I don't entirely trust that they will remain excluded from Copilot in future. This probably isn't the place to re-litigate this discussion, but GitHub's claim is that source licenses simply do not apply to what Copilot ingests.
That being the case, the only thing that distinguishes private repos in this context is a thin policy that can be changed at a whim (and perhaps without any announcement).
Also, one of the reasons I put my code on a host like GitHub is so that I can share it/show it off*. So using a private repo to avoid Copliot defeats some of the purpose of me using GitHub in the first place.
I think it's _exceptionally_ unlikely they would use copilot in private repos. Not for licensing issues, but for security issues. People include secrets all the time in private repos. Trying to get an AI that can train on that sort of data, but avoid including these secrets would be a technical nightmare, and they'd get sued as soon a secret inevitably leaked out. And they have no incentive to; there's _plenty_ of data available in public GitHub to make copilot amazing. And they have access to that data in a way no other company does, so competition is going to be waaaays away.
Outside of that, I see copilot like the future of search engines. Instead of searching for plain text matches, it lets you search for abstract and complicated patterns. The killer feature for me would be seeing for a given suggestion (or search result), where the suggestion was derived from, linking back to the code snippet in GitHub.
