Well actually, you can. Just get a syringe.
The problem is that it might just be too sloppy(sic).
I won’t argue that I despise much of what AI is pumping out to social media, because actually my focus is more on the idiots that ‘enjoy’ it.
I think people are the biggest problem, not AI.
An update on this:
due to incorrect dependency resolution for non-AUR packages; adding to that each AUR PKGBUILD must be reviews manually by the user and the user must build all AUR PKGBUILDs inside clean chroot(which will break most of the PKGBUILDs hosted on the AUR)
I’m guessing this explains why some people ended up with AUR versions of some cachy packages recently.
Good to see a heightened security focus too.
any suggestions to use paru/yay for all operations(-Syu, -S) is dangerous to the system
This is something I often tried to impress upon new users. There is a growing habit of people using AUR helpers to “bundle” software operations and that’s caused many issues with dependency resolution between native and AUR that newer users are just not equipped to handle themselves.
As far as the original topic here goes, I don’t think there was any harm in including an AUR helper in installations by default, but the issue seems to always be external influences(AI chatbots, YouTubers, etc) improperly pointing new users toward poor system maintenance practices that they would’ve otherwise been unaware of.
Ultimately, even though I dislike that the blame for users’ poor decisions is falling on poor paru’s head, it’s not like it really matters. We can just install our AUR helper of choice and go about our day. paru never really needed to be installed by default, even if it was convenient that it was.
As long as bootstrap requirements are met (some way), it’s good. However, at least for me, choice of distro is the recipe(s) they use in software pre-load/configuration and the winner is the one that requires me to do the least in getting a system up.
I agree with this. I’m a staunch supporter of the Arch philosophy yet I still use CachyOS instead because it comes much closer to my ideal system state. I suppose it’s a matter of subtracting less from CachyOS than I add to Arch to get there. However, I would mostly attribute that to the CachyOS Kernel and its repositories.
paru ultimately plays a small role in the setup cocktail, but it does play a role for me. I use only three AUR packages but ideally I want all three of them installed the first time I load into my desktop environment, so making sure I have an AUR helper is one more step I have to do that I didn’t before.
I don’t know if it will eventually amount to me making a switch, but I can’t deny that:
- shelly
- cachyos-packageinstaller
+ paru
+ octopi
are points against CachyOS. And I suppose it’s a little sad that until a couple months ago three of those steps didn’t exist. You’d like to think that your chosen distro moves more toward your preferences than away. But that’s life I guess!
Here’s the trick: I don’t need to put it back in the tube. I can simply cap the tube, clean up, and move on. I don’t HAVE to use LLMs, and no “inevitability” argument holds enough water to fill a coffee cup.
EDIT, after having read further replies: This is where I agree with @CacheMeIfYouCan , in that none of this slop being pushed on us would have more than a fart in the wind’s chance if people in general were more discerning, less credulous, and better trained in critical thinking skills. The tool is just a tool. A good one, a bad one, an ethically unsound one-- the real issue are the people treating it like a prophet, or signing away all their mental processes to it and letting their minds atrophy.
Couldn’t agree more with the above. I don’t, however, agree with your general sentiment that the LLMs are not “inevitable”. While YOU personally may not have any use of them, they are, nonetheless, here to stay. I would argue that, taken your point on personal responsibility, LLMs can help people be more creative and clever. While simply copying blindly any script that a bot spits out is just as stupid as copying ANYTHING off the internet blindly, the use of LLMs by people who are intelligent and knowledgeable can unlock a lot of human potential. I am sure that there will always be the bad actors and slop in the LLM world, any new tech is usually an equal part a blessing and a curse, but with the careful curation and self-education people can harness the enormous potential they offer.
To bury your head in the sand about this is akin to the buggy whip manufacturer thinking that the Industrial Age is just a “new fad” that will go away in a few years.
Ergo: Sooner or later, AI will (must) eliminate humans (Why does this remind me of paru right now…?) .
Nothing is inevitable.
See above what you quoted from my words about the credulous ascribing this new tech unfounded mystical status.
You say it can help people be more creative and clever, but the clever have little need of an unreliable tool that can simplify (with errors) tasks they can already handle, and the creatives—well, the creatives—I’m a musician, an artist, and a writer, and I’m here to tell you that in every one of those circles, the vast majority want to see these things left by the side of the road with the passion of a million suns. That people think they could shortcut cultural context, the honing of craft, the journey of discovery and understanding that allows a creative person to build their skill set and then veer onto their own path, while connecting with other humans through art—there is no shortcut that doesn’t ring hollow, that isn’t uncanny valley. We aren’t afraid of LLMs taking our jobs, because they aren’t. We’re concerned with the environmental damage their data centers are causing, the economic impact AI startups’ trash-tier business models will have on everyday people when the bubble collapses (it’s begun, by the way), and way that such failed businesses are propped up and couched as “inevitable” in an almost-religious context.
There is no shortcut through art here. Many of the top traditional publishers, editors, authors, and professional organizations outright ban LLM use, both in the writing of the stories published and in the administration of publishing them. Art galleries, contests, and prestigious awards do the same. “AI music” is nearly universally shunned, because without the human story connecting the listener to the artist, you might as well be listening to nothing at all.
In short, creatives are staunchly against. For further examples, see film actors on picket lines, activists who fought past corporate overreach and greed stepping in once again to resist this fictitious “inevitable” narrative. They’re also protesting the theft of intellectual property to train LLMs. I have primary sources indicating that AI startups have spent billions illegally buying books, music, and other media in order to shovel it into their lackluster models. Models which, to date, have not revolutionized anything as release of the iPhone did. The hype and marketing are all about why you need an LLM, not why you need THIS one over THAT one. If the things were truly so good, they wouldn’t have to sell you on the concept of buying it in the first place. No car manufacturer ad tries to sell you on buying a car; they sell you on their particular car, and if these startups had produced anything of value, we’d see those stories. A company would make headlines claiming they let AI design a new product and it would be a huge success. This has not happened. The media, complicit in all this for some idiotic reason, simply parrots what startups tell them without fact checking the financial books or veracity of any claims, and the claims are almost always in the future tense. Vaporware.
I could go on and on, but we’re far off topic for this thread, and at any rate, the thread should be closed as the discussion/debate is ended anyway—paru has been removed in the latest build.
If you want to debate the usefulness of LLMs, tag me somewhere else. I laid out a screed in the vent/rant thread already.
You’ve chosen your hill to die on, and it’s an odd choice. Dagrod has it closer to the truth - and the ultimate aim.
Do you grab a rock and start chipping away artfully to create a beautiful shiny stone axe to cut down a tree? or do we have better tools at our disposal?
You made an utterly ridiculous statement right there. Just watching a TV series about factories - have you seen how eggs are sorted now? Nobody trusts PEOPLE to do that stuff, because people are not the best tool for the job. Cameras are cheaper, they can LOOK at hundreds of eggs, a sophisticated LLM can make judgements and remove any suspicious items in a fraction of a second…
I see AI software taking great strides in farming, fully autonomous tractors and drones, cowsheds and other smart buildings - doing a lot of work eliminating humans, but also improving a great deal on what humans are capable of.
There is no competition in many areas - there are aspects of medicine where automation is the GOAT - ultra-reliable detection beyond human ability. The only arguments are around the ‘intelligence’, and the cognitive dissonance that comes with LLM’s pretending to be personable, apologetic, polite, and self-referencing instead of simply communicating as neutral processors - cutting out speculation (or at least labelling it as such), prefacing and opinionated, and with too much filler language.
I imagine if I’d had an AI tutor at home, so that after going to college in the '80s I could come home and test myself…
I have a Thai wife, but she is no teacher - I started self teaching Thai years ago, initially just using text to speech in messenger apps, practicing speaking through text to speech in translator apps.
Now I can have a full conversation with an LLM - because Language is what it excels at most… so it can parse search results in Thai and Chinese and think about them in English.
There’s no doubt in my mind - it’s no fad… it’s just vastly over-estimated in terms of actual intelligence which is why I prefer to refer to them as LLM’s.
They’re just models, tools.
No one disputes the benefits AI can bring to us humans… but … where will it stop being merely a tool …?
We humans are the ones allowing ourselves to become dispensable. Every half-baked piece of crap churned out by the latest AI model gets hailed as gold. Sorry, I know you can’t put the toothpaste back in the tube, but I reserve the right, at least for myself, to boycott this whole shitshow as much as possible.
So yes, I think the discussion of ‘AI’ is suffering from the broad scope, and the brutal stabs are not being defined well enough.
Just a brief exercise in opening a browser page at YouTube and searching for something is increasingly likely to pull up Slop before any genuine results… recently searching for information about an accident at the TT Races pulled up a ton of YouTube results - with absolutely no genuine results at all, beyond brief statements from the news. Sometimes people are genuinely fooled (thinking the ‘AI Voice’ is likely a translation tool - which is a good use of AI) if the video comes out looking half decent.
So the biggest danger to me seems that we’ll get flooded.
The next issue is obviously the character of LLM’s outputting their information with formatting and language that implies confidence and intelligence. This is where I feel uneasy about using LLM’s to help with thinking skills…
FYI…
With yay one can set show menu before install.
Snipped from ~/.config/yay/init.lua
-- yay 13.0 Lua hooks for the AUR security stack.
-- Options
yay.opt.diff_menu = true
yay.opt.edit_menu = true -- prompt to review PKGBUILD before build
yay.opt.clean_menu = true
yay.opt.clean_after = false
yay.opt.sort_by = "votes"
yay.opt.bottom_up = true
-- Warn about AUR packages with PKGBUILD modified < 3 days ago
yay.create_autocmd("UpgradeSelect", {
desc = "flag recently modified AUR upgrades",
callback = function(event)
local cutoff = os.time() - (3 * 24 * 60 * 60)
for _, pkg in ipairs(event.data.upgrades) do
if pkg.repository == "aur" and pkg.last_modified >= cutoff then
local hours = math.floor((os.time() - pkg.last_modified) / 3600)
yay.log.warn(pkg.name .. ": PKGBUILD modified "
.. hours .. "h ago — review carefully")
end
end
return { exclude = {}, skip_menu = false }
end,
})
Line of interrest.
yay.opt.edit_menu = true
My speech…
It’s not a hill to die on; I’ll co-opt that metaphor and reply that it’s a hill I can see far from, a nice place to plant a few shade trees that’ll pay off later.