Could an AI be conscious? Separating the evidence from the vibes

In 2022 a Google engineer went public with his belief that the company’s LaMDA chatbot was sentient. Google disagreed and he lost his job. The episode is a useful warning: language models are trained on human writing, so they’re very good at producing the sentences a conscious being would say. That’s exactly why their own reports count for so little.

Serious researchers approach it differently. A 2023 report by a group of nineteen scientists and philosophers took the leading theories of consciousness, turned them into checklists of “indicator properties,” and assessed AI systems against them. Their conclusion was that no current system was a strong candidate, and that nothing obvious rules out building one later. Both halves of that sentence matter.

This section sits on that line. We don’t claim today’s AI is conscious, because there’s no evidence for it. We also don’t claim it never could be, because nobody knows what the right test is. What we can do is explain the theories, read the studies, and point out when a viral transcript is doing the work an argument should do.