Why do you even care? We use crawler on open source for year without anyone shocked about that. It's not even somethings they hide for years. They always claim to train on open source code. Open source is open source anyone can read the code. It's not like thy copy past code in they're program (even thought no one can say if they filter project base on license) it's just training. If the code appear in the end product is purely because the code was use by a lot of people and so can we really say that the code is protected by right. To me for this kind of code it's like trying to protect a color.
Also Copilot is not the only one training on opensource it's concurrent do it too. And has far as i understand they never say that they used Github only. From my understanding has soon has you're code is in read format and open source they could use it. And since a lot of crawler use gitlab and github i assume they at least do both.
To me it's a great idea. I don't really like copilot but for ML you have to get a lot of sample so why not ? Comment on Amazon are use for ML. Picture on FB too. I assume when you put somethings public you expect that some people will make stat on it. If they can make stat why can't they train a program on it?
Why do you even care? We use crawler on open source for year without anyone shocked about that. It's not even somethings they hide for years. They always claim to train on open source code. Open source is open source anyone can read the code. It's not like thy copy past code in they're program (even thought no one can say if they filter project base on license) it's just training. If the code appear in the end product is purely because the code was use by a lot of people and so can we really say that the code is protected by right. To me for this kind of code it's like trying to protect a color.
Also Copilot is not the only one training on opensource it's concurrent do it too. And has far as i understand they never say that they used Github only. From my understanding has soon has you're code is in read format and open source they could use it. And since a lot of crawler use gitlab and github i assume they at least do both.
To me it's a great idea. I don't really like copilot but for ML you have to get a lot of sample so why not ? Comment on Amazon are use for ML. Picture on FB too. I assume when you put somethings public you expect that some people will make stat on it. If they can make stat why can't they train a program on it?
Also i get arround this :
sfconservancy.org/blog/2022/jun/30...