On Laundering Licences and Code
A "Second SCO" waiting to happen
Yesterday: The Label 'Anti-AI' (as used by Linus Torvalds against people concerned about legitimate risks)
Who's even checking the source (origin) of the code inserted into Linux?
"Perhaps another reason why Linus Torvalds also takes this stand is because he knows that machine generated code, which only concerns "AI" and not compilers nor editors, is very problematic with regard to the GPL license. The "AI" companies are not recognized under copyright law so copyright systems do not have a way for the machine generated output to contain the grant of permissions which the GPL license requires in order to perform e.g. redistribution. By removing the previous obligatory confession of usage of "AI" tools in the development of the Linus kernel, the problem with the GPL is sidestepped. This can only be fully intentional and not just because Linus Torvalds considers it otherwise idiotic. He knows full well that if the requirement is maintained, nobody can legally use any "AI" generated code in the kernel. Without the requirement, you cannot prove that committed code has been copied from code which is under a license that is incompatible with the GPL. This is sheer hypocrisy. Just search for old videos with Linus Torvalds explaining why he thinks that Linux was such a massive success, he himself considers this to be because of the GPL." -Linus Torvalds, shame on you!
Code licensed under some terms can be plagiarised with permutation to make it seem sufficiently unique or differentiated (obfuscated). It has long been possible. Bypassing the licences by various loopholes predates slop - with or without some degree of automation. Is that "fair use"? Irrelevant.
Fuzzing has also long been possible. Many bugs can be found using brute force, but you then need to spend vast amounts of time separating the wheat from the chaff.
The latest LWN release of articles (from the paywall) includes several about slop, including one about Debian, setting aside one about the kernel.
Debian volunteers are rightly upset that they're expected to do the jobs bots throw at them. As mostly unpaid workers (most aren't paid anything at all), why bother?
They call those "reviews", which is a euphemism for chaff.
There are various licensing problems associated with those "Plagiarism Engines", but there are also human factors. At what point does a project walk blindfolded, praying and hoping that unaudited code is 1) not a bug (back) door; 2) not a licence risk; 3) not buggy and 4) not a performance bottleneck that will become difficult to pinpoint later as nobody understands the code or general execution flows?
Linux (the kernel in particular) is becoming slop-ware and Torvalds boasts that instead of finding bugs the slop is now producing "fixes" (i.e. code).
Should we cheer or shudder?
Aside from the licensing risk (this series of ours mentioned the SCO aspect), who's there to assure quality? Many veteran developers leave and the code is spewed out by GAFAM LLMs that GAFAM employees said publicly were making their infrastructure unreliable.
The worst part, to me, is the licensing, not the quality. With billions of copies of Linux running out there every day (even in Android) it's an attractive target for lawsuits and there are already patent attacks that OIN did nothing to protect against (this week's news!). What if copyright lawsuits come next? Who's even checking the source (origin) of the code inserted into Linux? This code didn't "grow on trees"; it was grabbed/grafted. The original developers can see it and prove it. Beware slop fanatics bearing gifts. █