Comment by ngriffiths | Hacker News Reader

Comment by ngriffiths | Hacker News Reader

Parent Full thread

> Additionally, not all writing serves the same purpose.

I think this is a really important point and to add on, there is a lot of writing that is really good, but only in a way that a niche audience can appreciate. Today's AI can basically compete with the low quality stuff that makes up most of social media, it can't really compete with higher quality stuff targeted to a general audience, and it's still nowhere close to some more niche classics.

An interesting thought experiment is whether it's possible that AI tools could write a novel that's better than War and Peace. A quick google shows a lot of (poorly written) articles about how "AI is just a machine, so it can never be creative," which strikes me as a weak argument way too focused on a physical detail instead of the result. War and Peace and/or other great novels are certainly in the training set of some or all models, and there is some real consensus about which ones are great, not just random subjective opinions.

I kind of think... there is still something fundamental that would get in the way, but that it is still totally achievable to overcome that some day? I don't think it's impossible for an AI to be creative in a humanlike way, they don't seem optimized for it because they are completely optimized for the sort of analytical mode of reading and writing, not the creative/immersive one.

> An interesting thought experiment is whether it's possible that AI tools could write a novel that's better than War and Peace. A quick google shows a lot of (poorly written) articles about how "AI is just a machine, so it can never be creative," which strikes me as a weak argument way too focused on a physical detail instead of the result. War and Peace and/or other great novels are certainly in the training set of some or all models, and there is some real consensus about which ones are great, not just random subjective opinions.

I am sure it could but then what is the point? Consider this, lets assume that someone did manage to use LLM to produce a very well written novel. Would you rather have the novel that the LLM generated (the output), or the prompts and process that lead to that novel?

The moment I know how its made, the exact prompts and process, I can then have an infinite number of said great novels in 1000 different variations. To me this makes the output way, way less valuable compared to the input. If great novels are cheap to produce, they are no longer novel and becomes the norm, expectation rises and we will be looking for something new.

I'm inclined to believe that the difference that makes the upper bound of human writing (or creativity) higher than that of an LLM comes from having experiences in the real world. When someone is "inspired" by others' work or is otherwise deriving ideas from them, they inevitably and unavoidably insert their own biases and experiences into their own work, i.e. they also derive from real-world processes. An LLM, however, is derived directly and entirely from others' work, and cannot be influenced by the real world, only a projection of it.

> Would you rather have the novel that the LLM generated (the output), or the prompts and process that lead to that novel?

The "process", in many cases, is not necessarily preferable to the novel. Because an important part of the creative process is real-world experiences (as described above), and the real world is often unpleasant, hard, and complex, I'd often prefer a novel over the source material. Reading Animal Farm is much less unpleasant than being caught in the Spanish Civil War, for example.

I agree with you.

I also think it's a matter of time before we start constructing virtual worlds in which we train AI. Meaning, representations of simulated world-like events, scenarios, scenery, even physics. This will begin with heavy HF, but will move to both synthetic content creation and curation over time.

People will do this because it's interesting and because there's potential to capitalize on the result.

I thought of this in jest, but I now see this as an eventuality.

> People will do this because it's interesting and because there's potential to capitalize on the result.

I don't know why anyone admits to thinking this. For one, there's nothing stopping you from making movies or writing stories now. You're not suddenly going to develop creativity or interesting ideas using LLMs, either.

Also, think it through. If everyone can yell at computer until movie fall out, there will be millions of them and nobody will pay for anything.

I don't want AI content, but there's a market in the belief that people do.

Is there? It sound like a bunch of uncreative people wishcasting.

> The "process", in many cases, is not necessarily preferable to the novel. Because an important part of the creative process is real-world experiences (as described above), and the real world is often unpleasant, hard, and complex, I'd often prefer a novel over the source material. Reading Animal Farm is much less unpleasant than being caught in the Spanish Civil War, for example.

I think you misunderstood what I meant by "prompts and process that lead to that novel". I am talking about the process that the "author" used to generate that novel output. I am more interested in the technique that they use, and the moment that technique is known. Then, I can produce billions of War And Peace.

I suppose the argument is that, the moment there's an LLM that can produce a unique and interesting novels, what stops it from generating another billion similarly interesting novels?

> Then, I can produce billions of War And Peace

You cannot and will never lol.

This so fundamentally misunderstands (1) the point of writing a novel and (2) what makes a novel interesting.

A novel isn't just a buncha words slapped together, bing bam slop boom, done.

What makes a novel interesting is the author and the author's choices, like all art. It's the closest you can get to experiencing what it's like to be someone else. You can't generate that, it's specific to a person.

The GP assumes that an LLM is able to write such novel. So I was working from there. My thesis is that even IF LLMs are able to produce "novelty", it will become the norm and we will simply demand even more exotic novelty.

> An interesting thought experiment is whether it's possible that AI tools could write a novel that's better than War and Peace. A quick google shows a lot of (poorly written) articles about how "AI is just a machine, so it can never be creative," which strikes me as a weak argument way too focused on a physical detail instead of the result. War and Peace and/or other great novels are certainly in the training set of some or all models, and there is some real consensus about which ones are great, not just random subjective opinions.

It can have anything you like in a training set, you still can't build specific human experiences.

I haven't read War & Peace -- I don't have the patience for Russian literature -- but a much more accessible example is the Vorkosigan series by Lois Bujold. She uses a lot of Tolstoy lol.

While you can read them as fun military scifi, that's not why the series is so good and so famous. In her books, humanity invented two critical things: wormhole FTL travel and uterine replicators.

A lot of the series is exploring how people actually would use and abuse those two things. And then on another layer the books are about her thoughts on parenting, marriage, power, inheritance, and so on.

Good art isn't about accepting someone's opinion that it's good art. Good art impacts you. I think about things differently after those books.

You cannot write a good novel using the algorithmic mean of a lot of different stories.

> Today's AI can basically compete with the low quality stuff that makes up most of social media, it can't really compete with higher quality stuff

But compete in what sense? It already wins on volume alone, because LLM writing is much cheaper than human writing. If you search for an explanation of a concept in science, engineering, philosophy, or art, the first result is an AI summary, probably followed by five AI-generated pages that crowded out the source material.

If you get your news on HN, a significant proportion of stories that make it to the top are LLM-generated. If you open a newspaper... a lot of them are using LLMs too. LLM-generated books are ubiquitous on Amazon. So what kind of competition / victory are we talking about? The satisfaction of writing better for an audience of none?

> if you get your news on HN, significant portion that make it to the top are LLM-generated.

You mean this anecdotally I assume.

This makes me think of the split between people who read the article and people who _only_ read the comments. I'm in the second group. I'd say we were preemptive in seeking the ideas and discussion, less so achieving "the point" of the article.

FWIW, AI infiltrates everything, i get that, but there's a difference between engagement with people around ideas and engagement with the content. it's blurry i know, but helps to be clear on what we're talking about.

edit: in this way, reading something a particular human wrote is both content engagement and engagement with people around an idea. lovely. engaging with content only, is something else. something less satisfying.

There are very few things worth reading submitted to this site. The only meaningful thing I'm glad to have read was the "I sell onions on the internet" blog post. Everything else I've forgotten, mostly VC marketing fluff or dev infighting in open source; hardly anything worth noting.

This place is up there with reddit, it's all lowish calorie info; 90% forgettable, 10% meaningful but you have to dig quite quite deep to find it.

To be fair, it has gotten harder, but when the meaningful stuff does happen, it is hard to beat. Some of the audience can have rather pointed takes. And if it is then somehow topped by 'off the beaten path' guy, it really makes it for me ( in the sense that maybe not all is lost quite yet ). I still sometimes reel from 'manifest bananas' guy.

Tens of millions of people, if not hundreds now thanks to the popularity of the television adaptation, have been waiting 15 years now for Winds of Winter to get published. If AI is such a good writer and can replace anything, write Winds of Winter for George. I don't really give a shit what's ubiquitous on Amazon. Nobody will remember any of it in a century the way we remember War and Peace. People will remember the Song of Ice and Fire books.

I think it's fine. As said above, most reading isn't done because people are looking for thought-provoking, deeply emotional multi-decade experiences with nearly parasocial relationships to major characters. They're just looking to avoid the existential dread of being alone with their thoughts for more than a few minutes. There's room for both twinkies and filet mignon in the world and filet mignon alone can't feed the entire world anyway. By the same token, if we expected all journalists to write like H.L. Menken, a lot of people wouldn't get any news, but the world still deserves to have at least a few H.L. Menkens and I don't think they'll have an audience of "none" even if their audience is smaller than Stephanie Meyer or whoever is popular today.

If it were me, I don't know man, does nobody on Hacker News still care about actually being good at anything as opposed to just making sales and having reach? Personally, I'd rather be Anthony Joshua than Jake Paul, even though Jake Paul is richer. Shit, I think Jake Paul himself would rather be Anthony Joshua

>The satisfaction of writing better for an audience of none?

The satisfaction of writing for an engine. The last of what could still be recognized as a real human being writing. There’s no competition with AI, but also no resignation and no fear of being limited compared to the vast knowledge of an LLM. Even in a context of an "audience of none", somewhere there will be a scraper tool interested in my writing. And if it gets hallucinated... wow!

Reply on news.ycombinator.com