What Writers Get Wrong About AI

00:00:00: There's a sort of clipped, sassy tone that AI writing like LLMs produce.

00:00:06: That is kind of tech-inflected and bit weird And you can identify it I think Trying

00:00:13: to be punchy but not quite putting together.

00:00:14: It's

00:00:15: trying to be punchy.

00:00:15: But also it's punchy Like an idiot Is trying hit on you.

00:00:19: You know what i mean?

00:00:20: They're just saying stuff.

00:00:24: Hi!

00:00:25: I'm Tobias Dabacorn and your listening to Air Lift from Berlin Review.

00:00:32: Last

00:00:32: one Something embarrassing happened in the English-speaking

00:00:36: literary world.

00:00:38: Well respected magazine published an award winning story, while none of the editors noticed many readers suspected that it was written by AI.

00:00:47: Lauren explains this scandal but I promise This is not another episode about how AI Is killing writing.

00:00:54: It's actually what most writers even The good ones are Not understanding About new technology And what they can do to keep writing human and even make it more interesting.

00:01:06: Here is my conversation with Lauren Euler.

00:01:09: Hi, Lauren!

00:01:10: Nice to be back.

00:01:11: Great to have you in the studio.

00:01:14: So there was a major scandal in English-speaking literary world this May involving AI and cheating.

00:01:23: What happened?

00:01:24: Okay well I love to explain a literary world scandal to people who don't know.

00:01:31: Yeah, so basically in the middle of May this short story prize called The Commonwealth Short Story Prize which awards A little bit of money I think between three thousand and five thousand pounds To uh short stories produced by unpublished or barely published writers from commonwealth countries.

00:01:50: i believe there are Five regions.

00:01:51: each picks a regional winner And then there's a big prize that has not been announced yet.

00:01:56: now The Commonwealth Short Story Prize is apparently, I didn't know this relatively well regarded as a short story prize.

00:02:04: And such the regional winners are every year published online by Granton Magazine, the storied UK literary magazine.

00:02:14: But at the same time, as with many things in the literary world they're very prestigious.

00:02:18: but also no one cares.

00:02:19: Okay so Granta publishes on their website not-in-print just On Their Website The five regional winners of this Commonwealth short story prize.

00:02:28: and someone super sleuth posted something that was like This is interesting.

00:02:34: it seems That an AI generated short story has won a prestigious literary Prize And what followed were some screenshots pretty what we would call purple prose, but kind of like manically purple prose written by a guy from Trinidad and Tobago named Jameer Nazir.

00:02:56: And he had won the Caribbean regional Commonwealth Story Prize.

00:03:01: That was second story as well that was detected to people now?

00:03:03: They were actually three.

00:03:05: There were three

00:03:06: out of the five.

00:03:06: Three Out Of The Five people became obsessed with like rooting out AI in these stories, their right wing literary sphere that exists now.

00:03:16: obviously there was some racial elements they would play on indicating kind of woke.

00:03:23: post-colonial literature has gotten out control and look how bad it is.

00:03:27: but just to get this clear those are actual the short stories, it's not that there is a fictional made-up name like an anonymous person.

00:03:39: So I believe he has given interviews since but this created total uproar right?

00:03:44: Lots of literary journalists try to contact him get some quotes from him.

00:03:49: um i saw various articles which are more journalistically sound than others about people having DM'd Tim and other people who had won regional prizes in his competition.

00:03:59: Basically it caused like a huge frenzy because everybody is obsessed with AI and also Everybody, you know.

00:04:05: I think there's a desire to puncture the bubble of literary prestige Like the teeny-tiny little area of literary prestigious that still exists.

00:04:13: Let's talk about Grand Theft as well.

00:04:15: Because basically rather than people focusing on the Commonwealth Short Story Prize Which chose these winners A lot of the burden responsibility fell on Granta Magazine For not having

00:04:26: noticed.

00:04:27: Yeah.

00:04:28: They're editors publish these stories without looking further into how kind of hallucinatory to use an AI word, they were.

00:04:37: And then after a few days of internet frenzy where everybody was condemning Granta and condemning the Commonwealth Story Prize being racist, Granta issued quite bizarre or bad statement saying that the winning short story through Claude, the LLM from Anthropic asking it if he thought that this story was AI generated.

00:05:02: and Claude said probably.

00:05:03: And then the publisher of Granta basically threw up her hands and she said It's possible.

00:05:08: we'll never know for sure If this is AI generated or not?

00:05:13: Just

00:05:14: a technical question.

00:05:15: I mean We've been you know confronted with AI generated Text for almost a decade now

00:05:22: probably

00:05:23: and it's true that especially early on It was so so obvious I'm less sure whether this is So easy to establish.

00:05:34: That some things clearly AI.

00:05:35: Now, so the people that are making an assessment of basing in on clear like structural stylistic elements where they say This sounds so much like

00:05:43: I'm

00:05:44: a hundred percent sure this must be AI because it sounds like.

00:05:47: Is that?

00:05:47: Yeah, and there are certain ticks.

00:05:49: but also there's certain structural things that does in the way that creates what we would call mixed metaphors which don't make any sense.

00:06:01: The issue when talking about like, literary writing in AI is that many literary writers are very bad at what they do naturally without AI.

00:06:12: They're just naturally quite bad...they write exactly the same kind of nonsense.

00:06:16: metaphors.

00:06:17: people say are definitely AI- they love to do not this but that constructions and an M dash.

00:06:24: I remember when i was a university into like, twenty-ten and I was in a writing class.

00:06:29: And there is long discussion about whether we all use too many M dashes because they're so fun to use right?

00:06:35: You send me that sub stack piece by the writer Sam Chris.

00:06:38: Yeah That came out couple of days ago.

00:06:40: What's title?

00:06:41: again?

00:06:41: it's like...I know you are using AI.

00:06:43: Oh no don't use AI or else

00:06:45: your house

00:06:45: will kill you.

00:06:46: If you use AI for your writing then i'll come up with yours.

00:06:49: Yes

00:06:50: So uh..i read this in preparation and its definitely nice piece.

00:06:54: he has great style.

00:06:56: He´s good critic But he's so sure that is able to detect AI in the day.

00:07:01: I, The flaw of AI writing a sort of like baked-in and stochastic mechanism predictive continuity incidences.

00:07:13: Yeah That he says exactly what you say at the mixed metaphors And that this looks slightly nonsensical guesswork.

00:07:20: He says that This will...this is recognizable and forever will be.

00:07:25: Therefore, you know stop using it because everyone sees that is obvious and I've been editing Text for fifteen years now.

00:07:35: And It's true That sort of like Analyzing the metaphors a writer uses and finding where they are Sort-of Like off Because They Are Combining i don't Know imagery or elements that Don't quite go together.

00:07:49: Maybe he's right at The in this particular cases that He has analyzed.

00:07:54: But on a sort of like more philosophical or theoretical level, the misuse and non-standard use of language is what always amounts to poetry or poetic or literary use in general.

00:08:10: So I'm just skeptical about our argument that we will forever be able to detect AI writing... No!

00:08:17: Well also sceptical because they think there are lots things that you know AI has trained huge data sets of human writing, you can also sort of train your own model to do like more specific kind.

00:08:35: People use this term AI plagiarism to mean like the use of AI and something that you claim is human-generated.

00:08:41: But actually what AI often does, it's literally plagiarized in an old fashioned way Which takes sentences from other pieces of writing And puts them out as if they are Something that the AI has generated itself.

00:08:53: So kind of cuts off the middleman of plagiarist where You can say oh I didn't steal that sentence From this writer It was just... The AI stole right?

00:09:02: Sam Chris In.

00:09:04: in trying to make this critique which he's made several times and several different outlets He had a really great piece in the New York Times magazine.

00:09:09: A few months ago about um AI style, he Had This Great Harpers cover story About Silicon Valley going to Silicon Valley And he captures Really well like The kind of crazy making Like hallucinatory Or or like the contagiousness Of the hallucination aspect?

00:09:27: Which is like you read something that's AI or You use AI and you're like I feel insane, but he's taking for granted in a lot of these pieces that like most people would find what is produced insane.

00:09:42: But most do not have an unique deep passionate relationship with language the way Sam Chris and I are doing it.

00:09:56: You mean also

00:09:58: literary writers?

00:09:59: Yeah, yeah.

00:10:00: People like I would say what distinguishes a literary writer from journalist or someone else is that they pay attention to the nuances of language and use language to create effects

00:10:12: so...

00:10:15: But but i think most even literary writers don't do that.

00:10:23: They just don't.

00:10:25: If literary writers, critics like if writers and people in the humanities in general would like to seriously combat.

00:10:33: The encroachment of AI into our tiny cultural sphere you kind of need to learn how it actually works.

00:10:41: because everyone's like I'm sure It's good for math!

00:10:43: Like iIm sure its' Good For This!

00:10:45: But most People don't even know How it Works.

00:10:47: And I think the primary example Of this... I just wrote a very long essay about AI A slightly different topic but I Just Wrote About This quite shocked to find while I was researching this that many, many articles about AI and writing in literature don't even understand like the basic premise of what happens with.

00:11:12: They say that it works by predicting what word is most likely to come after another word.

00:11:19: So you have a prompt and then it just sort of works like buy-it Like a vast predictive model.

00:11:23: That's

00:11:23: some crystals

00:11:24: It's not words, its tokens.

00:11:27: so it's not even using like what like linguists would understand as language.

00:11:33: Sometimes if see it's working with linguistic units which is the linguistics term but sometimes?

00:11:40: Letters.

00:11:41: now at this point and also I should say that the technology changes very rapidly.

00:11:46: so when we're talking about This it might not be true in two weeks.

00:11:48: That's just you know,

00:11:49: whatever.

00:11:49: And I think they were emergent.

00:11:51: but at this Point there are some emergent properties of It where even The tokens based decomposition Of language yeah structure is Like one stage But then other things happen?

00:12:06: And i'm Not sure if Even the engineers who make the models Are a hundred percent sure of what happens at the even higher levels.

00:12:12: Yeah, they don't really understand

00:12:16: how it works!

00:12:18: I mean we also do not understand human mind work right.

00:12:20: so its kind fair.

00:12:21: but this thing about is like the devaluation of language and meaning conveyed through language has been happening before this, certainly.

00:12:32: And so there are lots people who want to be writers that I don't really care about as such like...I wanna create some kind of phantom story or phantom piece of writing that functions as a piece That doesn't actually convey a lot of meaning.

00:12:55: Um, and I do think that there's lots of like writing that does this kind of... Like.

00:12:58: oh it sounds cool!

00:13:00: It just sounds poetic-it sounds lyrical right?

00:13:03: And i think the story um, that Granta published by Jameer Nazir called The Serpent in the Grove at an all-even thing I've said what is called Does have this kinda childish Like delight at the sound of words that have kind of little to do with like, The meaning of the word together.

00:13:22: Yeah

00:13:23: which can produce interesting poetry sometimes but usually the poets who are most successful With that strategy also know what the words Know how language functions and they break their rules.

00:13:35: They know the rules before they could break them as it were.

00:13:38: That's an interesting point.

00:13:39: if we zoom out on the question of literature use and meaning and status of writing in general, life like the way economy or any sort system works.

00:13:51: It's that riding... In general so many professions.

00:13:57: what gives you professional status and proves your job is a piece of writing.

00:14:03: How so?

00:14:04: Well I mean okay this comes from Germany but amount reporting that people have to do in any sort of like public job, even companies or universities.

00:14:21: You have to prove someone else.

00:14:22: you've done this and then there's not so far-fetched analogy for the common wealth price.

00:14:38: It's not by accident that they're competing for a literary prize, which ideally will give them some sort of statues.

00:14:46: And then they hand in piece writing.

00:14:49: so the painful process of writing down things is often... In literature and literally it shouldn't but I think happens also outside of literature That painful processes actually only undergone by a lot of people because they need the end result, because it does something for them.

00:15:08: Yeah

00:15:09: but also I think that's sort of like perverse about it is most of those people don't like writing and want to have do these kind reports or summaries you know what i mean?

00:15:19: For their job!

00:15:20: And then just do anyway...and then for this AI my understanding its total game changing experience where they dont' have spend hours being so anxious about sending an email But with writing as an art form, it's really funny because it just demonstrates the very human tendency which is they want to be a writer without having any writing?

00:15:48: Yes.

00:15:48: But I also thought you were going to say, oh in bureaucracy or sort of normal jobs one has to produce a bunch of writing to prove their legitimacy and the role.

00:15:58: And then like ah well for writing you have to produce much of writing For literature.

00:16:02: You've produced a lot of writing To prove your legitimate Like the legitimacy Of your soul Or spirit Because it's supposed come from some process of thinking philosophically creatively, whatever that is unique to you.

00:16:18: Whereas in bureaucracy uniqueness it's definitely like a weakness.

00:16:23: why am I just creating all these sort of rhyming this?

00:16:25: It's coming as the uniqueness is a weakness.

00:16:32: Yeah but what?

00:16:32: What Is Going To Happen Then?

00:16:34: um i mean its an academic writing The Joke.

00:16:40: you spend five or six years of your life writing a PhD that in the end, thirty people will read.

00:16:45: So it's also riding without a readership and literary writing.

00:16:49: I mean to the extent that the literary market functions still is the case that literally writing ideally has done for people who read it.

00:16:57: let us assume like the AIs we'll get better.

00:17:01: There are all sorts of problems with that.

00:17:04: The resources being consumed for making it more potent and powerful, but let's bracket them.

00:17:10: maybe Let us assume the proficiency in this text machine is going to get better.

00:17:18: as you said You can feed or train your personal agent With your own writing And then it emits a style.

00:17:27: Where does this lead literary culture?

00:17:31: I think that there's a huge resistance within literary culture and within the highest tiers of literature to AI.

00:17:41: If you're just listening this podcast, do not talk to me about AI, I will kill myself.

00:17:48: Like this is how many literary people feel- They

00:17:49: don't wanna talk about it

00:17:50: anymore because they've cut out their head?

00:17:52: It's also because its kind of an overwhelming subject.

00:17:54: right and you're like well i can't do anything about it.

00:17:56: And if your are my age or your age You have lived through pretty significant technological changes in your adult life Specifically the rise social media and a complete transformation Of your day today with your cell phone.

00:18:12: I don't want to do that again.

00:18:13: Like, you know what i mean?

00:18:14: We complain and resist social media so much And now it's just part of our lives.

00:18:20: People

00:18:20: are tired.

00:18:21: yet another technological

00:18:23: revolution.

00:18:25: This is not how the forces of history work or technology works at all.

00:18:30: But why couldn't we stop an email Email and Google Maps, you know I think a lot of people would agree.

00:18:38: these are things that definitely improved everyone's lives But like text messaging.

00:18:43: I don't even know maybe having a phone That you could call someone if your late?

00:18:46: That's kind of useful.

00:18:47: i don't Even know it and I love text messaging.

00:18:49: but like If we went back to Nineteen ninety nine yeah And had that level of technological advancement

00:18:56: peak of human

00:18:57: wouldn't be awesome.

00:18:58: It Would Be So Sick.

00:18:59: so Like I Think You know, you feel like it's just go.

00:19:05: Like again I can't do.

00:19:07: the social media thing is sucked It sucks Social Media sucks and there Is this kind of futile resistance?

00:19:13: I think among literary people especially because they don't understand how technology works And They Don't want to

00:19:18: learn.

00:19:19: But i think that is no.

00:19:20: but This is indicative Of a larger sort of Philosophical problem or over question Because its out There.

00:19:29: you Know AI is built in to various systems.

00:19:33: And it has been, frankly for a long time because the recommendation algorithms are also form of machine learning and so forth much less sophisticated that LLMs.

00:19:41: but

00:19:42: well I mean there's

00:19:43: been

00:19:45: no.

00:19:45: no since the fifties.

00:19:47: like this is Like It's Been in Development For A Long Time.

00:19:49: The Tendency To Anthropomorphize Think That A Machine Has Empathy For You Is Called The Eliza Effect Which Was Coined In The Seventies Or Something Like This.

00:20:00: So Like even like the idea of an AI therapist, which people are sort-of up in arms about as well.

00:20:07: And I know...I actually know people who use ChatGPT to work out their problems with their girlfriend before they have a serious talk with her boyfriend Which i think is incredibly perverse.

00:20:16: But all this isn't new at all but it's mass adoption and scale.

00:20:29: exponential scaling is new.

00:20:31: Yeah, because I think this actually goes back to like the core of the problem which is you can't trust these things on a mass scale!

00:20:42: You can trust them maybe with people who are sort-of developing them for good reasons.

00:20:47: but this cheating tool to like write a book.

00:20:56: There was the New York Times article of couple months ago that said there is a surge in self-published books and in book production, between twenty twenty four and twenty twenty five people self published a million more books than they had year before.

00:21:12: it's slop.

00:21:13: yeah I mean its also just.

00:21:15: someone said oh she uses AI right.

00:21:18: two hundred romance novels.

00:21:21: What's the point?

00:21:23: No one even likes reading.

00:21:24: Do you know what I mean, this is what i don't understand.

00:21:28: like no-one wants to read two hundred books in a year but it's a scale thing.

00:21:36: In an interesting way The problem for the price story of the Commonwealth prize Is analogous that any teacher or professor who was giving writing assignments.

00:21:46: and we have a piece coming up by the philosopher Lily who was at Yale and she writes about the use of AI in analytical philosophy.

00:21:56: And interestingly for me, I wasn't aware of that.

00:21:58: She says that I mean, analytical philosophy In particular has A negative attitude towards writing.

00:22:06: anyways Yeah,

00:22:07: if you've ever read any pieces by

00:22:09: Analyne

00:22:11: Callas incredibly dispiriting.

00:22:12: Yes i studied mostly Continental and it's awhile ago.

00:22:15: but So the idea is that it's all about making points and basically, yeah It's like a very formal egg kind of language.

00:22:25: And then the writing that happens in between the points you make.

00:22:28: Why not outsource to a writing assistant?

00:22:32: She's trying to put some resistance on what she feels an enormous amount of peer group pressure actually from her analytical philosophers who are all AI maxing And

00:22:46: I know some academics who are AI, they're not matching but their AI moderating.

00:22:52: AI

00:22:53: moderated?

00:22:53: Well that's what she is trying to do.

00:22:56: and then the other point of course it's that... She speaks also from a point-of-view with her teacher.

00:23:03: obviously any sort like homework exercise assignment which gives students can be most likely going to be an essay set will be handed in.

00:23:12: so forms evaluation.

00:23:14: Um, academic culture are also changing.

00:23:18: And if you want to make sure that people actually read and have digested in metabolize the points that their papers I'm making You need to evaluate them

00:23:29: Orally yeah or with a blue book right?

00:23:31: Yeah

00:23:32: Blue Book is in the U S of those.

00:23:34: It's like it's a hand written exam where you write an essay in there blues so we call them blue books.

00:23:38: Well,

00:23:39: that's how I did all my exams and

00:23:41: yeah.

00:23:42: I mean i did them in.

00:23:43: That's what?

00:23:43: I didn't in university as well.

00:23:45: um I also think like It's interesting because They're In My generation.

00:23:54: there was a real skepticism of using Google Like at All or Using like Wikipedia like you shouldn't be like.

00:24:01: There is this kind of like primacy about using like physical books And going to the library and using The Dewey Decimal System.

00:24:08: but the thing Is The people were still using it all the time, right?

00:24:12: And what happened was there was a style of writing in English at least that arose like during the twenty tens which was very heavily influenced by Wikipedia.

00:24:22: Yeah and It's just sort of encyclopedia Like Very broad general like clicking around kind of structure and also the style Of writing Was very straightforward.

00:24:33: maybe they weren't plagiarizing from wikipedia but They were writing as if for some kind of mass curious but fundamentally like not that literate person, right?

00:24:45: Even if they're describing really complicated concepts.

00:24:47: That even though the research was doing this rabbit hole thing where were bringing in all different kinds of weird unusual references and stuff.

00:24:56: And I'm not against that kind of research at all.

00:24:59: as long you evaluate what you produce using traditional fact checking means.

00:25:07: The thing I wanted to say is like people wrote, the way that technological advancement works was.

00:25:11: humans came up with Wikipedia.

00:25:13: They created a Wikipedia style of writing and a Wikipedia sort of network right?

00:25:18: And then people started using Wikipedia and they mimicked the things that humans had made which they have done.

00:25:24: so seem sort-of make it seem like robot or automatic Right.

00:25:29: So same thing will happen with AI.

00:25:33: I think writers like Sam Chris and I, and you know Senegal.

00:25:36: Like sort of there's a fatalism that surrounds the topic is That it's not just that everyone's using AI And It's going to be surrounding us like everything Is gonna Be written in this kind Of like manic like saucy tech world wait way.

00:25:52: But it's also that people are Gonna start writing like that naturally because they also don't read anything else.

00:25:59: So then, there are going to start writing like that and no one is gonna be able see the old way where we were.

00:26:05: like this sentence needs some kind of special verb with a unique relationship language.

00:26:12: you

00:26:13: know?

00:26:14: I do not know if it's most likely scenario... The analogy between Wikipedia and AI LLMs really works.

00:26:23: Because have you ever tried edit or create a Wikipedia article?

00:26:31: No.

00:26:32: The amount of friction that the pretty, extremely ingenious- It's

00:26:38: great system!

00:26:39: Wikipedia editing community has put into getting things in to actual pages.

00:26:43: it is so tedious and long.

00:26:45: there are many checks... You're being

00:26:47: a little bit German about this.

00:26:48: I'm not making you literal

00:26:49: And no but the point i want make Some in some way like all the human work that went into creating, maintaining expanding editing and all this friction.

00:27:05: That went into the Wikipedia article somehow shows in the end product.

00:27:10: And I want to be optimistic about the fact that Human readers will sort of like derive more value from that than From the effortless creation of infinite amounts of texts via

00:27:25: The thing about AI, or LLMs is that they are made to limit the amount of quote unquote work.

00:27:33: That user has to do and to limit them out.

00:27:36: have worked it?

00:27:36: They use her as a see right.

00:27:38: so there isn't no sense like the vast technological advancement that's behind these things.

00:27:45: It supposed to be frictionless formulaic in Union knows on some level you want to produce a text Hitting certain bullet points like it's doing likes, you know the has core values right

00:27:58: yeah.

00:27:59: So I think Like they're not this time.

00:28:02: No, I don't mean that They're the same but I mean at the principle of We get this technology and writers then mimic Yeah The technology kind of like You know?

00:28:11: I also use look if like I don' t really used Wikipedia But like, you Know i've certainly read a wikipedia article before but I try Really hard to write like That in that tone.

00:28:19: Lots of people don't feel that way.

00:28:21: Yeah, right?

00:28:21: Lots People are like.

00:28:22: this is how writing should be and they will feel the same way about The Writing That Is Produced by LLM's And it does make you Feel Like Incredibly Fatalistic Because You're Like I Don'T want to Be Confronted With This All The Time.

00:28:33: yeah

00:28:34: So Your The Editor Of A Magazine.

00:28:36: yeah

00:28:37: You Get a Lot Of Submissions.

00:28:38: You're Very Busy.

00:28:39: Yes You're Wearing A lot of Hats.

00:28:41: yes um and you still like you know as As Anger and scrutiny around this issue increases.

00:28:50: There's only more pressure to make sure that you don't publish something That has been plagiarized by AI, like does it bother?

00:28:58: You doesn't worry.

00:29:00: so if we go back to the Commonwealth story The Commonwealth Prize scandal Some people are sure that this is a I others didn't detect it And going forward We will see probably a situation where its becoming harder and harder To tell.

00:29:16: Now you're asking me as a magazine editor of Billion Review, what do I make if this?

00:29:21: How did I make sure that the stuff that's getting submitted is not written by AI.

00:29:25: I would say in specific case for the magazine we are doing... We mostly interested on a specific style like cultural criticism and writing And it's cases with lot of magazines.

00:29:38: You've been edited by number of magazines But also have certain idea about how to style So I think we are less likely to be duped by... We do occasionally also publish literary writing, but very few of it.

00:29:55: To a certain degree trying to iron out and streamline pieces in the editorial process as human editors is unfortunately more and more emulated also by AI.

00:30:13: Yeah, well I think it was something we didn't talk about.

00:30:15: is people who use AI as like a starting point right?

00:30:18: And then they edited what it produces to make it sound like more human or add their own twist on it.

00:30:24: That's actually much more common than a purely AI generated story.

00:30:27: I mean i will say that there are some non-native speakers of English whose working language in the other half isn't German.

00:30:40: the American or British, or English-speaking academic system also.

00:30:45: And so many people we work with and who are good writers in principle but they're not stylistically at a level as someone like you would be or person has done an MFA language.

00:30:56: So for that vast vast was actually majority of the english people working right now.

00:31:04: the use of technology to like make sure that at least what you're saying is not outright wrong or sort of standard using it as a simple proofreader,

00:31:14: yeah.

00:31:15: Is I think a different use case than giving just a few bullet points and then having right your text?

00:31:21: And also- The problem this really... It's a slippery slope Yeah.

00:31:26: And it is ultimately down to the responsibility or the work ethic, or whatever like.

00:31:32: that's a self conception of any writer how far they want to go in using the technology?

00:31:38: Well this I mean This is interesting because i was wanted to ask about translation.

00:31:41: right Of course anyone who lives In A country where They don't speak The native language

00:31:45: relies on translation

00:31:46: Is Relying On Google Translate and Techno.

00:31:48: Like In That Technology Is Getting Really A lot Better Than It Used To Be.

00:31:52: I Remember When You Would Use It.

00:31:54: It's

00:31:54: Wonderful.

00:31:56: Nevertheless, like

00:31:57: it's done.

00:31:57: I mean i can speak a lot about that because part of my professional life in the last uh fifteen years was also as a translator and the google translate is notoriously worse than the german startup deep

00:32:13: l yeah

00:32:14: and its also sort of an unspoken secret.

00:32:17: everyone in culture uses deep L for at least first drafts.

00:32:23: translations Of course, also do that.

00:32:26: But it depends very much on the nature and kind of text you're trying to translate.

00:32:32: how good The Deep El Draft then is I would guess at this point a majority of professional translators maybe not in poetry or literature but if its about functional texts non-fictional um, I don't know.

00:32:46: Um cultural institutions that have to produce catalog texts and all that kind of stuff if that stuff gets translated most of the professionals use D-Belt for their first drafts then they post edit.

00:32:56: And the question is Is that inherently problematic?

00:33:01: Does it change the way language works?

00:33:03: i think It does have an effect on The German Language Because in culture The work language Increasingly is English.

00:33:14: I think, stylistically.

00:33:16: our friend Vincenzo Lattronico wrote about this over a decade ago.

00:33:19: About international art and English as sort of like International style that has emerged.

00:33:26: And the thing with this kind of english is That it's very denotative kind of writing.

00:33:33: It's saying plain sentences.

00:33:36: If you machine translate that into German, the German that comes out of it is very dull.

00:33:42: Because it doesn't have usual syntactical variations in the German language where we can put other parts to the front and so forth.

00:33:51: So I think via machine translation from an already leveled down type of English this is affecting

00:34:03: The thing too, I think that it does create.

00:34:05: It still creates kind of like weird things right?

00:34:08: Even now

00:34:09: they're really

00:34:11: little not even glitches just sort in and sort of pretentious twentieth century magazine editing.

00:34:17: They would call them infelicities Right!

00:34:20: Just sounds a bit weird.

00:34:22: International art English is kinda like a lychee-like Weird Thing where if you are native english speaker And hear the example i heard someone use last Berlin Biennale was the positions that art takes.

00:34:36: We would never say it, but also not any non-native English speaker's fault like you don't know what we don't...

00:34:43: Maybe this is a retro contamination from German because this is what German curators always say they

00:34:50: don't speak

00:34:50: about individual artists

00:34:54: And, but I mean there's you know sort of and i'm like oh that's interesting.

00:34:56: where is it coming from?

00:34:57: It's kind of an interesting.

00:34:58: Like if your interested in language these kind of infelicities or mistakes That maybe are going to be ironed out by the use machine translation Are interesting for me!

00:35:06: I've always loved English as a second language.

00:35:10: German speak English like naturally.

00:35:12: I think it's really interesting.

00:35:13: It shows you a lot of things about your language and there's something also Interesting, which i wanted to bring up in the context of AI In general Which is that now many people are leaving spelling or grammar mistakes And for example their hinge profiles.

00:35:26: they're dating app profiles Also Like an application For Something.

00:35:32: You kind Of Want To Make It Clear Someway That You Didn't Use AI.

00:35:36: So You Put In A Mistake that AI wouldn't make as opposed to like a word usage one, like grammatical mistake.

00:35:43: That it just won't make and demonstrates you are human?

00:35:47: Yeah

00:35:47: there was also this wave of using an AI program artificially via AI, putting in some errors.

00:35:59: I mean it's crazy!

00:36:00: It's just funny...I also think if you're a literary writer and you write novel or fiction or something and want to use different ways people talk?

00:36:07: And you wanna write some dialogue for example like of course make mistakes the way they talk right.

00:36:14: Or we were gonna talk about this Ben Lerner novel which never talked.

00:36:18: but there is particular voice that can't really be captured.

00:36:25: It can kind of be captured by AI, but like not quite.

00:36:27: Not very well yet.

00:36:29: and part of that voice is the kinds mistakes people make.

00:36:32: And when I say oh i've always loved English as a second language it's because I really liked to see the lineage Of learning english in what the result right?

00:36:41: Those are particular mistakes That native speaker wouldn't make and shows you something new about language.

00:36:48: Yeah so how?

00:36:49: Okay Lauren You have been A published writer for ten or fifteen years, you grew up in a literary world where AI texts were not omnipresent.

00:37:02: But today the number of self-published books that has been on the market for last four years have exploded and there's so many more people now apparently trying to become writers And The Bottleneck is same as ever much more streaming into it.

00:37:20: So its likely going ever more difficult for the people who are serious about writing to even break through.

00:37:29: Do you think that is the case?

00:37:30: Well, yes and no because I think if your creating like my writing in particular a rather niche product That ideally- Like If You're A Literary Writer... You kind of corner small market.

00:37:44: Yeah, and like that's how you... You know there's people are reading magazine articles.

00:37:49: Sometimes they're still a viral magazine article.

00:37:52: If you saw five thousand copies of the book in your first year That is considered huge success in literary publishing.

00:37:58: specifically I think for long time There has been A category error In talking about publishing or books in general.

00:38:09: because Realistically, you know maybe I can trick some of the people who read self-published romantic novels on amazon.com Into buying my novel by using like a really sassy like sexy Synopsis and i'm not opposed to that.

00:38:24: Like if someone buys my novel accidentally thinking it's like Some kind of thriller And then they discover something They might think about It.

00:38:31: That' fine.

00:38:32: Also money is money.

00:38:34: Buy it?

00:38:34: I don't care.

00:38:37: I think you can think of it as almost like a local market, right?

00:38:40: Where the restaurant down the street doesn't need everyone from around the world to come there.

00:38:46: It just needs enough customers to sustain itself and be kind-of comfortable financially.

00:38:51: And i think that lots people who have this weird resentment There are lots literary writers or high culture producers.

00:39:02: They're resentful.

00:39:04: They have some imagined idea of like the past when everybody was reading.

00:39:08: and you know, actually I think The complaint that people aren't reading serious books has been around for a really long time And there's never been like a huge market for it.

00:39:17: You know?

00:39:17: Like where-like?

00:39:18: the worst case scenario For the literary novel.

00:39:20: as i see It is something like opera and like opera sustains itself through patronage.

00:39:24: Fine!

00:39:26: Its also very different from opera because the barrier to entry Is much lower but for both consumers & producers.

00:39:34: If I look at it from the German perspective, yeah.

00:39:37: Um i would say that Independently of AI something has happened and its to do with social media over The past?

00:39:44: I don't know.

00:39:45: five or ten years a lot Of people in the literary world aren't not as much writers As they are broadcasters so way To break through and become a published writer is actually first of all create A personality.

00:39:59: this can be done via magazine pieces but a lot of It is clearly a good performance on Instagram.

00:40:06: And then make yourself look so interesting that literary agents will approach you and beg you to write the book, it happens all the time!

00:40:17: Then they come to ghost,

00:40:20: there's another... What happens is these writers are not going name names but book because the book is one item that allows them or their publisher, or agent to monetize in a tangible way.

00:40:36: The attention they have... It's also

00:40:38: legitimizing force

00:40:39: right?

00:40:39: That are harvesting on Instagram and I think for the literary culture this has been terrible.

00:40:48: Luckily there was movement with it now but we've seen interesting Choices for literary prizes.

00:40:56: For example, last year's Buchpies was awarded to a really high-literary stylistically innovative

00:41:04: book at

00:41:04: the Hotel Menger.

00:41:05: I've

00:41:07: heard some bad things about it.

00:41:09: Well maybe AI was responsible for all the complicated Konjunktiv II that the book is written in but... ...I hope we were sort of like over this hill.

00:41:21: It goes back to, I mean my main question is like...I think the statues of written text are really changing and AI's contributing.

00:41:35: AI isn't only reason for it but i don't see any way in which you know this trend can be reversed.

00:41:43: Yeah!

00:41:46: For me personally what it has made me do And it's similar to like I work as a ghost writer sometimes and have to write these kind of boring straight forward, like memoirs.

00:41:54: Yeah

00:41:55: that then they don't always get published the way that I write them anyway doesn't really matter.

00:41:58: but basically Like It makes me much more experimental As a writer because i feel like i need To like Get it farther and further away from like A normal seeming piece Of writing doing formal experimentation specifically.

00:42:12: There is just you know AI can't do That.

00:42:14: maybe you can tell it to do that, and it can produce something that looks like that.

00:42:18: But I just can't really like produce... It can't produce a literary text that is making an argument through its form right?

00:42:24: They can't do that yet!

00:42:26: Yeah i mean..I just don't know how they could do that because that requires an ability actually read not just input output literature.

00:42:34: so For me, that's kind of fun.

00:42:39: And like there's a kind of the like.

00:42:41: anyone who is vaguely optimistic about this It's like oh maybe it means set.

00:42:45: The actual literary people will just get circling weirder and weirter.

00:42:48: I'm like the weirdness and experimentation Will be more rewarded because it so different from those status quo AI type

00:42:58: stuff.

00:42:59: Okay to wrap up sometimes often with recommendations We didn't want to recommend any...

00:43:06: I have a good book, actually.

00:43:07: You have a Good Book on AI?

00:43:09: The book i would like to recommend is Empire of A.I by Karen Howe who's a tech journalist did a lot of work on open AI and AI companies for many years.

00:43:21: And then produce this book.

00:43:23: because I do think if you want to understand how the technology is working, What's at stake from a capital perspective?

00:43:40: And what are the motives of these companies, like how're they operating?

00:43:45: and she does really good job explaining that.

00:43:49: The kind-of eyebrow raising anecdotes one loves to hate reading.

00:43:56: yet I've, in preparation for this episode.

00:44:00: I went back to read Nick Bostrom's book Superintelligence from two thousand fourteen which sort of like launched the discussion about the singularity and super intelligent AI taking over world control.

00:44:14: And what's interesting about his book is that it's mostly actually a theological discussion Which reminded me of philosophy class classes on Descartes and on Emanuel Kant.

00:44:25: Maybe we should just go back to Descartes, actually?

00:44:28: I think it's worth doing because there are a lot of problems that he tries solve which is what we're confronted with now.

00:44:35: but another thing...I think the question we could discuss in an other occasion was whether AI can be funny.

00:44:42: Can't It Be Funny?

00:44:44: There's this guy on Instagram who really, really funny whose handle i cant remember so ill look at up.

00:44:48: But you've probably seen his videos where he has this really flat affect.

00:44:54: He's a blonde guy and here is what happens when I tell AI, I'm being hit by car And then the AIs like oh no!

00:45:03: You know?

00:45:03: Like it was just really funny conversations with the AI, that they obviously is not intending.

00:45:10: It can't intend to do anything but it's not supposed to be funny.

00:45:14: But the disconnect between what you expect from like a super intelligent robot and their results which are often moronic The impossibility of understanding creates humor And its still human.

00:45:28: using this tool for creating humor I mean if this is an interesting question like in order to be funny Like do you have to be intentional?

00:45:35: This is why it's not.

00:45:36: Oh, you know how children are funny.

00:45:38: They don't know like they're making a sex joke But they don't understand the innuendo and think that just saying something And your adults are laughing Is That The Same Thing?

00:45:50: We'll find out.

00:46:06: essay by the writer and researcher Ines Marchi on image manipulation and propaganda.

00:46:11: It's called AI Uglies!

00:46:14: You can find Ines' Essay in all our reviews & podcasts episodes at BLNReview.de Wherever you're listening, please take a moment to subscribe to Airlift.

00:46:24: Give us a like and positive review – it actually does help.

00:46:29: This episode was edited by Kate and Robert I'm Tobias Harbourkorn editor of Billion Review.

00:46:34: See ya next time.