Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- see also, his : https://hughhowey.com/the-end-of-an-era/by jgord
- "So here’s what I would love: A bookstore for humans by humans. A Humazon.com. Or even just a BookBud recommendation service where the authors are fully vetted (they’re your buds!)."
--------------
To do what the author of this post asks for is incredibly difficult. You can't reliably use a LLM to detect the output of other LLM's. You can't trust user reviews to distinguish AI from human writers because AI can easily produce reviews. You'd basically need to form a central gate-keeping authority that decides who is allowed and who isn't. This would be resource intensive and fraught with problems. (e.g. How many authors would sue if they were blocked as the result of a false positive? How many using LLM's to write their material would brazenly sue for being blocked too?)
It really bothers me that one of the biggest problems with LLM's is how difficult it is to filter out what they produce.
by beloch - This comment was AI-generated, wasn't it?
Turtles all the way down I swear.
by majorchord - This is a remarkably magnanimous and even-tempered post. Most authors (e.g. Charlie Stross) really hate AI, feel personally affronted by it, and lose their cool when they talk about it.
I agree with Howey's sentiments. AI-written books aren't bad in themselves -- they're not ontologically bad -- but they tend to be, by their very nature, derivative. Narrative quality is, as yet, still quite poor. And self-publishing on Amazon.com is full of that stuff. (I've long since stopped buying self-published stuff on Amazon on account of low-quality content, and Amazon's review system seems severely broken.)
Humazon would be a welcome thing. Or even a "human written" flag for books on Amazon.
by A_D_E_P_T - I think AI detection will increasingly be an arms race .. so we might need other solutions, eg :
yes, Im aware the second can be increasingly faked .. eg train a DNN or LLM to mimic brushstrokes and oil paint handling- blockchain attestation [ crypto-graphically signed by human author / artist ] - proof of work in the traditional sense, eg video of artist at work, stages of the painting from sketch to under-painting, brushstrokes, layers, glazingIn my case, I think videos of the painting process are a reasonable, if temporary, solution.
by jgord - > I see nothing wrong with readers gobbling up AI-written books for the rest of their lives.
> Because to me, authenticity makes for a better reading experience.
This pair of sentences gave me pause, until I realised that they're not in contradiction - the author is deliberately non-judgemental of _others_ making a personal choice that _he himself_ would prefer not to make. Remarkably rare, nowadays.
by scubbo - They are in contradiction, because the author starts by saying how much he enjoyed the Stratemeyer Content Mill books (full disclosure - I did too) and then says that if only he'd known...
But he didn't. And it didn't affect his experience until he found out.
Which strongly suggests there was nothing wrong with the reading experience, and no lack of "authenticity" in the text itself, on its own terms.
It was a chimera.
As far as most readers knew, Carolyn Keene, Franklin W. Dixon, and Victor Appleton were all authentically real people.
Discovering they weren't didn't change the quality of the writing - it only changed the implied relationship with the author.
But that's a parasocial experience, not a feature of the words on the page.
I suppose the implication is that because the book went through the mill the author and publisher somehow didn't really mean them.
Perhaps they were tainted by money instead of being products of pure enlightened generous impulse.
But how can you not mean something like a Tom Swift book?
The series was incredibly successful and had real social influence. Would it have been more successful if Victor Appleton wasn't a made-up person?
Why do we expect authors to be scrupulously physical, when creating imaginary people and imaginary worlds is literally the job description?
- I've been researching AI detection (and also made a game to see if humans can quickly detect AI writing just based on style in a short passage: https://prose-or-con.com/ as a way to gather data and to sate my own curiosity about whether I can actually detect AI writing reliably...I can, but not as reliably as I thought).
Storyscope (https://github.com/jenna-russell/storyscope) is a really interesting concept, and one I'm in the process of reproducing and expanding on, even though it is less accurate on direct detection than something like BERT or Binoculars. It is interesting because the usual obfuscation tactics, even having a human transcribe and rephrase the story don't work; it looks at the story itself, rather than word use and grammatical quirks, etc. Plot, agents, temporal structure.
So far, AI writing is detectable to a pretty high degree, though I think it'll be a war of attrition that AI eventually wins. And, all the detectors are the same tools one could use to create undetectable AI writing; AI loves to iterate in a loop, seems like iterating to rewrite to be undetectable is a soluble problem (though models currently probably would end up writing worse and worse to avoid detection, and the current best models have begun integrating watermarking, pushing back the defeat of human writers for some time).
by SwellJoe - This was an interesting test, thank you for posting the link. I also found the AI writing in it harder to spot than I expected, but I suppose I'm mostly tuned in to the low-quality Claude-style slop that's most commonly posted to HN. The coding-oriented models seem to be optimised for feedback-driven code creation at the expense of the writing style (which stinks and is quite easy to spot).by tom_
- There is also this, Wikipedia's AI or not quiz:by MonstraG