Re: debian-user Question about: Debian puts artificial intelligence to a vote; developers will decide how to accept AI-generated or assisted code
Joe <[email protected]>
| Newsgroups | gmane.linux.debian.user |
|---|---|
| Message-ID | <[email protected]> |
On Fri, 21 Aug 2026 14:43:54 -0600 Charles Curley <[email protected]> wrote: > On Fri, 21 Aug 2026 19:16:59 +0000 > [email protected] wrote: > > > it wasn't my intent to ruffle anybody's feathers > > Understood. > > > only that i am concerned about what my computer is doing that i > > didn't consent to > > In that case, I suggest you be less concerned with AI per se and more > with malware, whether produced by artificial intelligence or the > natural kind. > > Tools are amplifiers. Throughout history most tools have amplified > muscles: the block and tackle, the hammer, the screwdriver, etc. Some > tools amplify the mind: writing, the computer, and now AI. Any tool > can be used for good or ill. I suggest one should be concerned with > malice and not with tools used with malice. > Yes, but I think that AI is again vastly increasing the scope for malice. Back in the days of the BBC Micro, Apple II etc. there were a few tens of kilobytes of operating system code. A Win95 installation was about 25MB to begin with, and a Win11 installation is typically 80GB in size. That's an enormous increase in size and a corresponding much better hiding place for anything malicious. We know that current Windows contains a lot of spyware, MS is even quite open about it now. AI as used in Open Source products may indeed produce just source code which could theoretically be audited, but inevitably, it would be audited by another AI. We all know how many humans could be bothered to audit the comparatively tiny chunk of extremely critical code which contained the Heartbleed bug. So when you enter your script and get back a couple of hundred megabytes of code, what is in it? Who had a say in what is in it? Since the LLM will have been trained on a large chunk of the Internet, which we know is not completely error-free, how trustworthy is it? AI is indeed a tool, but it is a tool which you pick up without knowing either who manufactured it or what it is capable of doing in addition to what you expect.