Turn off/Block AI features

[John F]

Originally Posted by DiapDealer

The fork will eventually be abandoned and then they'll need to come back. But by all means, use a strictly cosmetic fork of Calibre while it lasts because principles.

I'll be interested to see how long the fork(s) last(s), and how soon the defectors return.

And it will be interesting to see what happens with the DRM removal tools.

[DiapDealer]

Unless needless changes (needless in the sense that they are unrelated to ai support-code removal) are made to the external plugin framework in the fork, the DRM removal plugin should remain completely unaffected.

[John F]

Originally Posted by DiapDealer

Unless needless changes (needless in the sense that they are unrelated to ai support-code removal) are made to the external plugin framework in the fork, the DRM removal plugin should remain completely unaffected.

It has been ~2 weeks since the fork. I don't recall any posts about the fork actually being modified. So the defectors currently aren't stripping DRM?

What I meant was stand alone DRM removal tools will need to be made, or use commercial software, for defectors that don't want to support "calibre" at all.

[DiapDealer]

Originally Posted by John F

It has been ~2 weeks since the fork. I don't recall any posts about the fork actually being modified. So the defectors currently aren't stripping DRM?

What I meant was stand alone DRM removal tools will need to be made, or use commercial software, for defectors that don't want to support "calibre" at all.

There's been several commits to the fork remove the any ai support code, but there's been no release. So unless people are building their own versions of Calibre from the fork (no small feat) I doubt there's been any defectors yet.

I guess I don't know what you mean about needing standalone DRM removal tools. Calibre has no integral DRM removal capabilities. Never has. It is all done through add-on third-party plugins. Which would still work with the proposed fork of Calibre. Continuing to use them wouldn't be "supporting" calibre in any way.

[JSWolf]

Unless the fork is pre-built, I cannot see many using it.

[ownedbycats]

I suspect the fork won't last very long tbh.

[John F]

Originally Posted by DiapDealer

There's been several commits to the fork remove the any ai support code, but there's been no release. So unless people are building their own versions of Calibre from the fork (no small feat) I doubt there's been any defectors yet.

I guess I don't know what you mean about needing standalone DRM removal tools. Calibre has no integral DRM removal capabilities. Never has. It is all done through add-on third-party plugins. Which would still work with the proposed fork of Calibre. Continuing to use them wouldn't be "supporting" calibre in any way.

I assume the plugins only work with calibre. If someone will not use calibre (or a fork) how would they remove DRM. At one time, the "tools" included stand alone AND plugins?

[DiapDealer]

I've not seen any "not going to support calibre at all" sentiments. There's the fork effort and there are those who have said they're not going to update calibre past the last ai-support-code-free version, but that's all I've heard. I suspect the number of people actually willing to give up their primary drm-removal option over this particular tempest in a teacup would be quite low. As in nothing-to-see-here low.

[mitsie]

I don't know if you remember, but I created an AI Plugin collection to send your eBooks to a Local LLM about a year ago. It was a game-changer for compiling summaries and completing your metadata automatically. It also allowed you to pull down custom metadata from books and upscale your low-resolution ebook covers directly in Calibre.

It was all local, and you didn't need an internet connection to use the plugin.

I'm dead against BigTech AI Companies harvesting books and stealing authors' work for their own profits, so I went out of my way to block all the main endpoints for all the cloud-based services, Google, Deepseek, Anthropic, ChatGPT ect.. I encoded the domain blocks and made it difficult to reverse engineer the plugin, so you could only use the Plugins locally on your own machine, and the author's work would be safe.

https://www.digitalassassins.co.uk/n...book-software/

I was shunned at the time by the Calibre community, banned from the Sub-Reddit for daring to introduce AI into Calibre, and blocked from uploading my AI plugin into the official Calibre Store..

Kovid Goyal at the time said AI will never make it into Calibre and argued with me, and I quote "A.I. has the IQ of a 3 day old Corpse"..

He said AI is pointless and is stupid; it doesn't deserve to be in Calibre.

9 Months later, he starts adding AI tools that communicate with BigTech AI services to send authors' work from your machine, to be stored on cloud servers for AI Companies to make money from.. Such as Google Gemini and OpenRouter.AI

It's kind of Ironic really !!

[Sirtel]

Originally Posted by mitsie

9 Months later, he starts adding AI tools that communicate with BigTech AI services to send authors' work from your machine, to be stored on cloud servers for AI Companies to make money from.. Such as Google Gemini and OpenRouter.AI

AFAIK no books are sent to any AI - just the title and the author. I tried to get an AI to summarize a virtually unknown fan fiction collection in my library. It told me it couldn't find any information about it. This means the AI is not getting the book itself, just the title/author information.

[mitsie]

Originally Posted by Sirtel

AFAIK no books are sent to any AI - just the title and the author. I tried to get an AI to summarize a virtually unknown fan fiction collection in my library. It told me it couldn't find any information about it. This means the AI is not getting the book itself, just the title/author information.

I thought the internal quotation in Calibre reader allowed you to select a passage from the book, as much as you like, to send to the AI? This is sending part of the "authors work" with the title and author to a cloud server for an AI to store as part of its training data..

There was a News Article where the financial officer of a Fortune 500 company used ChatGPT to work out the companies finances, he only uploaded a snippet of the financial report.. another researcher typed into Chat.. "What is the financial earnings of Company X" ChatGPT was able to regurgitate that financial report information to the researcher. Word for word.

If enough people upload big enough snippets of a book, that is uploading the authors work to those companies.. especially if the AI can piece those snippets back together from Memory..

[Sirtel]

Originally Posted by mitsie

I thought the internal quotation in Calibre reader allowed you to select a passage from the book, as much as you like, to send to the AI?

If you mean copy and paste in the conversation window, you can do this without any AI/calibre integration anyway. You don't even need calibre for that. Just select a passage in your reading app, copy it, go to an AI provider, paste and send.

[mitsie]

Originally Posted by Sirtel

If you mean copy and paste in the conversation window, you can do this without any AI/calibre integration anyway. You don't even need calibre for that. Just select a passage in your reading app, copy it, go to an AI provider, paste and send.

Yeah well, that's why I made the plugin I designed local, so you didn't have to send that information to BigTech companies, even a snippet.

These AI updates are so far behind the times anyway, Kovid may aswell of saved his time and effort on them. With Claude or OpenClaw ect. it's actually getting quite scary what you can do.. Now you don't need an API or to design plugins for outdated software with clunky GUIs anymore..

The AIs are now gaining the ability to control your computer or Docker container including the GUI themselves. No need for API keys or building custom plugins anymore. You can literally say to an AI Agent via Telegram Message:

"Find me the top 10 Vampire books in my collection and load them onto my Kindle, provide summaries and artwork to the ebooks if they don't have them"

Watch on your computer as the AI firest up the Chrome Browser, searches "Top 10 Vampire Novels" makes a list, opens up Calibre itself, checks the metadata for all the Vampire books in your collection. Downloads metadata and book covers then cross references everything to make sure it's correct. Any books it can't find a summary for, it opens the ebook and reads the full thing and types out a summary for you.. Embeds the metadata in the ebooks, then fires the books to your Kindle device, all while using the UI exactly like a Human would.

AI integrations are going to be pointless soon, as the AI can do exactly what you can do on a computer, automated..

In a few years Calibre will obsolete anyway, once AI bots have control over operating systems and the ability to code faster than a team of thousands of humans, it will build it's own eBook management system and do everything calibre can do and more..

Many software packages will fall by the wayside, why do you need a Human Interface to manage your ebooks, when AI can sort your ebook library out quicker and without human intervention with its own AI built software it just chucked together in seconds running under the hood?

[un_pogaz]

If the AI improves. Which is not very the case now, we start to reached the limits of the model.

We can given AI more power and control, which creates an illusion of progress, but this does not change the fact that they are merely trained monkeys, dangerous without supervision.
This leads to a paradox: the people best suited to use AI are those who are already competent in their field, and in this case, supervising AI is a waste of time and energy.

And if the bubble doesn't burst (it's starting to get tense.).

[mitsie]

With computer games, the computational power needed for the next generation of games requires more computational power, this is achieved by advancements in both technology and energy efficiency.

Just like to render more mathematical equations in games, we need more computational power to improve AI. The improvement in computer games, drove the market to provide more raw processing power, those Graphics card power and efficiency improvements are what drove the AI boom.

Saying providing more power to AI is a paradox, but currently Enterprise AI has nowhere near the processing power of the human brain, AI is restricted by our inability to provide the processing power. In the same way computer game resolution will not match the DPI of our eyes due to restrictions in the rasterisation processing power needed to provide a rendered image at 32k resolution.

It's not a paradox, just a limitation.. Higher parameter models of AI require higher VRAM and processing power to run, home users are limited by their Hardware.. currently cloud based AI is far superior, but only a matter of time until home users have the same Graphics card power as Enterprise users and the game changes..

Jobs are being replaced on mass, which drives more investment in AI technology.. some jobs you only need a supervised monkey to operate a computer ��