I imagine if this attacker wasn’t in a rush to get the backdoor into the upcoming Debian and Fedora stable releases he would have been able to notice and correct the increased CPU usage tell and remain undetected.

I think ideas about prevention should be more concerned with the social engineering aspect of this attack. The code itself is certainly cleverly hidden, but any bad actor who gains the kind of access as Jia did could likely pull off something similar without duplicating their specific method or technique.

as long as you’re up to date on everything here: https://boehs.org/node/everything-i-know-about-the-xz-backdoor

the only additional thing i’ve seen noted is a possibilty that they were using Arch based on investigation of the tarball that they provided to distro maintainers

if the stakes are so low then blocking them is as low-stakes as not, so why make a fuss about it?

i also remember having the cube around the same time in OSX somehow but I forget the method

just having a couple brands and models to look into is pretty helpful. I know i’m sick to death of massive screens with touchscreen buttons for everything. i drive my step mother’s recent model BMW sometimes and trying to just change the radio station is a chore that involves clicking accept on a screen with a huge disclaimer warning you not to get distracted by the touchscreen while driving before you can even operate it.

Plenty of new cars don’t have an obnoxious infotainment screen as well.

maybe list them for OP instead of hinting at their existence.

The line between problem and solution for C should be 30 miles long.

deleted by creator

if it did, they wouldn’t be nerds anymore.

Yes that is what they are good at. But not as good as a deterministic algorithm that can do the same thing. You use machine learning when the problem is too complex to solve deterministically, and an approximate result is acceptable.

the comic is about using a machine learning algorithm instead of a hand-coded algorithm. not about using chatGPT to write a trivial program that no doubt exists a thousand times in the data it was trained on.

I didn’t drop the prompting. over half that comment is specifically an analogy about prompting. are you ok

You can ‘explain’ all the technical details you like but nothing is going to change the fact that it was put out as it is, after careful work to make it as close as they could to how they wanted it. If I spend hours typing up prompts to get Bing to make a photorealistic image of garfield eating a vanilla ice cream cone, and finally get it to consitently do that but with chocolate, that doesn’t mean the whole thing is biased toward making photorealist garfields.

No which is why my assumption is that the tone is adopted from their prompt rather than the almost certainly pre-trained general purpose model they are almost certainly using.

They can tweak the prompt in order to make it sound how they want. Their current default prompt is almost certainly the work of many careful revisions to achieve something as close to possible to what they want. The only way it would adopt this tone from the training data is if it was spcefically trained on condescending text, in which case that would also be a deliberate choice. I don’t know how to make this point any clearer.

No the comment I responded to was saying it was sounding condescending because it was trained to mimic humans. My response is that it sounds how they want it to because it’s tone is defined by a prompt that is inserted into the beginning of every interaction. A prompt they tailored to produce a tone they desired.

I didn’t say they specifically told it to be condescending. They probably told it to adopt something like a professional neutral tone and the trained model produced a mildly condescending tone because that’s what it associated with those adjectives. This is why I said it was only somewhat instructed to do this.

They almost certainly tweaked and tested it before releasing it to the public. So they knew what they were getting either way and this must be what they wanted or close enough.

There’s a default invisible prompt that precedes every conversation that sets parameters like tone, style, and taboos. The AI was instructed to behave like this, at least somewhat.

It’s an open source project. It has no investors driving it toward user hostile profit seeking which is the primary force behind enshittification. A large user base doesn’t cause it, merely triggers it where the cause is already present.