🍵 
Scubrats, I was hoping to last a few more years before adding a project to my plate to code a Calibre alternative.
Eh, I've been secretly planning an alternative for a while. Because what I needed all along was more coding projects.
@mirabilos @amin @rl_dane I don't see that as invasive now. Normal operations still don't use LLMs.
@amin @rl_dane Check out booklore. https://github.com/booklore-app/booklore
Hm, this does look intriguing enough to test out.
Particularly now that I've jailbroken my Kindle and have KOReader.
It does appear to be coded in Java and TypeScript, though. Which sounds like a pain to self-host (I avoid docker). :P
@amin @thelinuxcast @rl_dane would you elaborate why no docker? no containers in general? (I’m genuinely interested)
@tippfehlr @thelinuxcast @rl_dane
Yeah, containerization just hasn't been very fun for me to play with in general. And I generally find that software designed in such a way that it's easy to self-host without a container is also usually designed much more in line with my preferences and is less bloated. And it's easier to diagnose issues.
With docker specifically my main gripe is that it completely bypasses firewalls and rewrites their rules.
Or just read the developer's response. 📕
Thanks for sharing. As "AI Integrations" go, I give him credit for making it about the least odious that I've seen.
Still not particularly happy with it, but it's a far cry from MS or even Mozilla.
Very much agree with you! - That's why I started looking into the issue.
Opt-in is the best choice IMHO.
The Calibre AI plugins can be disabled in the Plugins section of the Preferences.
@norbu @amin @rl_dane this, and the response from https://polymaths.social/@rl_dane/statuses/01KBV08J9E6B8E8WFN9D7F5PMR…
As "AI Integrations" go, I give him credit for making it about the least odious that I've seen.
… are true.
However, it’s still an explicit endorsement of the TESCREAL bullshit, and I’d prefer to use something that does not even have an optional “AI” integration. (I’m not a Calibre user, have yet to find any need for it, so I’m commenting a bit from the outside.)
On the other hand, here’s a link to a very interesting thread about this from not only the inside, but also from the books and reading and user perspective. This is definitely food for thought, and good to have read in full (follow the link contained), even if one has already a set opinion on so-called “AI” per se and/or this integration, even if one doesn’t even use it:
https://buc.ci/abucci/p/1765053767.148776
The above link is a toot linking to https://wandering.shop/@xgranade/115671289658145064 (a large thread split into many small posts 😾) and excerpting a best-of. Still worth reading through the entire thread.
(Thanks, @abucci and @xgranade and @whitequark )
@mirabilos @amin @rl_dane @abucci @xgranade not sure I agree with much said in the thread, like "fundamentally changing the relation" or "from a tool to read books to ...".
I still use Calibre the same way as ever before, mostly to manage my libraries and load then onto my reading devices.
The permanent whining about AI is strange to me.
@norbu It's not whining, it's caring about art. You seriously don't need to make apologetics or excuses for AI, *especially* not in the replies to someone who's already made their opinions quite clear, along with the basis for those opinions, and why they're important to me as a reader and a aspiring author.
@xgranade art will not be destroyed by a piece of software that gets added another piece of software. You mix up a technology (AI) with a certain product (name your favorite GenAI). It's like "TVs must be abolished because we have FOX news".
@norbu That is a gross, gross misunderstanding, and frankly just plain insulting. It's also incredibly rude to say that to someone who just told you they're an aspiring author.
Making excuses for AI by shitting on art is a great way to get blocked.
Agree with most of what you said, do want to point out that the last bit (firewalls) is not a problem if you use podman as your container platform.
Apologies if you know all that already, just thought it might be nice to know.
@tippfehlr @thelinuxcast @rl_dane
- replies
- 0
- announces
- 0
- likes
- 1
@mirabilos@toot.mirbsd.org @amin@polymaths.social @rl_dane@polymaths.social @xgranade@wandering.shop
@abucci @mirabilos @amin @rl_dane
I am affected, both at work as well as maintainer of oss projects and gsoc mentor. I'm well aware of the pain of LLM slop. But I have also developed enough AI based systems that actually help people.
So AI, but also LLMs, are tools, and I hear far too many rants that are just meaningless broadband accusations that AI is bad and and and.
Those are tools. Like TV. Like guns. If they are used badly, it is bad. But a blanket "AI is bad" simply doesn't cut it.
@norbu @abucci @mirabilos @amin
As my "Intro to Graduate Humanities" professor was always fond of saying, "Il faut préciser."
There are always nuances to every argument, but I think your initial reply smacks a bit of tone policing, to be honest.
Yes, there are a multitude of types of "A.I.," and not every application is bad. But since most "A.I." out there was incredibly unethically sourced, and involved in what will almost certainly be just the next yottacorporate pump-and-dump scheme, I don't understand your frustration with people's outrage.
I am, to be very frank, incredibly frustrated that 99% of the people I encounter simply do not give a shit about how something is created, procured, or used, as long as it makes their day 1% easier.
This is why I say, "Replace 'AI' in the sentence with 'slavery' and see how it reads."
Because ultimately, the carelessness of the consumer is the same with both, and the inhumanity of "capitalism" is the same.
@rl_dane @mirabilos @amin (taking out abucci since they mentioned "I'm out")
> But since most "A.I." out there was incredibly unethically sourced,
Actually NO. LLMs might be unethically sourced, I completely agree with that. But that is not "most AI". There is **far** more AI out there then the LLMs - only it is far less visible and a kick point.
That is the whole of my argument, please say "LLM" if you mean LLM, and not "AI".
If you put "LLM" in there, I am 100% behind what you say.
I think you're being semantic-pedantic, to be honest.
"AI" is a broad term that covers everything from Pacman's ghosts' pathfinding algorithm, to ELIZA, to the Perceptrons, to whatever fresh hell Altman is dreaming up next.
But I'm responding to "AI" in its modern marketing-driven parlance, which is almost exclusively LLMs, which is (AFAIK) the issue with Calibre.
@rl_dane @mirabilos @amin Sorry no, modern AI is much more than LLM. I mentioned a few things, like TDA for stability checks of buildings, deep auto-encoders for bot blocking, classical NN for improving search - these are just three examples where I have worked on - and *NONE* of them uses LLMs.
In fact, I would go as far that more than 90% of all AI in production is **NOT** using LLMs.
The marketing hype for LLMs is there, but production deployments are a different thing.
I see generative AI as, among other things, a tool to further degrade collective reason, dialogue, deliberation, and reflection, which is one of a million reasons to resist it. I think it's especially important to resist generative AI wherever it threatens to degrade literacy, since reading is one of the best ways to we know to improve collective reason et al.
@norbu@mastodon.social @mirabilos@toot.mirbsd.org @amin@polymaths.social
@mirabilos @amin @rl_dane @norbu AI is what people say when the want venture capital, even if no machine learning is involved