Using AI for reading

[Sirtel]

Originally Posted by rantanplan

Don’t anthropromorphize LLMs. They are not intelligent, they don’t have emotions, they don’t panic. They have been made to appear that way, for our benefit, or rather the companies benefit.

I'm pretty sure j.p.s is perfectly aware of this and the comment was meant to be ironic.

[John F]

Originally Posted by Quoth

So you want threads to be fans posting agreement of the Original Poster, even if it has questionable content and you don't want any alternate opinions or actual facts if they contradict the OP or a fan?

...

A nice discussion is fine. Going on, and on, and on, and on, .... in multiple threads becomes tedious.

Especially in the second post of the thread. Did you even answer the OP's question?

That would kill any forum. ...

And so does proselytizing. There is a reason why there is a P&R forum: people get tired of hearing these things repeatedly. People don't want to voice their opinion if it just leads to the same regurgitated opinions/facts (at least some).

You didn't even help a little bit by suggesting alternate resources to solve Barty's (and maybe others) needs. AI may be evil, but if it is the only alternative, or works the best, ...

[Quoth]

Echo chambers and opinion Ghettos are not the answer.

[rowe]

It's a fine sentiment. The irony is that an echo chamber is exactly what gets created when debate is conducted in bad faith.

For instance when the OP simply asked for practical uses of AI in reading the immediate response wasn't a contribution to the topic but a blanket, off-topic dismissal.

When another user suggested that these repeated, broad condemnations might be better in their own dedicated thread their position was misrepresented as them wanting zero alternate opinions and a forum run like a 'dictatorship'. That’s a classic straw man argument, not a discussion.

And when challenged directly resorting to personal attacks, calling another member a sycophant.

That's the behaviour that kills a conversation. It's not the presence of disagreement that creates an echo chamber. It's the refusal to engage honestly with what others are actually saying.

[arjaybe]

Isn't text-to-speech these days using AI to read?

[Quoth]

Originally Posted by arjaybe

Isn't text-to-speech these days using AI to read?

Mostly not, and it's not the same sort of AI as generative AI/ LLMs.

[jhowell]

Earlier this year Amazon added a “Recaps” feature that uses AI to summarize book series up to a certain point so that readers can more easily pick up where they left off after a break. I can see that being useful for series that I read as they come out, often with a year or two gap between books.

For those against AI it is interesting that Amazon states to KDP self-publishers, “Can I opt in or out of Recaps for my books? The Recaps feature is always on to provide readers with a consistent experience across series they read, so there is no option to enroll or unenroll.”

[Dopedangel]

For the last decade, I've been searching on and off for a book I read over 25 years ago when I was about 13 or 14. Every time I saw a post from someone trying to find a book, it would prompt me to search for mine again. My memory of the story was vague, and I was never sure if I was remembering it correctly. Recently, after seeing another such post, I decided to ask one of these LLMs. It found the book from just three sentences.

LLMs are excellent at information retrieval. Their main limitation is the quality of their training data, but as these issues are resolved, the models will continuously improve. So I personally don't understand the hate for it. Now being wary of the changes it will bring to the world I understand as the older we get the more we are stuck in our ways and anything or anyway of doing something differently is wrong.

[DJUNGELSKOG]

If the book in question is a 1000-pages long doorstopper then it’s highly likely that some kind soul has already posted a chapter summary or a reading guide online. I can just refer to that without consulting a hallucinating chatbot.

Otherwise it’s a small 300-pages book and I can just re-read the initial chapters…

And if it’s an e-book then I can just use the search functionality that was present on any device or in any app long before the chatbots became a thing.

[Graham44]

Originally Posted by jhowell

Earlier this year Amazon added a “Recaps” feature that uses AI to summarize book series up to a certain point so that readers can more easily pick up where they left off after a break. I can see that being useful for series that I read as they come out, often with a year or two gap between books.

For those against AI it is interesting that Amazon states to KDP self-publishers, “Can I opt in or out of Recaps for my books? The Recaps feature is always on to provide readers with a consistent experience across series they read, so there is no option to enroll or unenroll.”

So does that mean if you self publish on Amazon they automatically scan your book with AI?

[Barty]

Originally Posted by Quoth

It is inaccurate.

A huge waste of earth's resources creating it and wasting your time. Also using pirated content.

"Hallucinations" is a marketing term to hide the fact that by design it doesn't accurately replay the inputs.

I don't really dispute anything you're saying here

Originally Posted by DiapDealer

Three similar sentences in every thread where the letters A and I appear together. Cumulatively, it's a freaking campaign.

Well I think people are entitled to their causes, as long as they're civil and don't attack people personally. Sort of like Jon's anti-anti-cover crusade. (Not that I'm equating the two.)

[Graham44]

Personally I wouldn't use AI for anything -

To answer your questions/ observations - I very rarely start a book and then put it aside for any period of time, but on the few occasions I have I was shown a neat trick years ago (can't remember by who) to jog your memory about a story (works for text books too), read the first and last sentence of every chapter up to where you stopped, it's quite often enough to jog your memory.

I have read a fair amount of historical fiction and I too would come across descriptions, items etc that I was unfamiliar with but I've always found a web search to be the best solution as they come up with many different descriptions, examples and possibly even pictures which you can quickly cross reference with each other. Having several sources to look at usually helps me to remember something as well.

[Sirtel]

Originally Posted by Graham44

I have read a fair amount of historical fiction and I too would come across descriptions, items etc that I was unfamiliar with but I've always found a web search to be the best solution as they come up with many different descriptions, examples and possibly even pictures which you can quickly cross reference with each other. Having several sources to look at usually helps me to remember something as well.

Yeah, same for me.

I wouldn't want to rely only on an AI for this - how would I know it isn't just making it up? I'd have to search out other sources anyway, to be sure. So it's faster to just search to begin with.

As to the chapter summaries, the same thing applies. How would I know the AI isn't just hallucinating the summaries, or at least parts of them?

As things stand at the moment, LLMs just aren't reliable enough.

[jhowell]

Originally Posted by Graham44

So does that mean if you self publish on Amazon they automatically scan your book with AI?

Yes, at least for books that are part of a series. There is no way to opt out of it.

[DNSB]

Just for the heck of it, I had 10 2 page maximum summaries of the first 11 chapters of 20 books generated. About 50% of them weren't bad, 35% were iffy and 15% had me questioning if I existed in the same universe as the AI.