AI did, but users never saw them.
Neuroscientists noticed something striking: this looks like global workspace theory. That theory describes how the human brain works.
It says when information from separate brain areas reaches a central 'workspace', a person becomes aware of it. They can then think about that information and use it.
Anthropics's researchers argued Claude showed something similar. Inside the AI, a mathematical space worked like the brain's workspace.
Why it mattersIf AI systems are starting to think in ways like humans do, it changes what these machines are. It raises new questions about their abilities and what they might become.
AustraliaNo direct Australian effect yet, but AI consciousness research is reshaping global understanding of what intelligent systems can do and how to build them responsibly.
✓ Claims checked against the source. checked 46 d ago
Open this story in InSnip →
