AI Is Making Me Lie: One Writer’s Opinion — Do You Agree?

AI Is Making Me Lie

Artificial Intelligence: Controversial and Criticized

Guest author L.N. Hunter provides his observations about AI. See also a couple of his other guest posts:

How Critique Sites Can Help Any Writer, Amateur or Pro

What Do Americans Have Against the Letter “U” Anyway?

L.N. hails from the UK; therefore his grammar and spelling is based on UK English.

AI Is Making Me Lie!

Actually, the title’s a lie: submission calls which ask me to confirm that no AI was used in the writing of my work are making me lie, but that’s too long for a title. I always say yes to those things, even though — shhhh — I did use AI.

Before you wallop me with the permanent ban-hammer, let me explain …

Here’s one example of the assertion people are required to make: ‘I submit that I did not use AI or machine-learning tools to generate part or all of this work, and that this work is fully my own human creation.’

My intention is to talk about AI in this article, but ‘fully’ also has me worried: I use a computer for my writing — does that mean my work is not ‘fully’ human created? Taking it to an extreme, can I even use a pencil, or do I need to create one from scratch [1]? Some tools are acceptable and some are not — but the word ‘fully’ isn’t helping.

I feel that the term ‘AI’ is being bandied about in the same way as ‘chemicals’ in some parts of society: we don’t want no goldarned eeeevil chem-i-kals in our food. I want to show that an outright ban on AI makes as much sense as a ban on chemicals, but that we don’t fully understand what this thing called AI is.

Technicalities

There’s a technical difference between AI and ML (Machine Learning), and what we’re really discussing is the application of ML to creative writing. Since people are using the term AI, let’s just stick with that rather than getting sidetracked into the details. I’m not going to say much about the technology itself, just point out that the term AI covers a lot of ground.

Starting with the basics, I’m sure a lot of you use a spellchecker. Those used to be rule-based (e.g., rules to stem words in order to reduce the size of the dictionary required, important when computers were small) — that’s AI. More recent checkers use models built up by scanning vast tranches of text and compiling statistical models, and all but the simplest of statistical systems are labelled as AI. Some of that is no doubt sales hype and jumping on the bandwagon. Regardless, I think it’s safe to say that automated spelling and grammar checkers use AI [2]. Maybe I should revert to physical paper dictionaries just to be on the safe side.

But wait a minute, what’s running on my computer? I see I have something called Intel Dynamic Tuning Technology, which uses ‘advanced AI and ML-based algorithms to enable … optimizations for performance, thermals, and battery life’ [3]. What’s more, uninstalling it is discouraged as ‘Disabling Intel DTT may void the system manufacturer’s warranty, cause damage to the system, or risk higher external chassis temperatures than designed’ [4]. So, there’s an AI process running in the background, doing something I don’t understand to maintain the health of my computer — this means AI is involved in everything I do on the PC. Maybe I should write on paper. (Or use Linux; hmm, I’ll stick with paper.)

Research

As with many people, a lot of my writing involves some research, such as searching the web to find references for this article. Web search engines use AI. I suppose I could use material like the Encyclopedia Britannica instead, assuming I can find it in paper form.

No spellchecking, no computers? Surely that can’t be what publishers and editors intend, so let’s look at what they might really intend.

Perhaps it’s what AI’s being used for? Thesauruses (thesauri?) are useful beasts, but like spellcheckers, your favourite online one is likely to be using AI. Ah, but surely replacing a single word or short phrase is OK, while writing the whole story is verboten. Nonetheless, I note that the checkbox requires me to state I’ve used no AI at all …

Maybe the issue is nothing to do with AI, but is all about plagiarism, basing your work on material ‘stolen’ from others? Many existing online dictionaries, thesauruses, grammar checkers, etc., are as much trained on ‘publicly available’ material on the web as LLMs (Large Language Models — a form of generative AI), and therefore also based on ‘stolen’ work. Ah, you say, pirated books have appeared in LLM training data, but even if ‘Big AI’ manages to expunge the blatantly purloined material [5], will that make such systems more acceptable? I suspect not.

We’re Getting Nowhere

What is the complaint really about? Is it about stealing jobs? With all these AI-generated stories cluttering up online bookstores, we humans will struggle to make a living. It may come as a surprise, but few writers make a living from their writing. Yes, the situation’s worse with all these AI ‘authors,’ but is it really way worse? I make so little, alas, that while my income would reduce, it would make little difference (a fraction of roughly zero is still roughly zero). Money aside, I do have my first publication under my belt, which I acknowledge will become trickier for new folk with the overall number of submitted works increasing.

Perhaps it’s about homogenous, garbage writing? AI is going to improve — will it be acceptable when the quality of its output gets better? Probably not. As an aside, many humans can write garbage too. Looking at things from the reader’s perspective, is a good AI-generated story a better thing to have than a poorly written human one? That’s an argument for humans improving their skills, not for banning AI. In other domains, humans and AI are working together, getting better results than either alone [6] — maybe that could apply to writing. (Are we on that road already, through the use of spellcheckers, etc.?)

I view this as similar to when cheap electronic instruments appeared, disrupting music creation, or how phone cameras are democratising video production and letting people broadcast themselves. (Allegedly, TikTok is responsible for the poor quality of modern music [7]!) According to Sturgeon’s Law [8], 90% of everything is crap, so that means a lot of rubbish is produced thanks to these advances in technology — AI is little different.

Workload Reduction?

Something users of AI in writing are definitely doing is swamping editors and publishers with huge amounts of material, the vast majority of it being low quality. Perhaps the ban on AI is specifically intended to reduce their workload, which is perfectly fine and reasonable, and I guess they’re using AI as a proxy for ‘Don’t send rubbish stories.’

Is it simply the technology? Old rule-based systems and statistical models are OK, but LLMs aren’t? (The details of how LLMs differ from previous ML systems isn’t important here, just that they are more powerful.) That’s perhaps the closest we can get to specifying what we don’t want, but there are (and will be) other technologies that do the same thing. And LLMs can make for great thesauri or reverse lookup dictionaries — if other sorts of AI are fine for those, why not LLMs too?

Where Does This Get Us?

There is a fear of AI in writing (and in other creative spheres), which is perhaps justified. As a means of dealing with it, a total ban is unreasonable. If some AI is acceptable, such as spellchecking, what might the limit be? Perhaps the typical 10% leeway on word count applies here. I think the problem is that we can’t fully articulate what we want to ban, but I believe that trying to nail it down would be a mistake, as that will lead to an arms race: automated detection of AI-generated writing sounds like a tool that would help, but that’s a bit circular [9]. Despite their problems, solely to reduce the slush reading workload, maybe one definition of AI writing is whatever a publisher’s tool detects, giving the additional concern about being erroneously ban-hammered if my writing is misclassified.

So, after 1,000 words, I’m no further forward. AI writing is something we don’t like, but we’re not sure what it is — the closest definition of what’s bad that I can come up with is: LLMs used for directly creating significant parts of a story, but once again, I note that’s not what publishers are asking me to promise not to use. Whatever ‘AI’ is, I continue to lie and tick the ‘No AI was used’ checkboxes.

Note: no AI was used in the creation of this article. Besides a spellchecker, web searching, and whatever the Intel doodad is doing on my computer. In other words, yes, I did use AI, the same as in just about everything I write.

References

[1] https://www.wikihow.com/Make-a-Pencil

[2] https://www.online-spellcheck.com/

[3] https://www.intel.com/content/www/us/en/support/articles/000058479/graphics.html

[4] https://www.intel.com/content/www/us/en/support/articles/000090464/graphics.html

[5] https://www.searchenginejournal.com/are-chatgpt-bard-and-dolly-2-0-trained-on-pirated-content/485089/

[6] https://www.csiro.au/en/news/All/Articles/2023/July/collaborative-intelligence

[7] https://www.spectator.co.uk/article/its-official-modern-music-is-bad/

[8] https://en.wikipedia.org/wiki/Sturgeon%27s_law

[9] https://www.longshot.ai/blog/ai-detectors-accuracy

© L.N. Hunter

About L.N. Hunter

The Feather and the Lamp

L.N. Hunter’s comic fantasy novel, The Feather and the Lamp (Three Ravens Publishing), sits alongside works in anthologies such as the BSFA award-winning Best of British Science Fiction 2022, as well as several issues of Short Édition’s Short Circuit and the Horrifying Tales of Wonder podcast. There have also been papers in the IEEE Transactions on Neural Networks, which are probably somewhat less relevant and definitely less entertaining. When not writing, L.N. occasionally masquerades as a software developer or can be found unwinding in a disorganised home in Carlisle, UK, along with two cats and a soulmate.

Get in touch via https://linktr.ee/l.n.hunter or https://www.facebook.com/L.N.Hunter.writer


Are You Interested in Word Lists and Writing Tips?

If you haven’t done so already, please subscribe to my blog. (The link will take you to the subscription widget at the top left of this post.)

I usually post two to five times monthly, and you can discontinue your subscription at any time.

Find thousands of writing tips and word lists in
The Writer’s Lexicon series
and additional resources on my Facebook page.

Please don't be shy. Leave a reply.

Your email address will not be published. Required fields are marked *

17 thoughts on “AI Is Making Me Lie: One Writer’s Opinion — Do You Agree?

  1. Regarding. Artificial Intelligence: Controversial and Criticized

    New technology, AI, is to be managed, not feared.

    When new technology is accepted by the populace, it becomes unstoppable. Those who stand against new technology, over time, travel down a path from contradiction to hypocrisy, and finally to absurdity.

    Example: In the mid-1800s, symphony music by composers such as Bach, Beethoven, and Brahms was exclusive to the elites—those who had money and were considered the right sort of people. When the phonograph came to market, symphony music could then be heard by the masses.

    Conclusion: I see AI as a democratizing force for creative writing. AI offers a free expression of story ideas that might otherwise be lost. Just as the phonograph set a higher standard for music playback and inspired thousands to learn the art, AI can inspire and elevate creative writing. Did music programs exist in schools in the 1800s? No.

    My solution: The publishing industry should establish categories for traditionally authored works, AI-assisted works, and AI-generated works.

  2. I agree with it. I use ProWritingAid to spell and grammar check my work. I allow it to make rephrase suggestions which I sometimes use although usually I take both my original and the suggestion to see if one is better than the other and then combine the bits I like. And that, to me, is the thing… I keep control of my work at all times so the work I produce is basically mine.

    Anyway, what’s the difference between that and submitting your work to a professional editor who then spell and grammar check it as well as makes improvement suggestions? When my book was first published, the editor suggested I start it with a high action chapter so I did… all my work but that suggestion wasn’t mine; it was his.

  3. It’s okay for publishers to make rules around submissions where the bulk of the work has been machine-generated, which seems to be a significant issue that is upsetting folks. But a blanket ‘no AI’ is confusing. Publishers may need to define what they regard as AI.

  4. It’s nudging *my characters* in directions I don’t want my want characters to go. I laid out the plot, I laid out the map that I want my characters to follow, and it’s pointing them in directions I don’t want them to go. A.I. in a pain.

    • Are your characters sometimes a pain, too, Tom? I find that mine point me in directions I don’t want to go. But I hear you. I don’t use AI in any part of my creativity process.

    • This sounds like one of those cases where you need to engineer the prompt to give you the answer you want (to be more precise, to *not* give you the answer you don’t want) – but the effort involved in doing that could be as much as writing the text yourself.

      What you could perhaps try is asking it to generate a dozen different ‘next steps’ and pick the one(s) you like, then ask it for more like that. (Just guessing – no idea if it’d really work.)

  5. I haven’t heard it expressed more clearly. Even writing this short comment is triggering ProWritingAid to make suggestions. So it seems I can’t add a comment to a well written, thoughtful blog post without technically using a form of AI. (Although I ignored its suggestions – so maybe I can. But would it have been better if I hadn’t?)
    Well said LN. And hello from Gretna Green

    • Hi, Linzi. Thanks. I’ve never tried anything like ProWritingAid or Grammarly. I prefer to rely on my inner voice to guide me.

      As I look at the second sentence in the preceding paragraph, I see that WordPress doesn’t like my spelling of ProWritingAid. And I give WordPress the raspberry.

    • Thank you for those kind words.

      (I hate SPaG checkers in a browser window – they’re fine when I’m writing something substantial, but I’ll do that in a proper tool. Who cares about spelling and grammar in a short note like this one?! I’ve even turned off the limited checking my browser offers. Bah, humbug.)

      I must visit Gretna Green sometime. Thus far, I’ve only passed by on the way to other places in Scotland. Still, I’ve only been in Carlisle a few months, and until recently the weather’s been too grim to want to visit anywhere!

    • The term artificial intelligence covers a huge area and range of techniques. Early SPaG tools embodied a series of rules (e.g., capital letter at the start of a sentence), and such a collection is a form of AI. More modern ones learn from examples, which is a different form of AI (machine learning really). LLMs, which are the current ‘bad guys’ can be used for SPaG, though I’m not sure how many purely SPaG tools do use it – regardless, if they’re better at fixing typos than traditional AI tools, would we really want to ban them?

      That’s the sort of sublety I’m trying to get at in the article – not all AI is bad, and not all LLM uses are bad. (There’s still the issue of ‘stealing’ copyrighted text, but that is not specific to LLMs.)

  6. In case it’s not clear, I’m quibbling with the wording of no-AI requests, and saying little about whether AI is a good or bad thing.

    I saw another one this morning which started off talking about spcific tools: ‘We do not accept or publish work that has written with, or with the assistance of, writing software such as ChatGPT, Perplexity, etc.’ – though ‘with the assistance of’ is overly inclusive. It then went on to ask for ‘a statement indicating that you have not used AI to create your work’ – I guess ‘create’ is a good word to use to express what they intend. This one’s probably the best attempt I’ve seen so far.

    • I’ve started to add “100% Human-Generated” to the front matter of my books, but it’s like saying something is 100% non-GMO. GMOs are in our environment, and cross pollination, etc., have probably resulted in our unwitting consumption of genetically modified organisms. We can say we’re 100% opposed to GMOs as we shove a forkful of [fill in the blank] into our mouths. Are we being deceitful if that forkful contains an unidentified GMO?