Words, Words, Words(aeon.co)
aeon.co
Words, Words, Words
https://aeon.co/essays/literature-fans-should-welcome-ai-as-a-fellow-wordsmith
16 comments
Derrida was worried that people were merely imitating his style of thinking without the substance, but it's fine for AI to do it to humanity-in-general. OK.
I agree with the point that the content of the text matters, but most people care more about who wrote it. For example, no matter how much I write on my blog (www.makonea.com), no one reads it. But if the same content is written by someone famous, it gets cited far more often. That's the problem.
There are people who want to obtain information from text, and there are people who want "communication." In communication, the more famous the person is, the more enthusiastic the response tends to be. In other words, the purpose of reading is often different.
So I think there are two types of reading: reading for communication and reading for information.
Writing for communication elicits completely different reactions depending on who wrote it. That's one example. Writing by AI betrays that expectation. The problem is that there's quite a fuzzy boundary between what's written by AI and what's written by me.
On top of that, I want to add something about the "flatness" of AI writing. By flatness, I mean not pushing any particular argument strongly and appropriately acknowledging drawbacks as well. When I write my blog posts, I run everything through an AI red-team review. The final draft ends up very different from the initial one, but the writing loses the sharpness of my subjective bias. It becomes blunted. As a result, the sentences become flatter in places. This happens because the AI detects and smooths out my unconscious biases. (Of course, AI has its own biases too, but humans also chip away at AI's biases.)
That said, if I don't use AI at all, since I pride myself on running a technical blog, not detecting technical errors would also be a problem. So it's a dilemma.
My personal concern is that in the long run, a lot of people's writing will become flat. AI is genuinely useful as an encyclopedia, and while it can contain errors, on average it's often more accurate than I am, so it's hard not to use it. The problem is that the materials AI learned from, the writings of knowledgeable people, were shaped by a strategy of finding safe, low-cost landing points when engaging in high-cost criticism. In a social context, that's actually a fairly valid strategy. Strong claims invite strong backlash, but most people smooth things over appropriately for the sake of social harmony.
The issue is that human writing is generally rougher. But if we insist on preserving only human writing, we end up having to define humanity's roughness as mere "barbarism." Ultimately, I think human writing might even start imitating AI writing. And if that happens, what becomes of writing as "communication"? We write not only for information, but also because we want to talk to someone. Like when I write on HN.
And while AI expands the creative world, giving tools to people who want to create but didn't know how, like building a philosophical Frankenstein RAG chatbot, whether that tool is actually accurate ultimately requires knowing how to use the tool properly. In other words, liberation and accuracy exist in tension with each other.
As for cultural comparisons, I've worked with people from Korea, China, Japan, and the US. The fact that Korea and China are more lenient toward AI is partly because Korea, China, and Japan are direct beneficiaries of AI. For instance, Korean stocks are essentially a massive semiconductor leverage play, and the direct consumers of memory semiconductors are AI. China, meanwhile, sees AI as a new path to surpass the US.
I really enjoyed reading your text. Humanities scholars definitely write in a way that's fun and logically compelling, which makes it a pleasure to read. It really drove home for me that this is the kind of writing you need to produce to make money from writing. Fortunately, I'm grateful that my job is being a programmer.
There are people who want to obtain information from text, and there are people who want "communication." In communication, the more famous the person is, the more enthusiastic the response tends to be. In other words, the purpose of reading is often different.
So I think there are two types of reading: reading for communication and reading for information.
Writing for communication elicits completely different reactions depending on who wrote it. That's one example. Writing by AI betrays that expectation. The problem is that there's quite a fuzzy boundary between what's written by AI and what's written by me.
On top of that, I want to add something about the "flatness" of AI writing. By flatness, I mean not pushing any particular argument strongly and appropriately acknowledging drawbacks as well. When I write my blog posts, I run everything through an AI red-team review. The final draft ends up very different from the initial one, but the writing loses the sharpness of my subjective bias. It becomes blunted. As a result, the sentences become flatter in places. This happens because the AI detects and smooths out my unconscious biases. (Of course, AI has its own biases too, but humans also chip away at AI's biases.)
That said, if I don't use AI at all, since I pride myself on running a technical blog, not detecting technical errors would also be a problem. So it's a dilemma.
My personal concern is that in the long run, a lot of people's writing will become flat. AI is genuinely useful as an encyclopedia, and while it can contain errors, on average it's often more accurate than I am, so it's hard not to use it. The problem is that the materials AI learned from, the writings of knowledgeable people, were shaped by a strategy of finding safe, low-cost landing points when engaging in high-cost criticism. In a social context, that's actually a fairly valid strategy. Strong claims invite strong backlash, but most people smooth things over appropriately for the sake of social harmony.
The issue is that human writing is generally rougher. But if we insist on preserving only human writing, we end up having to define humanity's roughness as mere "barbarism." Ultimately, I think human writing might even start imitating AI writing. And if that happens, what becomes of writing as "communication"? We write not only for information, but also because we want to talk to someone. Like when I write on HN.
And while AI expands the creative world, giving tools to people who want to create but didn't know how, like building a philosophical Frankenstein RAG chatbot, whether that tool is actually accurate ultimately requires knowing how to use the tool properly. In other words, liberation and accuracy exist in tension with each other.
As for cultural comparisons, I've worked with people from Korea, China, Japan, and the US. The fact that Korea and China are more lenient toward AI is partly because Korea, China, and Japan are direct beneficiaries of AI. For instance, Korean stocks are essentially a massive semiconductor leverage play, and the direct consumers of memory semiconductors are AI. China, meanwhile, sees AI as a new path to surpass the US.
I really enjoyed reading your text. Humanities scholars definitely write in a way that's fun and logically compelling, which makes it a pleasure to read. It really drove home for me that this is the kind of writing you need to produce to make money from writing. Fortunately, I'm grateful that my job is being a programmer.
>That said, if I don't use AI at all, since I pride myself on running a technical blog, not detecting technical errors would also be a problem. So it's a dilemma.
Is it? To err is human. Embrace the risk of technical errors and what/how we learn from them.
>The issue is that human writing is generally rougher. But if we insist on preserving only human writing, we end up having to define humanity's roughness as mere "barbarism."
Is it truly "barbarism", or is it just... human? What's wrong with writing being a bit "rough around the edges"?
Is it? To err is human. Embrace the risk of technical errors and what/how we learn from them.
>The issue is that human writing is generally rougher. But if we insist on preserving only human writing, we end up having to define humanity's roughness as mere "barbarism."
Is it truly "barbarism", or is it just... human? What's wrong with writing being a bit "rough around the edges"?
Having the LLM rewrite your writing always, in my experience, loses any sense of what the author cared about, everything is given monotone weight instead of somethings more important than others.
But does your writing have to become blunted after a red-team _review_? You're running a technical blog and want to detect technical errors, that's great - a red-team review sounds like something that would tell you the technical errors and you'd fix those up without rewriting your words.
But does your writing have to become blunted after a red-team _review_? You're running a technical blog and want to detect technical errors, that's great - a red-team review sounds like something that would tell you the technical errors and you'd fix those up without rewriting your words.
Rugnir's points:
> "readers feel cheated when they learn that a text was written by AI. But will this attitude persist?"
Yes, yes it will. I don't say this as a value judgement, just an inevitability. When people can generate stuff themselves, it becomes less magical. You have to do something they themselves cannot, or would not think to do. Its possible to do surprising and insightful with ai, but you aren't going to get that with a single one paragraph prompt. If you aren't sure, you aren't doing it. By definition, very few people will reach this level with AI (because if more did, it would no longer be noteworthy or interesting).
I imagine social norms will develop that you chat with these yourself to develop something (including having it fetch human works on the topic for you to read), then write your piece in your own words/say it in your own voice.
> "an app that stages debates among these philosophers, allowing what many philosophers and humanists have long dreamed about: assembling the best minds in the world and having them talk to each other."
this is not actually assembling the best minds, its asking one mind to write fanfiction about what it thinks those best minds might have said. It might be interesting, just as such a work could be if authored by a human, but it is meaningfully different.
overall, I personally reckon AI's will likely be integrated into most peoples workflows and ways of thinking eventually. Some valuable things will be lost, or at least diminished as a result. But at the same time, it won't be a disaster, and life will go on. People will recognise the biases the tools have, and learn to have self control, restraint, and social norms around their use that lets society continue to exist in some form or another.
I don't say that because I think it's good (or bad!), just what I think will actually happen.
> "readers feel cheated when they learn that a text was written by AI. But will this attitude persist?"
Yes, yes it will. I don't say this as a value judgement, just an inevitability. When people can generate stuff themselves, it becomes less magical. You have to do something they themselves cannot, or would not think to do. Its possible to do surprising and insightful with ai, but you aren't going to get that with a single one paragraph prompt. If you aren't sure, you aren't doing it. By definition, very few people will reach this level with AI (because if more did, it would no longer be noteworthy or interesting).
I imagine social norms will develop that you chat with these yourself to develop something (including having it fetch human works on the topic for you to read), then write your piece in your own words/say it in your own voice.
> "an app that stages debates among these philosophers, allowing what many philosophers and humanists have long dreamed about: assembling the best minds in the world and having them talk to each other."
this is not actually assembling the best minds, its asking one mind to write fanfiction about what it thinks those best minds might have said. It might be interesting, just as such a work could be if authored by a human, but it is meaningfully different.
overall, I personally reckon AI's will likely be integrated into most peoples workflows and ways of thinking eventually. Some valuable things will be lost, or at least diminished as a result. But at the same time, it won't be a disaster, and life will go on. People will recognise the biases the tools have, and learn to have self control, restraint, and social norms around their use that lets society continue to exist in some form or another.
I don't say that because I think it's good (or bad!), just what I think will actually happen.
> write fanfiction about what it thinks those best minds
I asked AI what Einstein would be working on today if he were alive, the result was a predictably rosy scenario of modern physics problems.
However the reality was that Einstein (in his later years) became isolated from physics developments.
I asked AI what Einstein would be working on today if he were alive, the result was a predictably rosy scenario of modern physics problems.
However the reality was that Einstein (in his later years) became isolated from physics developments.
While I don't think there's much to be gained by asking AI for a verisimilitude of famous dead people, GPT 5.5 in response to that prompt ("What would Einstein be working on today, if he were alive") was more nuanced:
This is an educated guess, because we cannot know how Einstein would have reacted to post-1955 developments: Bell’s theorem, quantum information, the Standard Model, inflation, black hole thermodynamics, dark matter/dark energy, gravitational waves, and modern experimental cosmology. The missing information that would most change my answer is whether “today’s Einstein” would retain his later-life philosophical commitments — realism, locality, geometric unification — or adapt them after seeing modern evidence.
This is an educated guess, because we cannot know how Einstein would have reacted to post-1955 developments: Bell’s theorem, quantum information, the Standard Model, inflation, black hole thermodynamics, dark matter/dark energy, gravitational waves, and modern experimental cosmology. The missing information that would most change my answer is whether “today’s Einstein” would retain his later-life philosophical commitments — realism, locality, geometric unification — or adapt them after seeing modern evidence.
> I imagine social norms will develop that you chat with these yourself to develop something (including having it fetch human works on the topic for you to read), then write your piece in your own words/say it in your own voice.
This is exactly what the author was arguing for towards the end of the article:
"At the same time, we felt that, if used right, AI could enhance the process of writing and thinking. To this end, we teach students how to create AI agents that act as sparring partners, forcing them to sharpen their thinking by reacting to counter-argument and counter-evidence."
I do this myself, and find it highly useful for my own research.
This is exactly what the author was arguing for towards the end of the article:
"At the same time, we felt that, if used right, AI could enhance the process of writing and thinking. To this end, we teach students how to create AI agents that act as sparring partners, forcing them to sharpen their thinking by reacting to counter-argument and counter-evidence."
I do this myself, and find it highly useful for my own research.
It feels obvious, but people feel cheated when they have been cheated. If you learn that text was AI when it was presented as human written (even implicitly) then that's deception and people are right to feel deceived. Just as when "true story" films like The Salt Path turn out to be bullshit.
I think the real question is how people will feel about text that is explicitly labelled as AI generated.
I think the real question is how people will feel about text that is explicitly labelled as AI generated.
the problem we have at the moment is most text isn't presented as anything; the default assumption is that a human wrote it, because until a couple years ago, we knew of nothing else that could write. So people feel betrayed if they see a text written by ai, anywhere, unless real clear upfront told its AI. Thats the implicit element problem: all text is implicitly human written, even though LLMs probably author more words than the entire human race on any given day.
That's what I mean by a new social norm is being written, and I see it going the direction where damn near everything corporate or longform has "partially written by ai" on it as some kinda legalese thing (and is the default assumption where no disclaimer is present), and a select few human writers have to write "written only by humans", much though some loathe the need for such disclaimers.
there is another world, where the social pressure is too great, and ai use too embarassing, such that its pushed out of the public sphere in all but a few cases. I also see elements of this forming in some spaces (including most creative ones), though which of the two will win out culturally I can't tell - the first seems to have the governmental and business backing, the second grassroots support. I predict the first world. (though clear or obvious ai usage in the wrong setting will continue to be a faux pas - I mean that your disney films and your marketing releases and news articles will be (and are now) partially ghostwritten by ai, but it will still be rude to use an ai to write to your friends on whatsapp etc)
That's what I mean by a new social norm is being written, and I see it going the direction where damn near everything corporate or longform has "partially written by ai" on it as some kinda legalese thing (and is the default assumption where no disclaimer is present), and a select few human writers have to write "written only by humans", much though some loathe the need for such disclaimers.
there is another world, where the social pressure is too great, and ai use too embarassing, such that its pushed out of the public sphere in all but a few cases. I also see elements of this forming in some spaces (including most creative ones), though which of the two will win out culturally I can't tell - the first seems to have the governmental and business backing, the second grassroots support. I predict the first world. (though clear or obvious ai usage in the wrong setting will continue to be a faux pas - I mean that your disney films and your marketing releases and news articles will be (and are now) partially ghostwritten by ai, but it will still be rude to use an ai to write to your friends on whatsapp etc)
I found the first paragraph amusing.
> Since artificial intelligence went mainstream a few years ago, it has done double duty as a political personality test: tell me what you think about AI, and I’ll tell you who you are.
> Those worried about climate change focus on energy consumption.
> Those who denounce late capitalism see it as the ultimate example of corporate monopoly.
> Those concerned about racism have warned about AI biases.
> Those studying the effects of colonialism see it as yet another form of exploitation.
> And those tending toward doom have seen ChatGPT, Claude, Gemini and Grok as the four riders of the apocalypse.
I think one of the ways of going about thinking this is how we must've felt about being shared a google search (let me google that for you) or within programming, a stackoverflow link by say someone in the team or anything.
We wouldn't feel like they have done any effort because they haven't. The idea about readers feeling cheated is the same way as such and the attitude towards disdain of AI would be in same capacity as someone just giving us a stackoverflow link or google link. (tangentially, Ironic that both of these (SO and google) are now focusing on AI as well)
And in essence we humans would favour human effort. Not because its about the end output but perhaps the journey as well and more important than anything, if a human being is writing about something you enjoy or are passionate of, you firstly have a community and are able to talk with a actual person with hopefully nuanced opinions and just discuss things with things and most importantly, you wouldn't feel alone.
Human cinema,art (media) and music actually go through multiple invisible filters of rejections and previously centralized control and now algorithmic.
Internet has helped with it but it has once again been capitalized and the thing we know is that Ragebait and lack of nuance sells and so the economic incentive alone has led us to the state of the internet that we generally so speak of.
So in some sense, one can argue that even art created by humans have some algorithmic influence and preference but it has also been shown that there are people who are able to persist through with an idea and fight up against the algorithmic, at the end of the day algorithms although having power still serve the main purpose of money and if enough people want something and are willing to pay for it, someone is probably going to provide it who is genuinely passionate about it too.
So in some sense, real creativity and talent are already pushed down upon but it is part of the human spirit not in lack of it but inspite of all these negative conditions which is able to persist through. People love making music and they make it inspite of all the negativity and persist through. I am taking the example of music here and linking a relevant video for those interested[0]
I suppose the same goes for other art and I think that there can be two things correct at the same time. Internet is full of corporate-esque things and also niche things created by the spirit and love of mankind. I suppose that AI might have its impact on the corporate-esque part of things as I feel like some layoff letters are already getting written by AI in dystopian fashion but I also believe that it wouldn't be able to attack or impact the niche created by the spirit and love of mankind aspect that I mentioned.
But with corporations using AI they are pushing an already broken system down to an even more broken system and humans have shown time and time again that there is a breaking point for such scenarios from multiple revolutions throughout and one can argue that there is enough hate for AI already that it can constitute as one of the forming points. There has and should be more use with caution of AI even within corporate if they wish to survive long term as using AI unsustainably and hammering it everywhere both loses any sort of monopoly that they have, after all I could create an alternative of it with AI just as well and also at the same time, make their customers look for other alternatives which aren't filled with the slop-aganza of AI.
[0]: The music industry is broken: https://www.youtube.com/watch?v=Yx7baJMQuVA
We wouldn't feel like they have done any effort because they haven't. The idea about readers feeling cheated is the same way as such and the attitude towards disdain of AI would be in same capacity as someone just giving us a stackoverflow link or google link. (tangentially, Ironic that both of these (SO and google) are now focusing on AI as well)
And in essence we humans would favour human effort. Not because its about the end output but perhaps the journey as well and more important than anything, if a human being is writing about something you enjoy or are passionate of, you firstly have a community and are able to talk with a actual person with hopefully nuanced opinions and just discuss things with things and most importantly, you wouldn't feel alone.
Human cinema,art (media) and music actually go through multiple invisible filters of rejections and previously centralized control and now algorithmic.
Internet has helped with it but it has once again been capitalized and the thing we know is that Ragebait and lack of nuance sells and so the economic incentive alone has led us to the state of the internet that we generally so speak of.
So in some sense, one can argue that even art created by humans have some algorithmic influence and preference but it has also been shown that there are people who are able to persist through with an idea and fight up against the algorithmic, at the end of the day algorithms although having power still serve the main purpose of money and if enough people want something and are willing to pay for it, someone is probably going to provide it who is genuinely passionate about it too.
So in some sense, real creativity and talent are already pushed down upon but it is part of the human spirit not in lack of it but inspite of all these negative conditions which is able to persist through. People love making music and they make it inspite of all the negativity and persist through. I am taking the example of music here and linking a relevant video for those interested[0]
I suppose the same goes for other art and I think that there can be two things correct at the same time. Internet is full of corporate-esque things and also niche things created by the spirit and love of mankind. I suppose that AI might have its impact on the corporate-esque part of things as I feel like some layoff letters are already getting written by AI in dystopian fashion but I also believe that it wouldn't be able to attack or impact the niche created by the spirit and love of mankind aspect that I mentioned.
But with corporations using AI they are pushing an already broken system down to an even more broken system and humans have shown time and time again that there is a breaking point for such scenarios from multiple revolutions throughout and one can argue that there is enough hate for AI already that it can constitute as one of the forming points. There has and should be more use with caution of AI even within corporate if they wish to survive long term as using AI unsustainably and hammering it everywhere both loses any sort of monopoly that they have, after all I could create an alternative of it with AI just as well and also at the same time, make their customers look for other alternatives which aren't filled with the slop-aganza of AI.
[0]: The music industry is broken: https://www.youtube.com/watch?v=Yx7baJMQuVA
Significant omission: those worried about automating the production of bullshit at industrial scale.