New Kindle feature uses AI to answer questions about books | Hacker News Reader

New Kindle feature uses AI to answer questions about books

reactormag.com

84 pointsmindracer130 comments

If they're not using the book text to train models (keeping the focus on this particular new Kindle feature), where's the room for objection? My device, my content, it's none of the author's business how I read it, in my view.

Edit: Given I've been a reader of HN for some time, I am perfectly aware that on Kindle you don't own the content, just a license to the content. Don't need any more people pointing this out! Lol. In my house we still call owning a license to something that is not likely to be revoked "owning it".

"Amazon DID NOT answer PubLunch’s questions about “what rights the company was relying upon to execute the new feature was not answered, nor did they elaborate on the technical details of the service and any protections involved (whether to prevent against hallucinations, or to protect the text from AI training).”

> what rights the company was relying upon to execute the new feature

what rights does a bookstore clerk need to answer questions about a product on the store's shelves? what a presumptuous question

Yeah, the "but what about a human" argument doesn't really work here. Scale of data matters as always. And an Ai for kindle has the scale of 20 years of literature (and more if they just scrape the internet).

> Yeah, the "but what about a human" argument doesn't really work here. Scale of data matters as always. And an Ai for kindle has the scale of 20 years of literature

I haven't seen a convincing argument why not. There's millions of librarians with the knowledge of more than 20 years of literature under their belt. Why can they answer your questions about a book but the robot can't?

> Why can they answer your questions about a book but the robot can't

Robots simply do not deserve the same consideration and the same rights that humans have

It's really that easy. Humans deserve more rights than inanimate objects

Luckily we do not live in an allow-list based society where we need to ask permission for every new thing we invent. The burden is on someone to show that robot answers book questions is somehow bad, to justify outlawing it. And that has not been shown. Bringing up the ontology of humans having human rights has nothing to do with the argument at hand.

That way it should be illegal or discouraged to select text from a book and paste it elsewhere

Is the "clerk" scanning the books an digitizing them to generate other products using an LLM under the guise of "Answering Questions?" I believe this is the question being asked.

Companies like Amazon and Google have some really sticky fingers when it comes to intellectual property and personal data. I think it's worth asking these questions and holding them accountable for exploiting data that doesn't rightly belong to them.

> Is the "clerk" scanning the books an digitizing them to generate other products using an LLM under the guise of "Answering Questions?" I believe this is the question being asked.

That's what I mean by presumptuous. If that's really what they want the answer to, and what they object to, they should ask it plainly instead of alluding to it by asserting that there's some requisite but missing entitlement for the feature to exist on its face.

Either the Clerk would have read it, because they bought it, or borrowed it from the library.

I mean they could have read it on company time as well.

However, let us not use a straw man here. Its not some company clerk, its one of the largest company on earth using other people's copy right to make more money for them selves.

The author also gets a cut of this, no? It is the author's prerogative to sell their books to be read on a Kindle and they get compensated, maybe perhaps unfairly, when I choose to buy the book. Whatever happens after that, other then copying it and sticking it on Anna's archive is basically free game as long as I'm making derivative works and making money off them. Anything short of that, I'm good.

That's my thoughts on that, anyway.

You don’t need any rights to execute the feature. The user owns the book. The app lets the user feed the book into an LLM, as is absolutely their right, and asks questions.

>The app lets the user feed the book into an LLM, as is absolutely their right,

I don't think that's cut and clear yet. Throwing media onto someone else's server may count as distribution.

Sure, in the sense that any belief about the law isn’t cut and dried until a judge has explicitly dismissed it in the court of law.

How likely do you think it is that Amazon doesn’t have a pre-existing contract with these publishers to host these books on Amazon servers?

1. The user doesn't own the book, the user has a revocable license to the book. Amazon has no qualms about taking away books that people have bought

2. I doubt the Kindle version of the LLM will run locally. Is Amazon repurposing the author-provided files, or will the users' device upload the text of the book?

I am so confused by some of the comments in this thread. All these weird mental gymnastics to argue that users should have less rights.

“Oh, you think you should be able to use an LLM with a book you paid for? Well you don’t own and book.”

Ok, and you like that? You want even less ownership? Less control?

> All these weird mental gymnastics to argue that users should have less rights

We probably agree more than not. But users getting more rights isn’t universally good. To finish an argument, one must consider the externalities involved.

I don't agree with the way you're interpreting the comment. If anything I think it's BAD that you don't really "own" digital content.

I guess my argument is that Amazon shouldn't be able to have their cake and eat it too

You agree that we should own our digital content but it sounds like you don’t want this particular capability because… fuck Amazon.

I can totally understand that sentiment but I don’t think giving up end user capabilities to spite Amazon is logically aligned with wanting ownership of digital media.

> protect the text from AI training

Hasn't training been already ruled to be fair use in the recent lawsuits against Meta, Antrhopic? Ruled that works must be legally acquired, yes, but training was fair use.

>none of the author's business how I read it, in my view.

my favorite way to eat is give other people my food, and have them tell me how it tastes and what not being hungry feels like.

or to labor the point for the people that are having LLMs do their reading for them. Watching golf isn't playing golf.

Once you've bought that food and it's on your plate, how would you feel about the farmer who grew it coming up and forcing you to eat it with a specific fork or only using approved utensils?

You bought a kindle, they already did that to you.

It's not training on books, but it will answer questions about the book you're reading. Doesn't pass the sniff test.

>My device, my content

I don't think you own the kindle store and servers used to train the Ai.

Are you implying that an LLM needs to be trained on a specific piece of text to answer questions about it?

If you want proper answers, yes. If you want to rely on whatever reddit or tiktok says about the book, then I guess at that point you're fine with hallucinations and others doing the thinking for you anyway. Hence the issues brought up in the article.

I wouldn't trust an LLM for anything more than the most basic questions of it didn't actually have text to cite.

Having access to the text and being trained on the text are two different things.

Luckily, the LLM has the text to cite, it can be passed in at inference time, which is legally distinct from training on the data.

> It's not training on books, but it will answer questions about the book you're reading. Doesn't pass the sniff test.

What do you mean? Presumably the implication is that it will essentially read the book (or search through it) in order to answer questions about it. An LLM can of course summarize text that's not in its training set.

"Reads the book" is the issue, yes. It's possible they aren't training. Vit to be frank, we're long past the BOTD where tech companies aren't going to attempt to traon on every little thing fed into their servers.

Happy to be proven wrong, though.

There are LLM's that can process 1 million token context window. Amazon Nova 2 for one, even though it's definitely not the highest quality model. You just put whole book in context and make LLM answer questions about it. And given the fact that domain is pretty limited, you can just store KV cache for most popular books on SSD, eliminating quite a bit of cost.

You could also fill the context with just the book portion that you've read. That'd be a sure-fire way to fulfill Amazon's "spoiler-free" promise.

> My device, my content

I am quite sure Amazon doesn't sell you that.

>My device, my content

Afaik, while the device is yours, everything else on it isn't.

device is hardly yours unless you jailbreak it or collect bricks

I wish it was "my device, my content" but it absolutely isn't. If you want that you have to buy from a DRM-free source, and Kindle is the absolute opposite of that.

What does this have to do with the parent's comment?

Okay it's not 100% my device my content, so I shouldn't be allowed to run a local AI against the text?

IMHO you should be able to enjoy your books however you want. If you want to run a local AI against it, more power to you.

But my opinion doesn't matter. Only Amazon's does. That's the point I was making. The premise of "my device, my content" is flawed (because of the DRM Amazon uses) and undermines the argument.

Right, under that argument it's their content, their rules then - making this situation even more of a non issue because they're adding this feature themselves.

You don't mind having an llm owned by a megacorp lecturing you about the meaning of a book ?

"Yes this is a good question about 1984 by George Orwell, you could indeed be tempted to compare the events of this book with current authoritarianism and surveillance but I can assure you this book is a pure work of fiction and at best can only be compared to evil states such as China and Russia, rest assured that as a US citizen you are Free"

Not your content, it's Amazon's content, you only purchased a license to view it, which can be revoked at any time if daddy Jeff is not happy.

And I am not being cynical. That is literally what is on their web page, e.g.

https://www.amazon.com/dp/B0BTZT9PLM

Fun fact: the first book Amazon remotely removed from Kindles was… 1984.

https://www.theguardian.com/technology/2009/jul/17/amazon-ki...

Life mimics art.

The name Kindle suggests Fahrenheit 451. We're going to destroy books and here's the kindling.

Sure. But you knew what this comment was trying to say. It is obviously saying that what happens on the Kindle is between the customer and possibly Amazon, specifically that authors should not be involved. They got their money. That part of the transaction is complete. I know you realize this, it's annoying to read the constant "not your keys not your coins" reframe.

No. The author incorrectly thinks they "own" the "content" like with a physical book, which is the prerequisite for all the discussions following it. I pointed out, factually and correctly, that they don't own anything (other than the license) or have any control over anything.

Most of it is _not_ Amazon’s content. They don’t own the book, so they can’t sell you the book. Nemo dat.

> My device, my content, it's none of the author's business how I read it, in my view.

In practice, that's not the case though, e.g. publishers on Kindle can choose not to allow text-to-speech assistive functionality.

Audiobook publishers require/request this when you sell subsidiary rights. We’ve been able to push back citing accessibility concerns. I find it really annoying when not available for my own reading.

Couldn’t agree more. This is actually a super useful feature. I can’t think of how many times I’ve been reading a book and some minor character resurfaces and I’m like, who the hell is that guy? Now I can know. I can also get information on historical context. Who knows, maybe I can finally read Ulysses without having to have 5 other books.

In the case of a novel, or even certain text books, the author relies on the reader not jumping ahead. Especially murder mysteries and those kind of genres. There are artistic reasons for that, and it can wreck the work.

In my experience, AI summaries often miss points or misrepresent work. There is a human element to reading a well written novel. An AI will miss some of the subtleties and references.

But if I want to jump ahead and read the last page of a book first, is it reasonable for an author to tell me I can't do that?

From an artistic point of view, yes it is. It's a bit like doing a crossword with the answers in big letters next to it... It destroys the point.

I agree but for some reason, there are people who enjoy doing that. I think they should be allowed to do as they like.

In any case, Amazon claims this feature is spoiler-free and that would be easy to implement. It likely works by feeding the book into an LLM context, and they could simply feed in the portion you've already read.

Amazon is selling digital copies (or licenses, if you like) of the books, which means they need permission from the copyright holders. This permission is likely backed by a contractual agreement that covers some details about how Amazon presents the digital copies to the end users.

(This of course wouldn't be the case if they were reselling physical books.)

So what part of this presentation agreement could possibly apply?

> Edit: Given I've been a reader of HN for some time, I am perfectly aware that on Kindle you don't own the content, just a license to the content. Don't need any more people pointing this out! Lol. In my house we still call owning a license to something that is not likely to be revoked "owning it".

The amount of people completely - and likely intentionally - missing your point is both frustrating and completely unsurprising.

A quick reminder that this is part of HN's guidelines

> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.

> where's the room for objection?

I suspect most of the people arguing this way would be in favor of more end user rights if we were talking about anything except the right to use AI.

“Rights good, AI bad” somehow leads to the insane argument that it’s a good thing you don’t have rights over the book you bought.

“You don’t really own the book” is a crazy argument unless the person saying this wants the locked-down DRM world where you can’t own a piece of media.

This sounds useful for when you forget something that happened chapters earlier or when you space out and need to figure out what's happening. This feature should work for the user, author's shouldn't be able to deprive me of this tool.

Good lord, at this point just drone off in front of a Netflix show. How bad has it gotten that you even suggest that one can "forget what happened chapters earlier" ? This is not normal.

What kind of books are you reading? You're telling on yourself (and very arrogant about it).

My thought exactly. Not all books are the same, and I'm willing to bet that GP is not reading the same books that I am, and not with the same goals.

Even Pynchon's lightest work yet, the newly released Shadow Ticket, has me immediately reading it a 2nd time for many details I missed the first time through. Imagine the arrogance of shaming a Pynchon reader assuming they should never need to check notes or refer back to earlier pages.

It's more telling about the current state of affairs than the person who commented. Forgetting things is part of life, move on, we don't need daddy bezos sucking 1.21 gigawatts per request to tell you that some side character drunk a beer 12 chapters ago so you can enjoy the joke you just missed.

I don't mind bezos using 1.21 gigawatts per request, as long as it's only for a very short time.

Brutal on the crest factor though, you'll definitely get a snotty phone call from your power company if you keep that up.

Maybe. Depends on how many capacitors you have.

This is hardly that strange, life gets in the way for many of us. I too have many times wished for an easy way to recap a book I've had to put down for a week or two - this is by no means an endorsement of how Amazon have done it here, but you are making incredibly arrogant assumptions about how others enjoy books.

That's been happening to me since before Netflix licked their first envelope. Have some sympathy for people born during memory shortages!

https://www.amazon.com/Introduction-Quantum-Mechanics-David-...

I'm an avid reader. I'm reading The Silmarillion right now. There have been countless times where a short summary of an area/character/etc has been helpful. Luckily, in this particular case, there are very good Tolkien fan dictionaries that serve well.

As another example, I read the Aubrey-Maturin series earlier this year. Many times I would have liked a quick summary of a previous voyage or of a political plotline or something.

Don't be so judgemental.

I'm an avid reader and I can assure you that it's very normal.

We need to baby sit homoconsomators every step of the way or they get scared and confused

Forgetting what has happened earlier in a book you put down is very normal. Have you met your fellow humans?

If you have ample free time and few commitments and/or you read very short pop fiction, I could see how you might believe this. But there's a vast world of very long and dense literature, and also...people have kids and a life that gets in the way. Combine the two and...well, I can see why this feature would be useful. :)

Fantasy series seem like they've gotten longer and longer and it's often years between volumes. Many authors have started doing recaps of their previous books at the start of later volumes, but not all.

I could see this being useful for that.

I use LLMs for that all the time. Most frontier models have books trained in, so I just ask for a spoiler free recap or ask about certain characters. Works well in my experience, and made jumping back into Wheel of Time a lot easier

I want a character chart, social graph.

I can also imagine a character interaction graph, animated by chapter.

Oh, and pronunciation. The Sun Eater series is eloquent, but the names are inscrutable without having heard a few of the audio books.

> when you space out and need to figure out what's happening

Ok it's not just me that gets to the end of a page and it's like the page didn't exist.

On the other hand the times I use the search function on the ereader most are when I stumble across a continuity error. It would be interesting if a story-reading AI can be used to detect those. Not that I want there to be less human editing in books, if anything we seem to need more.

Or were away from the book for a while and are coming back to it. I've read 1000 page books that I just got tired of reading, so put 'em down for a bit to read something else. Anathem by Neal Stephenson comes to mind.

> To ensure a consistent reading experience, the feature is always on, and there is no option for authors or publishers to opt titles out.

> It also sounds as though authors and publishers were, for the most part, not notified of this feature’s existence.

This is perfectly reasonable fair use.

I'm starting to realize that a lot of content creators either don't understand fair use, or otherwise are unreasonable control freaks.

> This is perfectly reasonable fair use.

Well thats questionable actually

Sure the _result_ is transformative, but it had to consume the content in the first place to make a transformative. (grey area).

Yes, you _could_ argue that its a plain review, but you need to prove that its actually reviewing it rather than just quoting. But as its the machine doing it, that further muddies the water. Is it the end user whos generating the review? does the kindle license actually allow them to do that?

However, the other thing to note is that there is a contract between publishers and amazon that go over and above copyright. It will say how and where works can be distributed and how they can be processed. For example you're not able to distribute the book and then create your own audiobook version of it.

Please re-read my statement about being a control freak.

This is the lawsuit against the Diamond Rio all over again.

No I get that bit. But give how much amazon are keen to take the piss, I can see why they might indulge in such specific control freakery.

Afterall, I'd do the same thing, I don't trust amazon(the media part) with my IP

I’m looking forward to this. Especially reading old classics, or catching up on an old series and trying to figure out “is this character the sister or niece of the main protagonist? Outline their character development”

I used to have to read fan wikis to figure this out.

But it will especially be useful for all the textbooks I’ve bought years ago. Being able to ask it questions (to the content itself) is better than asking ChatGPT or Gemini because they don’t have the content (they’re summarizing summaries found on the web)

> I used to have to read fan wikis to figure this out.

I would much rather read a fan wiki than hope a LLM correctly understood a book's plot, at least with the current state-of-the-art of things.

Case in point: Amazon's own AI gets significant details of its own prestige TV show wrong:

https://gizmodo.com/fallout-ai-recap-prime-video-amazon-2000...

The Fallout fan wiki probably at least knows the Great War was in 2077.

There's also a few plugins for KOReader that achieve the same:

- https://github.com/omer-faruq/assistant.koplugin, which is forked from:

- https://github.com/drewbaumann/AskGPT

The first one even has prompts for quick recaps, summarize, translations, and more.

assistant.koplugin is pretty good. I've been using it for a while now.

Onyx has an AI feature in their boox which came very surprisingly with a firmware update a couple of month ago. It does the same thing - however: this needs constant access to the Internet to work and with boox you can switch it off. The article states that on Amazon apps/devices this will be perma-on and I'm sure that will do wonders for your data plan, particularly on phones. /s

It's super-annoying that the article begins with a photo of a Kindle e-reader, and it's only once you read the last sentence that you find:

  "Ask this Book is currently only available in the Kindle iOS app in the US, but Amazon says it “will come to Kindle devices and Android OS next year."

I was imagining a feature that allowed me to search across all my books, which is something that O’Reilly Learning does (actually it gives you answers from their entire range since their model is a license to access all content).

Come to think of it, given how early O’Reilly had this it’s shocking to me that Amazon hasn’t done this sooner.

The O’Reilly Learning search was simultaneously the best and worst of all the early LLM applications. They have tons of high quality content that underpins very useful answers. I’ve also found a bunch of worthwhile books by looking through the sources.

It’s the worst because their template response is extremely unimaginative. I can be asking process questions about managing tech debt and it still gives me a code sample with every response as though I was asking “how do I add this button to my app”.

I really don't understand why authors believe they have something to say about how I read their book.

They can't force you do anything, but a book or a piece of music etc, is often designed in a particular order so you get a particular effect.

You don't look up the end of a whodunnit before reading the beginning because that would make it kind of pointless.

And yet, some people enjoy doing that. I have no idea why but I think they should be allowed to do it.

In any case, Amazon claims this is spoiler-free, which would be easy to implement by feeding only the portion you've read into the LLM context.

There are many things that people are allowed to do which are not advisable. All a matter of degree I suppose.

Who asked for this? I thought Amazon was all about customer obsession, and I'm having a hard time imagining readers saying "You know? This book would SEND ME if it had a chat assistant."

Seems like a great feature. What I’d really like is a “recap for me till here” for books I started reading then stopped for whatever reason. I was reading Unsong for a bit (great book, very enjoyable) and then lately the baby has wanted a lot more attention so I didn’t get much reading done. I just want to catch up quick so I can continue.

LLMs are great for this, for the plot and character questions, etc.

Authors have nothing to do with it. It’s my device, my book that I bought. It would be like if YouTube banned a screen reader. These are at two different levels of the stack.

> LLMs are great for this, for the plot and character questions, etc.

The article links to a clear, direct counterexample of this claim. By Amazon, even.

https://gizmodo.com/fallout-ai-recap-prime-video-amazon-2000...

It's not that direct a counterexample. We have no idea what underlying data from the Fallout show they gave to the model to summarize. Surely it wasn't the scripts of the episodes. The nature of the error makes me think it might have been given stills of the show to analyze visually. In this case we know it is the text of the book.

> It's not that direct a counterexample.

Amazon made a video with AI summarizing their own show, and got it broadly wrong. Why would we expect their book analysis to be dramatically better - especially as far fewer human eyes are presumably on the summaries of some random book that sold 500 copies than official marketing pushes for the Fallout show.

For the reason I gave in my answer: it would be answering based on the text of the book. I don't expect it to be particularly great regardless because these features always use cheap models.

> For the reason I gave in my answer: it would be answering based on the text of the book.

Why would that not also be true for the Fallout season one recap video?

Did you read a word past the part of my answer you quoted?

Of course. I’m just not sure how “the Kindle feature’s cheap models are gonna be even worse at the task” helps your point.

Because text analysis is substantially easier than video analysis?

Amazon has the Fallout scripts, subtitles, internal show bibles, etc. all available to them.

I’ve found the complete opposite as recently as last week. When I ask a deep question about a book it will hallucinate whole paragraphs of bogus justification and even invent characters

Did you give it the text of the book and tell it to answer based on that?

Does this feature put the entire text of the book into the context?

I hope not, it probably only goes up to the page you are on in order to avoid spoiling the later content.

The LLM just working on its own is just generative intelligence. You have to ground it if you want the real stuff. The Kindle app has the text of the book and I'd want it to put that in the LLM context.

The entire book in the context at once?

That's probably 100k-150k tokens for most novels.

That’s more reasonable than I thought.

I occasionally buy DRM-free ebooks (tech books like O'Reilly) just to put it into NotebookLM, Claude etc.

While I feel a certain amount of empathy to the authors, it's a table stake at this point to be honest.

My kids have book reports and stuff. Lately I can use AI to generate non-trivial questions about the books and use it to quiz them without me knowing anything about the books. Been super useful.

How do you know they're non-trivial if you're not familiar with the material? Can you provide some examples? How long have you been doing it? Do your kids like it?

Just curious, not trying to attack you :)

As an author, this sounds amazing, if I can also get a look at the things people ask.

Especially if I can get some sort of stats on the questions.

Like, if there were a lot of questions about when a character did something, then I know I wrote that badly.

Or if people talk about a set of characters, then I know that those characters made an impact.

Or if no-one asks about the book itself, but about some plot point or worldbuilding idea.

In general, if I, the author, can get a peek at this data too, it would be of immense value to me (and my publisher).

That's a privacy nightmare. I absolutely would not want authors to have access to this information.

Google AI does a pretty good job of this already:

> I'm on page 750 of Anathem. Please give me a recap.

> You are currently reading the section of the book where the main characters have been launched into orbit aboard a repurposed military rocket and are preparing to board the alien starship, the Daban Urnud.

more recap details follow....

Given different printings and formats for books, I’d be very surprised if asking about a specific page number works reliably at all across books. I don’t even know if epub has page numbers embedded to keep track because the number of words on a page of an ereader is entirely arbitrary. My wife has her kindle in grandma mode or something. Only about 50 words fit on a page.

I would expect much more reliable results from chapter numbers though.

So...Are all Amazon books available on Kindle ? So...All books are content for the LLM behind it, I suppose ?

Welp. Seems perfect for a poison data effort !

Have you read any of the kindle-only erotic slop-smut that's going these days? The poisoning is well underway.

100% required on all Pynchon novels that's for sure.

ITT: people who really hate reading

Yes. I cannot work out who the intended audience for this feature is supposed to be.

School children and lazy people who can't be bothered to read properly.

Nah just tech fanatics stuck in 7 layers of bubbles who spend too much time in front of screens and not enough time with people in the real world

Reply on news.ycombinator.com