I continue to believe that the essential challenge in AI is trust, not just capability.

Obviously if you can give an AI system lots of data about your life, it can learn ways to automate and improve it. But do we trust AI companies with that data? Not really, not now anyway.

Ultimately there may be a technical solution to this challenge, kind of like how E2E encryption addresses many trust concerns around messaging.

But if companies pursue capability without addressing trust, they might find a harder time in the marketplace than they expect.