I'm curious what folks are thinking about this? I've seen a range of opinions on twitter. Does anyone know if open source maintainers / contributors / community members were an active part of the development of the feature?
For further actions, you may consider blocking this person and/or reporting abuse
Top comments (6)
Re: being included, my bet is that if you haven't opted out of the GH Copilot data collection yet, it's probably safe to assume you were somehow in the current training set. Even if you opted-out now, it might be too late, although it'd be interesting if researchers could confirm the data retention by writing open-source honeypot GH repositories. And unfortunately it seems like you wouldn't even have control of opting out if you didn't have control of a repository you contributed to.
Re: royalties, although I like the idea of people getting paid for their scraped code, it seems very unlikely given the challenge (e.g. imagine the case where someone got royalties for a function but they actually copied it from someone else's repo). ML is becoming so ubiquitous that "being part of the training dataset" seems like it's been normalized.
And even though functions have a lot of value, I think most of the value in software comes from applying those functions as a group and building something (or releasing versions, handling PRs, triaging bugs, and the many other things Lifters do!).
What an interesting new world we're in now 🧐
I personally feel like it's interesting that GH is able to monetize freely-accessible code content without paying the maintainers directly through royalties for this service. 🤔
Copilot has caused me some serious indigestion. I do know that they involved some open source maintainers in early demos but I got the sense they kept the circle very small during active development. It launched when I was OSI President and, I gotta say, no one in the foundation community was impressed – none had been consulted.
Both Copilot and things like DALL-E raise some tough questions. I've heard them described as "intellectual property laundering."
While GitHub's lawyers are of the opinion that they're honoring the letter of the open source licenses, I'm of the opinion that they're certainly not honoring the spirit of them – and that the legal questions remain unresolved.
But don't get me wrong, I'm no luddite. I had lots of conversation with Idan at GitHub, the fellow who lead the team that launched it, and I see the value of tools like this. Fundamentally, I think it needed to be something that people explicitly opted in to.
Amazon CodeWhisperer has entered the chat... siliconangle.com/2022/06/23/aws-de...
Add to the discussion
I am leaving GitHub, in part due to Copilot, but there were a lot of other issues too.
I wrote about my reasons here: