We don’t save any footage in Copilot Vision, so when you close your browser after a session, it all just disappears. Completely removes. But I’m wondering if and how to implement this in the future because a lot of people want this experience. If you could just say, “What is that photo I saw on the Internet the other day?” What meme was that? I think we’ll have to look at it one day.
However, Copilot Vision is ephemeral at this time. Over time we will have to experiment and see what makes sense in this regard.
What about privacy risks that arise when people otherwise share sensitive information with Copilot?
We store the logs generated by the call in the most secure manner, in accordance with Microsoft’s highest security standards. We save them because of course you want the conversation history.
You are also introducing Think Deeper, which will allow Copilot to deal with more hard problems. This is based on OpenAI’s o1 model, also known as Strawberry, right?
It’s like Strawberry, yeah. There is an OpenAI model that we have tuned to our more consumer-oriented goals and have it work in a way that is more consistent with our AI companion theme.
What are the differences?
OpenAI is much more focused on pure mathematics and solving scientific problems. We tried to focus on head-to-head comparisons and types of consumer analytics and things like that.
Or when you’re stuck on a hard problem or want to think about something, that’s when you can really provide a side-by-side comparison or perform a large-scale analysis.
Are Microsoft employees already using the fresh version of Copilot?
Yes, everyone benefits from it. A few days ago, we turned on general availability throughout the company. That’s why everyone uses it, giving lots of feedback. Our feedback channels are completely blocked. It’s great fun.
People will definitely remember Clippy, Microsoft’s last AI helper for Windows. Do people there see similarities?
Ha, well, I saw Bill Gates the other day and he was like, Do you realize you called this whole artificial intelligence thing wrong? This should be called Clippy. I thought to myself, man!
But that just goes to show how amazing people like Bill are. People who can see not just, you know, two years ahead, but 20 years ahead.
Are the fresh features a step towards so-called AI agents that perform useful work on the computer?
Yes, absolutely. The first stage involves the AI processing the same information you process – seeing what you see, hearing what you hear, and consuming the text you consume. The second phase is [AI having] long-term, enduring memory that creates shared understanding over time. The third stage is for AI to interact with third parties by sending instructions and taking actions – buying things, booking things, planning a schedule. We have these two features in an experimental R&D mode that we are working on.