News for hackers. Stuff that matters.

A comment on LLaMA: A foundational, 65B-parameter large language model. Parent comment

Re:LLaMA: A foundational, 65B-parameter large language model by ldhough on Friday February 24, 2023 @02:32PM (#34928906)
For a specific example some variation of "fast inverse square root" will usually get you the exact GPL licensed code from Quake III, comments included.
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by theRealMe on Friday February 24, 2023 @03:25PM (#34929615)
Do you mean the same code that has its own Wikipedia page where the exact code is written, comments included, and has probably been copy pasted into 100’s of other projects?

https://en.m.wikipedia.org/wiki/Fast_inverse_square_root

[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by robertlagrant on Friday February 24, 2023 @03:38PM (#34929787)
The point is that it's charging having been trained on open source code. What you're saying agrees with that, but your triumphant tone seems to be implying the opposite. Which did you mean?
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by Mizza on Friday February 24, 2023 @06:07PM (#34931499)
You mean this code?

https://archive.softwareheritage.org/browse/content/sha1_git...

Do you see that notice at the top of the file? It says:

==

This file is part of Quake III Arena source code.

Quake III Arena source code is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

===

but because it's been laundered by Microsoft, you think it's okay to steal free software and make it proprietary?

[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by warkdarrior on Friday February 24, 2023 @06:36PM (#34931802)
How is it made proprietary? The Quake III Arena is no more proprietary now then if it were stored on GitHub proprietary web servers. Copilot is just a fancy code index, that sometimes returns the original code and other times it gives you a modified copy.
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by Mizza on Friday February 24, 2023 @06:42PM (#34931869)
Because as you say, it provides original or modified code but doesn't provide provenance or license information. It's copyright laundering. After decades of fighting the community in the courts over shit like this, Microsoft just turns around and says well, it's okay when we do it? Foh.
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by imtringued on Saturday February 25, 2023 @05:16AM (#34935470)
The problem is you have to obey the license of the code even if you just take a snippet and Copilot does not reproduce the correct license.
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by charcircuit on Friday February 24, 2023 @07:59PM (#34932533)
That code didn't originate from quake
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by WithinReason on Sunday February 26, 2023 @07:03PM (#34951196)
"The algorithm was often misattributed to John Carmack, but in fact the code is based on an unpublished paper by William Kahan and K.C. Ng circulated in May 1986"
[ Parent | on HN ]
Re:LLaMA: A foundational, 65B-parameter large language model by ldhough on Sunday February 26, 2023 @01:05AM (#34943489)
Yes that code, I was replying to a comment claiming that

> Copilot has a tendency to regurgitate [code] verbatim, without said license.

and I think that is a pretty good example.

[ Parent | on HN ]