AI and Book Prizes > Likes and Comments
I guess the obvious solution would be to create an entirely own class, The Booker AI Prize (for example). It would be like the Olympics with Performance-enhancing drugs ... it is thing coming, isn't? Those interested could follow, and the rest of us, leave it be. Is this the right place to discuss how do people spot if AI is used? Would I see that just by reading, or is a tool of some sort needed?
I just read the book Seven, and I was thinking the language at the start of the book was really, really bad, but kept on reading and it dawn on me that the writer might have played me, suggesting the book was written by AI (not that it was, just pretending it was). Is that how one can see if AI has been involved?
Paul wrote: "A lot of book prizes, including the one of which I'm a Trustee, are wrestling with whether and how to circumscribe the use of AI. Particularly after one part of the Commonwealth Short Story Prize w..."I have some questions if I may:
1. How do we know if a book is written by AI? Given that AI-detection tools aren't always reliable, what would constitute convincing evidence of AI use?
2. How should a prize handle an author who claims that AI was only used to “edit” their book?
3. Should an author be required to provide drafts, notes, outlines, version histories, or anything of that sort to demonstrate that the work is genuinely their own? And if so, what should happen if an author refuses to provide such evidence?
4. If an author's work wins a prize and is later proven to have been AI-generated, should the prize be revoked? If so, should it then be reassigned to another shortlisted work?
5. Should an author who deliberately lies about AI use be banned from entering the prize again?
6. How much evidence should be required before a prize can disqualify someone for AI use? In particular, how do we distinguish reasonable suspicion from sufficient evidence?
7. I promise this is the last one!! How do we protect genuinely human writers with unusual or highly formulaic styles from being falsely accused of using AI?
Going back to the Enhanced Olympics analogy currently it feels like Book Prizes are trying the equivalent of saying that athletes cannot take (or should pretend they are not taking) any supplements at all - no vitamins, no protein powders etc. And a second comment Prizes disqualifying books because the cover was AI designed - which has happened a few times - seems silly to me.
Agree with the posts above. AI is a tool and it is going to be used, but to what degree and consequence is yet to be determined. What you are ultimately dealing with is potential confrontations with publishers. I think a blanket statement that supports human authorship should suffice and then take each case individually. The blanket statement should be vague enough to cover your actions in case a book is found to be AI generated but also to deter complaints of AI use from outside sources, (other jealous authors or someone just looking to troll) I think the goal should be to show enough support to appease frightened authors but respect inevitabilities, and to avoid inflating the issue. Another thought might be to generate a category for books that admit AI use but that is probably not possible.
Yes I don't think an AI category works other than for prizes with multiple categories. The rules that Galley Beggar's prize have seem fairly reasonable to me - I suspect not restrictive enough for a lot of people.
I suspect answer to most of the questions is you rely on an attestation from authors or indeed implicit in entry. And if it is later discovered they did use AI beyond the permitted extent you simply let it slide - the alternative of hounding people for their mistakes does not end well as we've sadly discovered in the UK recently.
I think Galley Beggar or the Booker blurb are fine. I find the UK Society of Authors one a little permissive, mostly because "research" and "outlining" are quite a broad church, and seems to imply that the only creativity in a book is its actual prose.I've seen a few lit mags say they have gone the route of asking authors to provide notes or drafts, and interviewing them to see if they can answer basic questions about their own creative work, and some stories have been cancelled on the back of this. I would favour this as a method of "checking" over running the work through an AI detection program.
Though realistically if anyone uses the internet for research, they are likely using AI now (indeed you have to actively switch it off on Google). And if they don't use the internet in 2026 what are they using?This is a good article proposing a loose 10 point scale for AI
https://www.the-hinternet.com/p/an-ai...
He has level 0 for AI use as
0 — You actually do live in a cave, or have found some other way to go about your life entirely unconnected to the internet or to any media that indirectly transmit habits of language formed on the internet. If that is the case with you, paradoxically that also cannot be the case, since by definition you cannot be reading this.
Where would people draw the line on AI use for a prize? NB the author of that article is pointing out that AI language is creeping into general use even at level 1 or 2.
On issue of checking - the approach I'd prefer for a Prize is trust.
Paul, the three policies you posted have in common that they cannot be independently verified. There are supposedly "AI detectors" out there, but as of now, they are not reliable. Those that do exist often produce false positives. So, I think the dialogue needs to be around what do we ask people to declare, what evidence can we demand, and what happens when someone lies.
The Booker's rule is a statement of values disguised as a rule. There are many tools out there (even before gen AI existed) that offer synonyms, antonyms, rhymes, definitions, spell checks, grammar checks, and more, and I'm not sure where to draw the line. For example, is asking for a different way to say something really a breach if the original thought came from the human?
What about AI-assisted research? This seems perfectly reasonable if someone is writing a historical fiction, for example, and wants to get the history right. Is it ok as long as they don't copy the result? The Booker rule seems to disallow any use whatsoever.
Galley Beggar's is the strongest because of the right to request additional information about how the work was made, plus the power to disqualify, withdraw, and reassign. Drafts, notebooks, version history, and a writer's ability to talk about their own decisions are the only real evidence available. A prize that has reserved the right to ask for them in advance is in a vastly better position than one improvising after an accusation is made.
However, I think all this checking should be done before any publicity makes it obvious that the author is being "investigated" for possible AI usage. Think about the damage it would do to a reputation if is even hinted that AI was involved. And if the investigation comes back clean, it seems like a horrible penalty for someone who didn't use it but has been "accused" (especially given the knee-jerk reactions of social media these days).
Also, accusations could be made in bad faith by rivals and in good faith by people misreading one of those unreliable AI detector outputs. A good policy should say what standard of proof applies, that detector output alone is never sufficient, that the process is confidential until concluded, and whether/how the appeals process would work.
The SoA scheme is the most honest. It explicitly permits brainstorming and outlining (and really what's the difference between doing this with AI and doing it with other humans. Just look at all the Authors' Notes at the end of their books, thanking a cadre of people for helping their work come to fruition.
Brainstorming or outlining is where it would be valuable to focus the trustees' discussion. Obviously, the use of AI to write entire books is out, but what about as a tool to help them think (and is this really so much different than discussing it with your spouse or first readers or friends or editors?)
A couple of stray thoughts that are not in the rules you provided:
The Commonwealth case happened in an open-entry competition, which is the hardest environment to police. It might help for those publisher-submitted prizes to require the publisher to sign off in addition to the author. Publishers have contracts with their authors, commercial relationships to protect, and their own reputations too. That is far more enforcement leverage than a tick-box from an individual entrant.
Another thought: it seems important to say something about the judges adhering to the same gen-AI policy. If the prize' is meant to be human judgment on human writing, then the judges should not be using gen-AI either (not sure this happens, but it could happen as it becomes more available and commonly used).
Consider disclosure instead of prohibition. A declared-use model asks entrants to describe any AI involvement rather than swear there was none. I would hope you would get truthful answers instead of absolute denials (and what good is a rule that everyone ignores?). Preserve discretion to judge case by case, and a false declaration becomes a clean, documentable breach rather than a contested interpretation of "substantive."
The first thing this question made me think about is how one would write a book about AI (AI as either a character or chatbot or whatever) and not use AI in some fashion to write part of your book (e.g., you're portraying an interaction between a human and AI, or AI and AI--like Moltbook which is like Reddit only for AIs to use). You would want the AI portion of said book, even if fictional, to closely resemble real world AI interaction/presentation/language. I guess you could model it off of previously release AI output/exchanges, but would that then count as "using AI." As Paul already mentioned, one's research (if online) is most likely already using AI. Of the examples presented above, I suppose I favor the International Booker's rule/guidance statement. It feels like authors should disclose whether AI was used for research, brainstorming, outlines, etc. without being disqualified.
Can AI lie? If not, then we could just ask AI if AI was used in this manuscript, problem solved ;)
Marc wrote: "The first thing this question made me think about is how one would write a book about AI (AI as either a character or chatbot or whatever) and not use AI in some fashion to write part of your book ..."I loved the novel Do You Remember Being Born?: A Novel by Sean Michaels which is about a poet hired by a tech company to interact with their AI and to co-write a poem with it, and I thought it was a delightful examination of the problem of authorship, and the author was very upfront about the AI side of the conversations being generated by/inspired by AI.
I keep thinking of things.... I've been reading a lot about AI and its history recently. I think a rule written to be true in 2026 will look either quaint or absurd by 2031, which means whatever policy is adopted should be reviewed regularly. As fast as the AI landscape is changing, I'd say each cycle merits a review.
Also, anything phrased as "no use of gen AI" ages badly because the tools get absorbed into everything (as you say, it's now the default for web searching and who knows what's next).
Anything phrased around authorship, that the expression is the entrant's own and they can account for how it came to be, ages much better, because it's the same test prizes have always applied to ghostwriting and plagiarism.
Thanks all - useful thoughts. Keep them coming.
I must admit I would tend towards zero restrictions and judge the quality of the output.
That's certainly an option, maybe even the best option since it would be hard to prove anyway. It would save a lot of time trying to figure out a rule.
Watermarking should help with more reliable detection, at least.
I have to say that as a reader, I'm pretty uninterested in engaging with things that even might be AI-written. I spend enough of my day job wading through them. I would be disappointed by a prize not making at least a basic effort to ensure actual people are involved.
What is the actual problem statement here? Why is it ok for privileged authors who publish either with big publishers or the very hands on small presses (Galley Beggar would be a good example here) to make extensive use of a human editor but that writers with other small presses or seeking to get their work known and without access to human editors, not to be able to use the same AI tools that are increasingly a normal part of most people’s jobs and personal life (I sometimes think I am the only one of my friends who does not use ChatGPT almost as a matter of routine).
Tend to agree but the literary world is very anti AI. In part as LLMs are giant plagiarism machines - the stories of books being bought, scanned, fed to AI for training and then pulped aren’t helping.
Although when one talks about authors’ influences, is that qualitatively different, even if on a more modest scale?
Also we do tend to like human competition. We don’t allow unlimited PEDs in athletics or indeed machines (I could easily win the 100m at the Olympics with my car). So for book prizes do we want limitations on mechanical aids, so we are judging purely human creativity, even if for wider books to read we may not?
I feel like if you have a free-for-all policy on AI you're going to end up essentially judging the literary value of LinkedIn posts. Or watching an Olympic event between robots.Both of those things could happen, but where is the interest for me as a spectator?
Why though re LinkedIn the quality of those is terrible so why would they make a prize list. There are without the help of AI a glut of low quality genre fiction but it’s not even submitted for literary prizes.
If AI starts helping make books that actually are better quality then that’s different thing.
But the robot point is excellent - I guess I would miss meeting the author so would not be keen on a book where there was no author to present it, talk about it … but one using AI as a tool to write a better book seems different.
Gumble's Yard - Golden Reviewer wrote: "But the robot point is excellent - I guess I would miss meeting the author so would not be keen on a book where there was no author to present it, talk about it … but one using AI as a tool to writ..."I get it and I know these tools will be used progressively more, but I'm skeptical they'll result in better books. Most AI writing has quite a flattening effect, pulling things more towards the norm...which I feel is the kind of writing we all bemoan a lot here and the kind prizes like this one were set up to provide alternatives to.
I wonder if we can say the Internet has resulted in better books? Research is easier, and it's easy for aspiring writers to access videos and blog posts with the kind of editorial tips that should even the playing field--but has that happened? Are books better than 30 years ago? And for that matter, are most books bad now in ways that AI is likely to solve?
Obviously these are big questions, but the fact that we (societal we, not Mookse) haven't answered them makes me want to be wary about inviting the robots into the fiction game.
I agree with you, Emily, about AI writing. There’s this idea that the writing will get better but I think the “fail” is written into the method from the beginning. LLMs are all about pattern matching. So there is a kind of fiction that can be produced that matches the patterns of lyric, metaphor-heavy prose, but the metaphors have the quality of madlibs choices. I can see such a story being read favorably by a prize committee the same way “found art” can be appreciated by a human being. Put a couple of old hubcaps on display in a museum and they become meaningful because human beings bring meaning to it.And on the other side AI will be able to write commercial trash, the equivalent of AI slop, and may even reach the level of passable fiction but it will always be pattern-based. It’s baked in.
The thing that is strangely compelling now about it is that it is so good at mimicking something real but I think people will adjust very quickly to seeing the artifice and rejecting it as inauthentic.
Gumble's Yard - Golden Reviewer wrote: "Why is it ok for privileged authors who publish either with big publishers or the very hands on small presses (Galley Beggar would be a good example here) to make extensive use of a human editor but that writers with other small presses or seeking to get their work known and without access to human editors, not to be able to use the same AI tools that are increasingly a normal part of most people’s jobs and personal life (I sometimes think I am the only one of my friends who does not use ChatGPT almost as a matter of routine)..."First, wow, that’s interesting that everyone is using ChatGPT. I have played with it a little and its answers are so syntactically repetitive that it drives me crazy.
The reason AI will not help writers who don’t have access to human editors is that AI will by definition have a set of rules about what is “good literary fiction,” and those rules will bend against innovative and unique and groundbreaking writing, every time. It’s by definition what AI does--looks for patterns in the past and applies it to the writing in front of them.
The second reason it is a trashy editorial tool, genuinely damaging and to be avoided at all costs by new writers, is that it is by design highly syncophantish. Even if you tell it to be harsh it will disobey and come back with “this is good, this is good...” or at best it will tell you it’s bad but when you say “no, you’re wrong” it can’t help but default to agreeing, "yep, you’re right to push back, it’s really good.”
It’s not just useless as an editorial tool. It’s genuinely harmful.
"It’s not just useless as an editorial tool. It’s genuinely harmful.”Okay this is disturbing to me, this is exactly the kind of oratorical flourish AI will write at the end of its little answers to questions! Exactly this!
Which is another thing that scares me about it. We are reading AI generated stuff all the time now and the rhythms are going to seep in and change how humans write.
Y'all make me feel like I'm the only one not directly using AI (obviously, it's already employed in my map/driving direction apps, my social media algorithms, my online search results, etc.). I believe I've asked ChatGPT one question and I did generate a dozen images from Midjourney a few years ago. I'll likely have to start using it for work in the next year or so, but it just generally feels like a technology being forced upon many of us and there seems to understandably be a strong backlash among artists, writers, musicians, etc.If it is as bad as many of you say it is in terms of writing/editing, than why would anyone want to use it, much less encourage its use for contests?
Do we know of any writers we like or respect who've admitted to actively using/embracing it? I know Olga Tokarczuk said she uses it for research.
Yes but they would not say in public. And it’s really not that bad as the above would imply.
The key to AI use is intelligent and iterative prompting. And you can of course easily train it on your own writing
I still avoid using AI wherever possible, but I am retired so nobody is forcing it down my throat.
lark wrote: ""It’s not just useless as an editorial tool. It’s genuinely harmful.”Okay this is disturbing to me, this is exactly the kind of oratorical flourish AI will write at the end of its little answers ..."
That article I posted above would say you have now obtained level 2 :-)
2 — You occasionally use ChatGPT for narrowly focused tasks, like finding out how the luggage-storage system works at the Düsseldorf Bahnhof, but never with the intention to share with others, under your own name, anything the LLM tells you. Over time, however, you begin to detect new idiosyncrasies in your own habits of expression.
Marc wrote: "Y'all make me feel like I'm the only one not directly using AI (obviously, it's already employed in my map/driving direction apps, my social media algorithms, my online search results, etc.). I bel..."It has many uses. Different solutions address different challenges and nothing is useful for everyone. Some are a form of whiteboard - helping keeping authors organized across a long-form writing project. There's a tool called, “Brainstorm,” to help authors come up with character names, magic systems, and locations. Nonfiction authors might use Claude to assist with integrating references. Multiple tools assist with grammar, style, clarity and structure (ProWritingAid, Hemingway Editor, Grammarly). I've heard that Sonix can be useful for analyzing audio interviews and recorded (oral) histories. Many of these tools simply automate tasks that authors used to spend hours doing manually, and may have little to do with the sentences in the literary work they ultimately produce. They are productivity tools. They may also reduce the amount of time that editors have historically spent flagging narrative inconsistencies.
Ai tools are also a big help with the marketing end of things - that which many authors abhor. Drafting social posts, marketing copy, book descriptions, queries.
I'm not pro- or -con-AI, but think those of us who care greatly about its use are best served by becoming familiar with the various solutions and not lump them all together as Satan's Toolkit. If an author saves time using chatgpt to transcribe handwritten notes into a file she can edit, we can quibble over whether we think it's prudent for her to give her content to chatgpt to help train it, but I can't see how we'd object on principle. (unless one's principle is, never use any trained LLM.)
Thanks, that’s a helpful perspective Carol. I think there will be those who would not want to use any of those tools but that feels more a political than an artistic choice. And book prizes run by charities (Booker, Women’s Prize, Queen Mary etc) aren’t really allowed to take political stances outside of their specific objectives.
Paul wrote: "Thanks, that’s a helpful perspective Carol. I think there will be those who would not want to use any of those tools but that feels more a political than an artistic choice. And book prizes run b..."
There will, indeed. I anticipate that the prize committees will find it difficult to come up with principles that accomplish their goals and make sense, and that are capable of enforcement.
I think we need to be careful not to use the PED analogy where it does not hold up. Namely, there are tests that can detect prohibited PED substances, where there is no such test for the use of gen-AI. A lab testing analysis for doping has a known false positive rate that can be driven very low, at the risk of missing abusers. Detectors for gen-AI haVe the reverse. They can only catch more by flagging more innocent people. Unless this changes, I think the PED analogy falls apart. Also, doping evidence is a physical sample, sealed, split into A and B, held by an accredited lab, contestable in arbitration. What is the equivalent for gen-AI? There is none.
By setting a moral position with no ability to enforce, what is the fallback? The best I can think of is requiring those submitting works to keep evidence of their process such as drafts, notes, version histories, correspondences with an editor, and an author's ability to talk in detail about why a something got cut or added.
The more I think about it the more uncomfortable I get with dictating the author's creative process.
What I'd suggest staying away from at all costs is an accusation method of investigation. Anyone can have a multitude of reasons for accusing someone, and we all know social media is a snake pit in this regard.
More reliable tests are coming, assuming the watermarking works as expected. Anthropic have already started putting it into their outputs. It won't withstand heavy editing, but it's something.I agree that an accusation-based approach won't work at all. By the time the falsely-accused are 'exonerated', it will be too late. More reliable detection, enabled by AI companies, is essential. Not just for this issue but for wider problems of misinformation.
Personally, I am profoundly uninterested in AI-generated writing. The very fact of human authorship is part of what makes books worth reading. Editorial involvement from a human, to me, is not the same, because that is a relationship between humans. There is of course a wide and blurry line of what 'AI use' means - I am definitely against any AI-generated text making it into manuscripts, but appreciate that AI-enabled tooling is hard to escape now.
I also think it's fine for this to be a moral stance.
Paul wrote: "That article I posted above would say you have now obtained level 2 :-)Ok it’s deeply disturbing that I can’t decide for sure whether I would have written that oratorical 2-sentence summing up the way I did, anyway, and AI just uses this common device, OR if I’ve been influenced even by the “messing about in boats” level of testing of the tech I’ve been doing the last few weeks.
What I’ve done is insert a passage from, say, Faulkner, or Sebald, and ask the bot to evaluate the writing. It has been discouraging to say the least to see the bot “workshop” these excerpts, the way maybe a really bad freshman composition teacher might critique student work.
I don’t think an AI bot can be reliably trained to a person’s writing, either, or that it can be expected to get better at anything useful, or do more than reflect back what the writer/asker knows already. What it is really good at is sounding good!
If I ask it leading questions and contradict it, it will come back with a reasonable response, but it’s mostly giving me what I implicity asked it to give me by the way I phrased my comments or questions. It follows every sniff of a leading question. Its usefulness is about is good as an astrologist’s usefulness. I went to an astrologer once and she was wonderfully helpful in a skill that I think a good therapist also must have. I think bots and astrologists and therapists maybe all work best on the same principle, of self-reflecting guiding.
I also have found it very difficult to reliably turn off either the syncophancy or the chattiness/mimicry of a warm editorial voice, which it does very well.
Oh at the same time my husband a lifelong programmer has been using an AI tool experimentally and, even if programming is the one area where AI seems to be undeniably useful, what he feels is that the bot guides him to a single answer/approach really quickly, where he, with his experience, can imagine 5-6 programming approaches that might be better. He can waste a lot of time guiding the bot to the right answer or just not. He is using a paid level of the AI where I just muck around in free space but given how AI works through pattern matching I think the problems are insurmountable when it comes to evaluating and advising a creative endeavor.
lark wrote: "Oh at the same time my husband a lifelong programmer has been using an AI tool experimentally and, even if programming is the one area where AI seems to be undeniably useful, what he feels is that ..."I agree that using it to write code or creative sentences has all of the perils you describe. I'm a transactional attorney and I've yet to encounter a product that is better than horrid at writing the sentences with which I make my career, primarily because the products sold into my industry - like all other LLMs - predict the next word based on volumes of inputs from poor training data. If the training data was restricted to well-written sources, they might threaten my job, but until that occurs...
What some products are good at is project management writing - synthesizing or condensing content from multiple files, creating tables from multiple files, including consolidating multiple calendars and the like, creating itineraries, drafting letters telling merchants to stop charging a bank account for a service I didn't procure, cover letters to be read by AI agents - the sort of daily life tasks that don't merit the level of creativity or time many of us allocate to them. And that don't result in creative works. Translations are creative works. TedX presentations are creative works. I'm not suggesting that only novels "count." I don't want to read AI-produced versions of any content to which someone is attaching their name and claiming moral rights as an author.
I don't think most readers care if a nonfiction writer uses an AI tool to help her prepare an index or reference table, aside from despising the fact that LLMs exist. In fact, if it saves an academic author 50+ hours of effort on the parts of her manuscript that aren't the book itself, we might see more woman authors get more non-fiction books published and begin to course-correct the gendered imbalance of male NF authors being substantially more often read by readers of any gender. But I digress.
From what I've been reading about AI in general (more than just gen-AI LLMs), we are only at the tip of the iceberg of what is possible. I'm not saying it's all good, but there will definitely be benefits coming down the road (e.g., in medical research, environmental research). We can complain and throw out many anecdotes of where it produces less than satisfactory output now, but until there are studies, those are just examples with no real weight behind them, and are often impacted by the viewpoint of the person offering them.
Mega-companies have spent vast sums of money developing AI to this point, and I think we need to figure out a way to use it well. The odds of derailing it at this point are unlikely to non-existent.
I will say that the Not X, But Y construction lark has mentioned is part of logic and rhetoric, which the models were trained on. The problem is that it takes a good concept that should be employed where it makes sense and uses it constantly, which just draws attention to the technique and waters down its effectiveness.
I work with lots of AI writing, as I work in business communications and business people love it. The Not X, but Y thing is a minor problem.The bigger problems are excessive length, a tendency to present everything as earth-shattering, repeating the same idea in different words ad nauseum, a tendency to lose the plot a bit and stop saying anything of importance after a while, while still sounding good, a strong tendency to make up examples and extrapolate ideas to the point of not resembling the idea any more.
You can tell it not to do these things, but generally it still will and because it always sounds good, you have to spend ages picking through it and questioning it. It is significantly harder to edit than imperfect human writing.
What it is good at is finding random examples of things from around the internet (that need to be heavily fact-checked but still it is often nice to have them) and proofreading. And translation. And writing a random nonsense introduction so you don't need to stare at a blank page.
Like a lot of things, it sounds amazing when you don't know much about the topic, and starts being problematic when you do. For laughs, I've occasionally asked it to write something in a "literary" style and oh dear.
I also haven't had much luck training it on "my" style. I tried with a series of twenty letters I had to write, but all it did was copy my precise sentence structures and sub in new words that sounded sycophantic and trite.
I also just find it weird that we don't expect our writers to...actually write. If you are a published writer you should know how to write by yourself, edit yourself and discuss and defend your work. We expect professionalism from other professions, why not artistic ones?
Is there another profession not now using AI extensively - if so then I think it will soon be an ex profession? And I feel like a lot of the limitations being described are a prompting issue and/or use of non frontier models.
Carol wrote: "I don't think most readers care if a nonfiction writer uses an AI tool to help her prepare an index or reference table, aside from despising the fact that LLMs exist."
But then the problem becomes how confidently AI will deliver wrong information. So we get stories about a freelance writer who wrote a “most anticipated books coming in 2026” with made-up books and US government agencies giving presntations where theor map of Africa has all the countries labeled wrong. There is no foolproof way of fact-checking the accuracy of AI generated info other than doing the work yourself.
I think a lot of the current issues with gen-AI will be resolved in future releases/versions, just as it may be possible to come up with a reliable "AI-detector." By the way, I hope I am not coming across as in favor of "expecting writers not to write". For me, the main issue is prohibiting all use of AI, even for brainstorming ideas or saying something in another way (at the phrase or sentence level). I think it can be helpful for those things. I do not think we (at the individual level) can stop AI from becoming ubiquitous, and future generations are likely to find it a normal part of their daily lives (just as the internet is for us now).
It's impossible to prohibit it for brainstorming and I agree it has some value there. But the examples above about having it design a fantasy writers magic system--they should really be doing that themselves. What is the actual point otherwise?
Gumble's Yard - Golden Reviewer wrote: "And I feel like a lot of the limitations being described are a prompting issue and/or use of non frontier models."Maybe, but people understandably get cranky about having to spend as much time prompting as they would writing, for a result they have less ownership of and that atrophies their brain, while essentially providing a knowledge transfer to dodgy companies.
Gumble's Yard - Golden Reviewer wrote: "Is there another profession not now using AI extensively - if so then I think it will soon be an ex profession? /..."My granddaughter, a first grade teacher, is not using it. When computer skills are taught to later grades, some version of AI or AI awareness will need to be taught.
Joy D wrote: "From what I've been reading about AI in general (more than just gen-AI LLMs), we are only at the tip of the iceberg of what is possible. I'm not saying it's all good, but there will definitely be b..."I doubt there will be any peer-reviewed studies because, what would be the point? studies of what? The performance data on each feature of each solution is collected by its owners and shared with the public via Gartner and the like. But also, within companies/customers using various solutions to process their own data, including their prompts. Almost every application and platform we use has Ai features and functionality. Hallucinations are endemic, but driven by user prompts.
lark wrote: "Carol wrote: "I don't think most readers care if a nonfiction writer uses an AI tool to help her prepare an index or reference table, aside from despising the fact that LLMs exist."
But then the..."
Yes,
The International Booker has added a very simple statement to its rules:
All books must be written by a human author and translated by a human translator/s. Books that include any text written using Artificial Intelligence technology are not eligible.
Galley Beggar's short story prize has a more detailed rule (Sam from the Press is very anti-AI, but his use of 'substantative text' is interesting):
We do not want to see AI generated work. Entrants must be able to warrant that they have submitted creative work conceived and written by themselves where structure, character development, plot, conception and narrative voice originate directly from themselves and that AI has not been used to generate or rewrite substantive text in their submitted story. Should concerns arise regarding authorship, the organisers may request additional information from an entrant about the creation and development of their work. The organisers reserve the right to disqualify any author who is unable to provide satisfactory information about authorship and originality and also to withdraw awards or publication and to reassign prizes where appropriate.
The UK Society of Authors also has an even longer self-certified 'Human-Authored' designation:
The text of the work was created by you as sole author or together with other human beings and that it is your own intellectual creation as the result of the exercise of your skill, labour and judgment.
The text, therefore, represents the expression of your creativity as the human author and it has not been produced by the use of a generative Artificial Intelligence (gen AI) model eg by the use of prompts.
The use of gen AI or tools with AI-embedded features does not disqualify a work from being ‘Human Authored’ under this scheme, provided that the tools have been used for assistive purposes only eg the use of gen AI-powered tools like word processing programmes, spelling and grammar checks, for research, brainstorming or outlining, or any other use that does not generate text.
What are people's thoughts as to what a Prize (including the prize I'm involved with) should do? Particularly Prizes that value literary experimentation.