People around me often say I do not let go of a thought until I am thoroughly satisfied. Maybe this why I spent the weekend still thinking about AI and consciousness.
Much of the conversation around AI focuses on intelligence. We debate whether AI will outperform humans, automate more work or transform industries. Those are important questions, but I increasingly think there’s another question underneath all of them.
If consciousness depends on biological matter, then no matter how intelligent AI becomes, it will never be conscious. But if consciousness depends on architecture, then biology may not be the decisive factor at all.
Current AI systems already satisfy parts of the model I’ve been thinking about. They can retain information during an interaction, reason across that information and modify future responses based on previous context. What they largely lack is continuous long-term identity, persistent autobiographical memory and genuinely intrinsic goals that exist independently of user prompts.
Those differences are significant, but they seem more like engineering gaps than conceptual ones. That doesn’t mean today’s AI is conscious, and I don’t think we should casually make that claim. It does mean that if our definition of consciousness is architectural, then we should be clear about which pieces are still missing instead of treating the entire question as unknowable.
Whether this model turns out to be right or wrong, AI is forcing us to become much more precise about concepts we’ve historically been comfortable leaving vague. In trying to understand machines, we may end up understanding ourselves much better.
That line of thinking eventually led me away from AI altogether and back to the human mind. Specifically, I’ve been wondering why our own emotional system works the way it does, but that’s a discussion for another post.
