I don't know this project and I don't understand what this is. Are they using AI to generate package info, and then using a bot to handle their Github issues? What does the project actually do, that something so useless was deployed?
Top tier comedy. Even the Github bot replies to the question are LLM nonsense: they simply agree with the above comment and apologize for not knowing the answer.
What's interesting is that this was created by Max Howell, creator of Homebrew who made waves a little while back for getting rejected by Google despite Homebrew's success. He talks about that here: https://www.quora.com/Whats-the-logic-behind-Google-rejectin...
To his credit, he didn’t let being rejected by the largest advertising surveillance company stop him from shipping surveillance software to millions of end-user workstations anyway, in the form of the nonconsensual spyware built into Homebrew.
The pkgx[.]dev website has some kind of `curl | sh` type installation commands to install packages. Given the AI generated descriptions of packages on their websites, I hope this isn't something other than an elaborate way to deliver malware to anyone who stumbles upon this website via a Google search.
I would highly encourage everyone that I strongly dislike to find random things with nonsensical artificial intelligence generated descriptions and then sudo curl | sh
> What is the point of dosubot? It agrees, and apologizes for not having provided a supposedly adequate response.
This is my problem with chat gpt - I can’t have anything resembling a ‘real’ conversation with the bot, because it refuses to engage with me in any way other than:
1) restate my last message
2) agree with what I’ve said to an excessive degree, possibly even complimenting me on having said it
3) ask me a leading question that encourages me to send another message, and now goto 1 and loop forever.
It ends up just feeling like I’m the target of some kind of time wasting prank.
I think the problem is that AI is wrong so often that it would be foolish for it to stick to its guns when you're correcting it.
Bing Chat did this at first, and started calling the user a manipulator and an abuser when you corrected it too much (regardless of whether the previous message made any sense or not). I found it really funny, but other people were distressed by it because I suppose they thought they were talking to something sentient.
The problem is probably specific to the default system prompt in the default ChatGPT UI version.
An open weights LLM can be made into a pretty engaging bot with the right prompt/model/finetuning. I use an uncensored LLM with a special system prompt which makes it answer directly without all the fluff.
Already getting this. Absurdist summaries of my company's history page that read like a 8th grade book report. Thing is I bet they use AI to parse responses, which means there's many opportunities for prompt-injection.
> Hey @pawamoy! Great to see you back here, diving into the intriguing world of pkg descriptions. Let's take a closer look together!
> [and so on for 94 pages]
Nope, y'know what, close issue, I don't care any more. (If I were pawamoy.)
This kind of crap's even worse than when you get a 'stale' tag for inactivity. Sure, true, nobody else has said anything, but why does that discount my buggy experience or whatever? I would understand if it was because I didn't reply for so long after a maintainer response, but when it's just because there hasn't been any update it's just annoying. Nobody likes '+1's, so don't force even more notifications from label added, pointless comment added to bump, label removed!
> This kind of crap's even worse than when you get a 'stale' tag for inactivity. Sure, true, nobody else has said anything, but why does that discount my buggy experience or whatever? I would understand if it was because I didn't reply for so long after a maintainer response, but when it's just because there hasn't been any update it's just annoying. Nobody likes '+1's, so don't force even more notifications from label added, pointless comment added to bump, label removed!
I absolutely hate repositories that do this.
You're sending a clear signal that you don't care about bug reports from users. I bumped one issue for about a year (I seem to remember it was security-related too, to make matters worse) whilst the bot repeatedly tried to close it. It never got fixed.
The one thing AI has done really well is generate all the signals of "intelligence" that SV tech folks care about. It's verbose and meaningless, confidently-asserted bullshit wrapped in the right packaging. It's like a Markov chain generator with a Stanford degree.
Readit News
Describes a lot of tech hype stuff in the last couple of decades.
Dead Comment
Deleted Comment
This is my problem with chat gpt - I can’t have anything resembling a ‘real’ conversation with the bot, because it refuses to engage with me in any way other than:
1) restate my last message 2) agree with what I’ve said to an excessive degree, possibly even complimenting me on having said it 3) ask me a leading question that encourages me to send another message, and now goto 1 and loop forever.
It ends up just feeling like I’m the target of some kind of time wasting prank.
Bing Chat did this at first, and started calling the user a manipulator and an abuser when you corrected it too much (regardless of whether the previous message made any sense or not). I found it really funny, but other people were distressed by it because I suppose they thought they were talking to something sentient.
An open weights LLM can be made into a pretty engaging bot with the right prompt/model/finetuning. I use an uncensored LLM with a special system prompt which makes it answer directly without all the fluff.
Behold, RimWorld's official forums : https://ludeon.com/forums/index.php?board=1.0
But yeah, it does seem to be getting worse as neural networks get better... :(
In the past we didn't have the resources to fill github with endless pages of blogspam, now we do.
> Hey @pawamoy! Great to see you back here, diving into the intriguing world of pkg descriptions. Let's take a closer look together!
> [and so on for 94 pages]
Nope, y'know what, close issue, I don't care any more. (If I were pawamoy.)
This kind of crap's even worse than when you get a 'stale' tag for inactivity. Sure, true, nobody else has said anything, but why does that discount my buggy experience or whatever? I would understand if it was because I didn't reply for so long after a maintainer response, but when it's just because there hasn't been any update it's just annoying. Nobody likes '+1's, so don't force even more notifications from label added, pointless comment added to bump, label removed!
I absolutely hate repositories that do this.
You're sending a clear signal that you don't care about bug reports from users. I bumped one issue for about a year (I seem to remember it was security-related too, to make matters worse) whilst the bot repeatedly tried to close it. It never got fixed.