Or maybe they didn't train it that way to be manipulative (although it's certainly a plausible explanation) but simply as an accidental artifact of trying to make it give honest answers?
LLM-generated images sometimes includes text from the prompt as literal text in the image, so perhaps this is the same sort of artifact? If they've told it to be honest, it responds by talking about being honest instead of actually being honest, because it has no actual understanding of anything.
This does make some sense if it's considered a valid fix to document that you have verified that Bluetooth is disabled on the servers and therefore not vulnerable. But that assumes that the scanning tool can be told about this kind of fix, so that it stops warning about it, which I guess it might not.
I know it's not the only problem (but it is IMO the main one), but this is why I'm a proponent of allowing operator overloading in languages.
Yes, you can't tell at a glance if `+` really still does what it should, but I find that problem no different from a library mis-naming a function and/or said function having strange side-effects.
`+` should do addition, period. No, not even concatenation if possible (I do find the lack of a separate concatenation operator to be a language flaw --- that said, I understand the use of `+` for concat in libraries targeting languages that offer no alternative).
I think the abuse (especially in C++ ... seriously, allowing overloading of the comma operator? --- and <iostream> working via bit-shifting streams by ${some_string} bits) gave everyone a bad taste and they kind of threw out the baby with the bathwater.
php has a separate `.` operator that concatenate and you get an error if you try to do it with a `+`. That's indeed better but most of the time you concatenate with `sprintf` anyway
Mostly. You can't do "show me this commit in the stack before and after the force push" though iirc. And you can't comment on the derived diff, even though that's what you'll be reading during re-review to see what changed.
GitHub constantly feels like features get requested with a one-sentence description, and then people who have never used any other major code review systems go build it without any further assistance or feedback, and eventually it just escapes containment and nobody tells the authors that there's a gigantic pile of bug feedback threads until almost a year later. Ten things get fixed, then another feature breaks out and focus shifts.
GitHub has a chance to get a generation of developers on to better coding practices. LLMs make it more feasible than ever to maintain high-quality patch series/stacked diffs — they can take care of all the mundane things like rebase conflicts for you. If I were in charge I would take this responsibility very seriously.
I guess the next best thing would be to port spr to this. (Probably time to start looking at this to be honest!) This makes me quite sad.
Not sure if it makes you feel better, but it seems there's at least a few startups trying to get a jujutsu-native forge going, by people who have seen and used the way Google's (and other big tech's) systems work.
That's really cool!
I tried getting on board with ERSC, but the timezone difference proved too much :(. At least it gave me the kick to try and build my own company, though!
It was definitely also the case before Microsoft, but I can't imagine it has improved since the acquisition. I try to avoid the site nowadays, it's randomly extremely slow and I would much rather spend my energy elsewhere anyway.
That is nowhere close to the experience Gerrit and Phabricator have, where you can diff two arbitrary versions in history. Comments on previous versions also get lost along the way.
This is the guy, if I'm not mistaken, who was fired for telling different groups people different and contradictory lies depending on what he needed them to believe. And now he's telling a reporter about what he's been telling the whitehouse, so he could be lying to the reporter, lying to the whitehouse or lying to both.
Well, yes. That LLMs are unable to distinguish instructions from data is a well-known and unsolved problem with LLMs in general.
This is one of the reasons it would be completely insane to give LLMs access to your data or rely on them for important tasks. But apparently that doesn't stop people from doing it anyway.
If you mean “would it allow arbitrary queries in the admin interface?”, then yes, it would. That’s why it’s the admin interface.
If you mean “would it allow arbitrary queries from normal view code”, then no it would not—not unless you string eval on the backend or something else dangerous.
LLM-generated images sometimes includes text from the prompt as literal text in the image, so perhaps this is the same sort of artifact? If they've told it to be honest, it responds by talking about being honest instead of actually being honest, because it has no actual understanding of anything.
reply