Hacker News Reader

Comment by antihero | Hacker News Reader

I mean, if it's autocompleting a fairly simple line, and can do that because it's analysed a lot of lines, I don't really see that as "stealing anything".

If you are using it to write whole complex functions thatare the same as other people's, I guess that is copying.

But if you do the second thing you are not a great dev, and would have probably ended up copy pasting it anyway.

I think the first use case is far more common, and creating boilerplate that is so generic you could never really attribute it anyway.

View on HN

> But if you do the second thing you are not a great dev, and would have probably ended up copy pasting it anyway.

If you do that on your own, it's your (legal) responsibility. If Copilot does it for you, it's GitHub's/Microsoft's responsibility.

Why should be GitHub's/Microsoft's responsibility? No one is forcing you to use copilot.

If I use grammarly, are they responsible for what am I aiming to write?

Does Grammarly gerate pages of content for you?

If I pay for grammarly, and it plagiarizes an existing work but represents it as an entirely new, independent work and I am unaware of the existing work that is being stolen, who is doing the stealing?

This makes more sense for text message auto complete: you just take the suggested next word after a one word start deed, it might reproduce a Wikipedia entry. But what did tub expect? The same would be true with grammarly if you somehow got it to produce a bunch of new text. You expected garbage, but somehow infringed on copyright instead. But I guess think the user deserves some responsibility in realizing their expected garbage output isn’t for some reason.

If you pay a shady character to get you a modern laptop for $100 you can't claim that you were unaware that it was most likely stolen and the fact that you paid for it something doesn't absolve you morally.

does shady guy have his name on the side of a building, and run ads: "buy my shady stuff" and then pay taxes and his earning? That kind of shady guy?

Sometimes. Like Amazon, widely known for their workers and 3rd party vendor exploitation practices.

You can no more claim ignorance of where the github copilot code comes from than where the Amazon's low, low prices come from.

Whether you care is totally on you regardless of whether you pay ir not. You pay for product or service not moral absolution.

Are thousands of amazon employees going to be in same docket with me?

So it's my job to check my supplier, to make sure lines from co-pilot are legit.

At the same time when fast fashion companies sell T-shirts made with slave labour, its not the company's responsebility to check what their suppliers are doing.

And if tesla autopilot kills you and your family its not their fault either.

Neoliberal morality - companies are never accountable for anything, it's heresy to suggest they should do their job properly.

Other than the first sentence nothing you wrote is true. If a company doesn’t do due diligence on their suppliers they face fines and possibly criminal charges. The news came out the other day that the NTSB is considering whether to require Tesla to recall all their vehicles with self driving enabled. Companies of all types face huge fines and civil liability for product safety issues.

have you never googled "slavery fast fashion"?

Zara's clothes sometimes have notes in their pockets from people being held as slaves, pleading for help. I havent heard of anyone going to jail

most of our electronic waste end up illegally exported to poor countries, again when was the last tomw someone faced the music for that?

> If Copilot does it for you, it's GitHub's/Microsoft's responsibility.

Is this true? It hasn't been tried yet I assume?

>If Copilot does it for you, it's GitHub's/Microsoft's responsibility.

GitHub/Microsoft says that it's still your responsibility.

>You should take the same precautions as you would with any code you write that uses material you did not independently originate. These include rigorous testing, IP scanning, and checking for security vulnerabilities. You should make sure your IDE or editor does not automatically compile or run generated code before you review it.

I'm not really sure how am I supposed to go about validating that I can in fact use this code that the magical black box barfed into my IDE using a bunch of different weights.

> GitHub/Microsoft says that it's still your responsibility

If Copilot is fair use, and has no restrictive license, than how is it anyone's responsibility?

If Copilot isn't fair use, it's Microsoft's responsibility.

(For copyright; for patent that's another issue, but you can violate patents by similarity without exposure or copying, anyway.)

Training Copilot is fair use, using Copilot is ???.

Let MS buy BlackDuck scanner and integrate in to GitHub/CoPilot. They could then suggest code and also scan it for any license violations, and give you both sides of the equation in the same tool.

Just what a horrible shady behavior.

Give us your money but you are responsible for the code that OUR tool generates.

You are responsible for your tool use. That's the same discussion as with whether uTorrent is responsible for your torrenting copyrighted stuff or with Tesla's auto-pilot. You buy the tool, you are responsible for what you create with the tool.

Napster was liable for copyright infringement.

Napster's raison d'etre was copyright infringement.

Which they were then liable for.

True, however, the users have been liable too. If my company gets sued because I used Copilot, it won't matter that much that the plaintiff also sued GitHub/Microsoft.

The when Oracle won its copyright lawsuit against google it was because of a 8 line bound checking utility function.

Source?

https://news.ycombinator.com/item?id=11722514

Thank you

Not only was that not the only code in question, you left out the conclusion! https://en.wikipedia.org/wiki/Google_LLC_v._Oracle_America,_.... It went to the supreme court, they concluded fair use, FU oracle.

I was talking more about ethics than the law. I try to live adjacent to the law (to an extent that is reasonable and sane) in all fairness, as it will never be in-line with my ethics.

The first can be automated without ML though. And once you use ML you cannot guarantee it won't copy-paste existing code.

This whole thing would be fine if GitHub hadn't just used all public code on their platform, ignoring all involved licenses.

The problem is, if they had used only code with a license that allows copying without attribution, there wouldn't have been a lot of code left...

Difficulty doing something legally doesn't justify breaking the law.

It changes the code for use. I'm not sure it can be considered a copy. It much like reading someone else's code and drawing ideas and patterns from that code.

It has been shown often enough that Copilot can reproduce exact copies of snippets.

I neither see it "stealing". The neuronal network was trained with code as input. It's creating code as output. The output has nothing to do with the input once it is trained. Do people dont know how neuronal network work?

It's like saying GPT-3 created text is copyright infringement, because some author used the same sentence in a book before.

Overfitting: One weird trick that copyright lawyers don't want you to know!

So if I fit a network to output entire chapters of a book when given the chapter number as input, I can print and sell copies of it that way?

1) Copilot is not designed to output the source code for a project source file 2) It does not re-create the whole source code, just parts of it (sentences, not chapters) 3) The source code license, e.g. BSD, works on "the code" - copying a line like "void main(void) {" will not trigger it, obviously

1. Create a neural network that produces an x264+dts stream of a movie 2. distribute it 3. checkmate copyright lawyers

>Do people dont know how neuronal network work?

I could ask you the same thing.

My problem is with the weights not being released. They are a derivative work of open source code in the most literal sense. The weights would not exist without those lines. Gradient descent is using literal derivatives.

> If you are using it to write whole complex functions thatare the same as other people's, I guess that is copying.

> But if you do the second thing you are not a great dev, and would have probably ended up copy pasting it anyway.

How would I know that the boiler plate I ask copilot to write for me is copied verbertim from a codebase, that neither I nor Microsoft has licensed to use?

Reply on news.ycombinator.com