If @GitHub (Microsoft) truly believes copilot isn't infringing on anyone's work, I want to offer them a chance to prove it: I'll donate $50k to a charity of their choice (or @EFF if we can't agree) if they release a Copilot version trained solely on Windows kernel source. 1/
https://twitter.com/bphogan/status/1411097686854488067
This isn't a joke. It would be amazingly helpful for device driver developers. This in turn would ostensibly benefit Windows users through fewer BSODs. Add the charity money in and there's literally NO REASON not to do it. 2/
So let's set some ground rules:
Independent verification that the newly released model is only trained on kernel source
It's the full kernel source, from all versions (leave Win2k out if you want due to that pesky Java thing)
All kernel drivers owned by MSFT are included too 3/
Independent verification that the newly released model is only trained on kernel source
It's the full kernel source, from all versions (leave Win2k out if you want due to that pesky Java thing)
All kernel drivers owned by MSFT are included too 3/
No shenanigans
The custom copilot version is published before July 31st
If they want, I'll even present the person in @Microsoft who made it happen with the charity check at their Blackhat party so they can milk maximum PR out of it. 4/
The custom copilot version is published before July 31st
If they want, I'll even present the person in @Microsoft who made it happen with the charity check at their Blackhat party so they can milk maximum PR out of it. 4/
If you are similarly convinced that this is a good idea, reply with your own pledge so it gets harder for @Microsoft and @github to ignore this.
Of course they're free to ignore, but that certainly doesn't speak well for the "it's just a model" argument... /FIN
Of course they're free to ignore, but that certainly doesn't speak well for the "it's just a model" argument... /FIN
• • •
Missing some Tweet in this thread? You can try to
force a refresh
