Are people at Microsoft already using this new version of Copilot?
Yes, everyone uses it. We just rolled out general availability company-wide a few days ago. So everyone is using it giving tons and tons of reviews. Our feedback channels are absolutely jammed. It’s a lot of fun.
People will remember Clippy, Microsoft’s latest AI assistant for Windows. Do people there see parallels?
Ha, well, I saw Bill Gates the other day and he said, do you realize you’ve misnamed this whole AI thing? It should be called Clippy. I was like, dude!
But I mean, it just goes to show you how mind-blowing people like Bill are. People who see not just two years ahead, but 20 years ahead.
Are the new features a step toward so-called AI agents that perform useful tasks on a computer?
Yes, absolutely. The first stage is AI that processes the same information you process – seeing what you see, hearing what you hear, consuming the text you consume. The second phase is [AI having] long-term, persistent memory that creates shared understanding over time. And the third stage is AI that interacts with third parties by sending instructions and taking actions—buying things, booking things, planning a schedule. And we have these two features in experimental R&D mode that we are working on.
Wait, you have an AI agent for Windows that can go and buy things for you?
It’s a long shot, but yes, we’ve closed the loop, we’ve made transactions. The problem with this technology is that you can get it to work 50, 60 percent of the time, but getting it to 90 percent reliability is a lot of effort. I’ve seen some stunning demos where it can trigger on its own and make a purchase etc. But I’ve also seen car crash moments where he doesn’t know what he’s doing.
Tell me more about a car accident. Did he blow out and buy a Lamborghini with Bill’s credit card?
If he used Bill’s credit card then I think it would be pretty funny. But no, like I said, we’re still figuring it out step by step. It is still deep in the decline of the labs. There’s a long way to go with these things, but you can count it in quarters, not years, I’d say.
What will be your biggest challenge in making the AI future you describe a reality?
The big thing here is figuring out how to create technology that you trust because it will feel like a very intimate and personal experience. We have to get the security part right, we have to get the privacy part right, of course. But I think the real thing is to try to design the conversation so that the agent can articulate boundaries so that they can say that this is not something I’m willing to participate in.
If we can achieve that, that’s the basis of a reliable experience, and then I think we can really get into the complex side of things, which is how to allow him to buy things for you on your behalf, or negotiate on your behalf, or contract on your behalf, or plan a schedule for you that includes three or four different stops within one Saturday afternoon. And you say, I believe you, co-pilot, did you get that? That’s really what we’re working towards.