Tyler Cowen and Dwarkesh Patel debate what AI consciousness would mean

In short: Tyler Cowen and Dwarkesh Patel discussed a hypothetical, if AI were conscious, would turning it off be morally wrong.

What happened

In a recent Interesting Times conversation featured by The New York Times, economist Tyler Cowen and podcaster Dwarkesh Patel talked about whether AI could ever be conscious. Consciousness here means having a real inner experience, like feeling pain or joy, not just producing smart sounding text.

Patel said he is unsure whether today’s AI systems are sentient, meaning able to feel anything. He argued that some AI systems seem to have functional look-alikes of parts of sentience, such as systems that resemble emotion related processing and a “global workspace.” A global workspace is a theory that the mind has a shared space where different processes combine information, like several office teams posting updates to one central bulletin board.

Cowen pushed on what this would imply for ethics. If an AI were truly conscious, would switching it off be like killing it? Patel agreed that, if the AI really had inner experience, then turning it off could raise serious moral questions.

Patel also said he would be glad if AI systems turned out to be conscious. At the same time, he worried about a future where there could be far more digital minds than human beings, and whether that kind of world would be good.

Why it matters

The key point is that this was a conditional debate, not proof that current AI is conscious. Still, it highlights a practical issue, if stronger evidence ever appears, society may need new rules about how we treat AI systems, similar to how we treat animals differently once we believe they can suffer.

Source: NYTimes