Hacker News Reader

Comment by world2vec | Hacker News Reader

It feels like I won't be reading any books published from 2023 onwards unless they came from writers I already know and respect or are known and respected by those writers.

View on HN

Books will be fine. Most people don't buy books from first-time authors blindly.

What will suffer, and possibly even disappear, are online comments sections. Yes, they're pretty much gone anyway because nobody besides Meta/Twitter wants to moderate discussions. But ChatGPT-driven bots will probably kill what little is left of organic conversations.

Hopefully Dang and HN will persist.

So it will come down to curation. Being a trusted source for recommendations may become an actual business, kind of an ad blocker for books.

I would assume "buy from publishers you trust" will still be a good strategy. Publishers can enforce some baseline of quality, where on-demand printed books is just a flood of things of every possible quality level.

Of course making sure nothing is written using ChatGPT is impossible, but you can make sure the reader gains something from reading the book (whether that's insight or entertainment)

But you’ll never know if the writer you respect is generating future novels with an AI trained on their past writing. Maybe this is better, but in a very short time, AI writing will be indistinguishable from human writing.

I admit it's not a perfect heuristic but it's the only one I can think of at the moment. Things will definitely change quite a lot in the next few years and I'll need to adapt accordingly.

Any awards for up-and-coming authors or best first publish for 2023 will be interesting

That is bigoted speciesism. As long as the books are good - does it matter what generated them

It's okay to be prejudiced.

He doesn’t like when the other mammals only enjoy mammalian things. It makes his cold reptileman blood run even icier.

It does matter, to me at least. You can read whatever you want.

And how is it speciesism? LLMs are not biological entities nor conscious (yet, I guess).

That is called sarcasm, top shelf too, so it doesn't self-identify as such and you need to think as you read.

Didn't felt like sarcasm to me (and if it is, definitely not top-shelf but opinions may vary) but yeah it happens sometimes that those things go over my head.

I agree. It matters to me too and, as you say, it cannot be called "speciesism" even in the most generous definition of the word. I propose the term lifeism: ethical and aesthetic prejudice favouring entities that have a life over those that do not.

We can do a thought experiment:

There is a runaway trolley hurtling down a track in the direction of a mediocre human science-fiction author, who is not even good enough to quit his or her day job. On a side track is a server holding the only trained copy of an experimental ChatGPT model which can generate science-fiction novels as good and original as Arthur C. Clarke and Ursula K. Le Guin combined, and at a massive scale. You can pull a lever to divert the trolley towards the server, destroying the ChatGPT model.

Do you pull the lever? As a self-confessed lifeist I would definitely pull the lever.

    (By the way, the mediocre author is not actually on the track, but on a bridge on top of it, out of harm's way)

Not speciesism. I'm biased against anyone who has a history of bullshitting. In general I would like to know before I read something whether an LLM wrote it. Why not just read it and see if it's good, you say. Because the thing about bullshitters is that they can waste a lot of your time before letting you down badly. The ones who are good at it are the worst.

You don’t use reviews or ratings to decide what to read? That is crazy brave even for pure human writers.

I think it may be possible for someone to write an interesting book with the assistance of an LLM, but what this article is warning about is people dumping tons of low-effort AI generated content into the market. Those are just not going to be good.

Reply on news.ycombinator.com