This was difficult to capture. Copilot was updated today through the Microsoft Store and now displays a message for a split second. Copilot stores our conversation to train its AI model and to “learn about my interests”.
It is basically the opposite of what Microsoft tells its customers. This is the only option in settings:
Copilot is like IBM Watson. The label gets slapped onto lots of diverse products. I am talking about Copilot as loaded in Windows if you hit the Copilot key on Copilot+ PCs. You can also get the same Copilot on copilot.microsoft.com – here is a screen recording which contains this message. Play close attention. It is only visible for a split second.
Is the “training on conversations” hidden behind the link to the privacy statement?
Anyway: Very shady stuff – consistent with the whole AI stuff
If that works as well as on Facebook I predict a disaster:
My sister owns and rides a horse, so I sometimes like her posts or even click on a link. Even went on the page of a horse training organisation when I was looking for a present for her.
Does that mean I’m interested in horses? No, not except a brotherly interest because I want to understand what she’s talking about. Yet I keep getting served ads for all kinds of horsey things. Clothes, food, whatever. Completely wasted on me.
And there’s more similar to that. A friend posting about a holiday in a place I don’t really care about, yet I still wish them a nice holiday. And boom, ads and pages about that place.
I fear the Microsoft AI will learn similarly, because it doesn’t know and understand context and meaning. Which means it will not really understand my interests and provide lots of useless information.