Live data from Hacker News

I let Claude Code write an entire book

github.com

61–70 of 84 posts

Re: I let Claude Code write an entire book

#61
post #15
post #5

Earlier quoted context omitted.

The actual author is Claude, and Claude has read a lot of books.

It is interesting to me that your emphasis is on _lot_, and not on _read_.

Why?

I emphasized "lot" because presumably the training data for a high-end model like Claude includes literally millions of books, far more than any human being has read even a few pages of.

To me, "Claude has read a lot of books" would read pretty oddly here. I suppose the idea would be to contrast with the (hopefully few) books it's written to date, but I can't think why the emphasis would be called for.

Re: I let Claude Code write an entire book

#62
post #16

Why would I read a book that a person couldn't be bothered to write?

Why would you choose not to read a book based on how it was written rather than the content? If it came out that Stephen king had been using AI for decades would that make his work any worse?

Actually LLM exists in the first place because creative people like him wrote down their thoughts. If no human ever wrote anything, LLMs wouldn’t be able to generate anything would they. Now because LLM are rehashing on human content the feeding value for us is less important. What would be nice is an LLM that writes a book that spans accros authors of different fields to aggregate and consolidate knowledge, saving us the time of reading all the books

Re: I let Claude Code write an entire book

#63

The premise is wrong. Humans don't listen to one another and nod unthinkingly. They criticize and validate relentlessly. Friends aren't mistaken for oracles. We have learned to not trust one another, except when one is speaking from deep experience and expertise in a given domain. AI is presented as an expert in every domain though, so we are lulled into a vulnerable state of unvigilance.

I really found the story in chapter 14 (recursive self-improvement) about the guy who got so addicted to self-improvement that he ended up in his own meta-reality unable to understand even himself because he was getting so much better and hacking his learning. A completely fabricated story with no basis in reality that I'm aware of but man there are a lot of bullet points to make it seem factual. What are we going to…

Honestly there are some interesting concepts and broad overviews of them but this is hardly a "book" but just a verbose LLM document that briefly lists a lot of concepts without sufficiently or consistently fleshing them out into actual meaningful chapters. Not to say that this sort of thing isn't potentially useful but it seems more like the starting point of an outline of a book rather than anything resembling a finished published book.

Re: I let Claude Code write an entire book

#64
post #16

Why would I read a book that a person couldn't be bothered to write?

because that's a lazy dismissive take on LLM usage. If you drive someone to airport, do you just say your cat have them a ride and say you had no part in it? Or do you just say you gave them a ride because of the time and effort it still took you, even though you didn't pick them up and throw them over your shoulder? Would you ride a self-driving Waymo car to the airport?

The problem with LLM generated writing is that, apart from a couple of tells, which high school and college students have figured out how to ask ChatGPT to use diction that befitting a highschool student, to avoid tells like "delve", that you can't reliably detect when something is basically entirely LLM generated vs half human and half LLM generated, vs entirely LLM generated. And if you can't actually tell that it's been generated, that you're instead trying to look for tells that it comes from an LLM instead of the content or the message itself, then whybare you even reading it?

If instead of setting it up to run entirely on its own, as this post did, you give it a scenario, writing a fiction book with ChatGPT is a fun way to spend a bit of time that's (imo) better than doom scrolling for the same amount of time. Give it a scenario and some theme and tell it you want to write a book about it, and have it ask you questions on where the book should go and then make a book that goes how you want. Want a utopian pollyanna view of the future? what a nitty gritty future that makes skynet look like paradise? Want aliens to visit? Want ChatGPT to give you an act three surprise that isn't a trope you expected? Whatever you want, it's just fun to play with (unless you just hate LLMs and can't have fun).

The question is, what do you do with this book that's now been written. If you had fun by yourself and don't share with anybody, was fun still had? If you only share the book with your LLM adopting book writing club, and you all take turns doing analysis of each others books, knowing they were helped by an LLM, does it still "count"? And what if you submit it, or not, to a publisher who accepts it, you get it posted to Kindle Unlimited, and you get a lot of readers? What then?

The very nature of entertainment is changing, from mass media, to personal media. Culture was already fragmenting, AI will only serve to divide us further apart from one another. Between AI for writing and images, as well as video, along with AI like suno for music, the only challenge we need face is the problem of connecting with other people when there's no shared cultural references.

If you and I have both read an loved a book/enjoyed a song/joined a movie/tv show's fandom, there's a basis for continued conversation. But other than adversity like addiction or a trip into the desert/mountain/Serengeti, soon we'll have even less to connect with our fellow humans over.

(and yes, I know there are a lot words here. I wrote this all by hand and didn't have time to shorten it)

Re: I let Claude Code write an entire book

#65

Earlier quoted context omitted.

Just like with mass-manufactured furniture, some of us seek hand-made items, as it feels more human, despite its flaws.

Realy? Does having flawws really make four better reading? Okay, I'll admit that hurt me to right (as did that) but writing isn't furniture, and other than a couple of tells which I haven't kept pace with (eg use of the word "delve"), the problem with trying to key off of LLM generated content and decide quality, is that you can't tell if the LLM operator took three minutes to copy and pasted the whole thing (unless…

I, personally, like mistakes in writing (as with in painting or singing) - I feel that it gives the art an additional depth, context, detail and comparison with author earlier/later works, other authors.

I believe that art function is to communicate - we create art, type letters, paint graffiti, verbal-vomit in online game PvP match to make a connection with other people.

So the mistakes are only adding to the art: "cooking this is difficult, and everyone do mistakes, but it's made with love and intuition, not blind recipe". Well, I can continue with examples of kissing but I guess I am repeating myself, haha.

I believe that being perfect is not human, and life doesn't have to be perfect. Getting better is great! But so is making mistakes.

(Or, dunno, maybe I have more to learn and I will some day think in a different way.)

Re: I let Claude Code write an entire book

#66
post #16

Why would I read a book that a person couldn't be bothered to write?

Behind this rhetorically articulated question, there is a simplistic view of content generation and consumption. First off, if you use LLMs, you are _constantly_ reading stuff nobody else bothered to answer you. So, it's not about who speaks, it's about the what. Then, what you deem as written by nobody turns out (not without irony) an average of so many (possibly) good writings and thoughts, gleaned from various corners of the web. In this regard, I particularly like--and want to shout out--the author's clarity on the work's attribution. It's not his. It is Claude's, hence, everybody's (in a sense). Finally, as a mere form of compressed human written productions, LLMs don't have agency over what they generate. So the prompting, the idea, as well as some editorial decisions are still attributed to the person behind this, hence making it unique in its own way. Instead of seeing as a piece that he didn't bother write, I see it as piece he chose to edit summoning the ghosts of every writer who ever put something on internet, which again, he correctly credits (to the limits of knowing what exact data Claude uses... which is another story).

Re: I let Claude Code write an entire book

#67
post #16

Why would I read a book that a person couldn't be bothered to write?

I find I'm willing to explore a minecraft world or puzzle through a nethack dungeon that nobody bothered to create. You could argue that humans made the biomes or defined the layout constraints, but humans also supplied the training data for an LLM. I guess it comes down to whether the art is any good, with procedural generation being mostly irrelevant? But perhaps a book is different

I mean the difference with Minecraft is each part that is procedurally generated was made by a human or involved human input into the design decisions.

Unless you are suggesting Notch was a generative AI model he made Minecraft.

Re: I let Claude Code write an entire book

#68

Earlier quoted context omitted.

Did you read more than the opening lines? When you've read it in its entirety, could you indicate on a scale from 1 to 10 what score it would get compared to published books you've read (including of course all the best and the worst ones)?

This is by definition slop. A mathematical average of training data. By design it contains nothing novel.

A small correction: it can contain novelty in the form of AI hallucinations.

Re: I let Claude Code write an entire book

#69

Earlier quoted context omitted.

It seems silly to expect someone to read the whole book before evaluating it when the creator didn't even read the whole book before publishing it

If someone claims "the book is total shit", it is entirely reasonable to ask if they've actually read it, regardless of who or what wrote it.

If you flip through something claiming to be a "book" and immediately see that a majority of pages just contain nonsensical bulleted lists, and furthermore see that chapter titles are printed overlapping with the book title on each page, you can correctly conclude the entire thing is a zero-effort pile of shit without wasting any further time to read it.

Re: I let Claude Code write an entire book

#70
post #61
post #15

Earlier quoted context omitted.

It is interesting to me that your emphasis is on _lot_, and not on _read_.

Why? I emphasized "lot" because presumably the training data for a high-end model like Claude includes literally millions of books, far more than any human being has read even a few pages of. To me, "Claude has read a lot of books" would read pretty oddly here. I suppose the idea would be to contrast with the (hopefully few) books it's written to date, but I can't think why the emphasis would be called for.

Yes, I understood your argument, and your emphasis is on the point there.

I am simply asking in what sense did Claude read all those books. I am sorry if I have hijacked the disscusion.

edited: spelling

Post reply on HN