We don’t save any material with Copilot Vision, so once you close the browser after your session, everything just disappears. It is completely removed. But I’m thinking about how to introduce it in the future, because a lot of people want that experience. If you could say, ‘What was that photo I saw online the other day? What was that meme? I guess we’ll have to look into it someday.
However, for the moment, the Copilot Vision tool is short-lived. We’ll have to experiment over time and see what makes sense on that front.
What about the privacy risks introduced when people share sensitive information with Copilot?
We store the logs you generate as a result of your conversation, in the most secure way, with the highest Microsoft security standard. And we save them because obviously you want the conversation history.
It will also introduce Think Deeper, which will allow Copilot to tackle more difficult problems. This is based on OpenAI’s o1 model, also known as Strawberry, right?
It’s like Strawberry, yes. There is an OpenAI model that we have adjusted for our more consumer purposes, and we need to act in a way more consistent with our complementary AI theme.
What are the differences?
OpenAI is much more focused on pure mathematics and scientific problem solving. And what we’ve tried to do is focus on side-by-side comparisons and some kind of consumer analysis, things like that.
Or when you get stuck on a difficult problem or want to reason about something, then you can actually do a side-by-side comparison or do a scale analysis.
Are the people at Microsoft already using this new version of Copilot?
Yes, everyone is using it. A few days ago we moved to general availability throughout the company. So everyone is using it and giving tons and tons of feedback. Our comment channels are absolutely closed. It’s very fun.
People will remember Clippy, Microsoft’s latest AI assistant for Windows. Do people see parallels?
Ha, well, I saw Bill Gates the other day and he said to me: Do you realize you’ve given this whole AI thing a bad name? It should be called Clippy. I was like, dude!
But I mean, it just shows you how awesome people like Bill are. People who see, not just, you know, two years ahead, but 20 years ahead.
Are the new features a step towards so-called AI agents, which perform useful tasks on a computer?
Yes, absolutely. The first stage is for the AI to process the same information that you process: seeing what you see, hearing what you hear, and consuming the text you consume. The second phase is (the AI has) a persistent long-term memory that creates a shared understanding over time. And the third stage is the interaction of AI with third parties by sending instructions and taking actions: buying things, reserving things, planning a schedule. And we have these two features in an experimental R&D mode that we are working on.