I recently read an interesting blog post by Maggie Appleton titled The Expanding Dark Forest and Generative AI. It begins by describing the negative effects that generative AI will have on the quality of writing and information on the internet before segueing into exploring how humans on the internet can prove that they are not language models. I largely agreed with Ms. Appleton’s takes from the perspective of a human being who produces a fairly high volume of writing without any use of generative AI, but I had some disagreements with the premise or framing of the second part of her essay while nevertheless simultaneously agreeing with much of the advice. Below, I will explain where I agreed, my disagreement with framing the issue in a way that suggests that people who write original articles or produce original media without being aided by language models have the burden of proving otherwise, and considering some of the meritorious advice in her article from the perspective of my alternative, still human-centric, framing.
Agreeing about the negative effects of language model-generated writing
I will start with where I share common ground with Ms. Appleton’s argument. I recommend reading her article in full and studying this useful image she put together showing the layers of the internet.
Ms. Appleton described the rapid progress being made by generative AI and opined that “[t]hese models became competent copywriters much faster than people expected…” I largely agree with the caveat that one’s mileage with “competency” in this context may vary. Most of the explicitly generated content I have seen – setting aside the examples where the AI something or other “hallucinates” – remind me very much of some of the articles I was reviewing in my two years as an editor for my high school newspaper. (Take as you will.) Ms. Appleton also discusses image generation, but I am less interested in that aspect here so I will set it aside (I have written about Midjourney’s censorship in a different context).
After describing some hopes that people have for the benefits of these language models, Ms. Appleton hits on the downside for the quality of writing on the internet:
You thought the first page of Google was bunk before? You haven’t seen Google where SEO optimizer bros pump out billions of perfectly coherent but predictably dull informational articles for every longtail keyword combination under the sun.
Maggie Appleton
(Note: “SEO” stands for “Search Engine Optimization.”)
This is entirely correct. Ms. Appleton later notes that the SEO spam situation had been bad enough when humans were doing it without AI assistance. I agree fully that being able to ask a bot to generate SEO gibberish is already not improving the situation. She adds a bit later:
We’re about to drown in a sea of pedestrian takes. An explosion of noise that will drown out any signal. Goodbye to finding original human insights or authentic connections under that pile of cruft.
The majority of pre-AI internet writing did not often make me think that original thoughts inspired what I was reading. This of course is one reason I use an RSS/ATOM feed collection to have original thoughts and reporting delivered to me instead. But, to be sure, if we had an original thought shortage on the internet before AI, AI will not help matters (I dare say it is not helping matters).
In the main, I agree in general with Ms. Appleton’s views on how AI-generated gibberish will affect the internet in full, granting I may have phrased one thing or another differently. One reason I never got into the subject beyond a few humor-oriented posts is because there is not much to say about AI blog spam that is not obvious. I will venture that most people who take the time to read the idle musings and essays of a random internet writer from Brooklyn are generally in accord in their sentiments against the current trends.
But what follows from AI blog-spam?
Disagreeing about the reverse turing test and the burden of proof
Before continuing, do note that I am not reading Ms. Appleton’s essay as declaring that human internet writers bear an actual burden of proving their humanity. I read what she wrote as a way to set up an interesting question about how we can prove our humanity while writing online. However, for reasons I explain below, I still disagree with her framing and think that there is a better way for internet writers to think about how to write originally for human beings.
Ms. Appleton begins her next section, which describes what follows from the general concerns she raised in the first part of her essay:
Our new challenge as little snowflake humans will be to prove we aren’t language models. It’s the reverse turing test.
Maggie Appleton
I did a double-take here, but I kept reading. She explained why she reached this conclusion:
[W]e will become deeply skeptical of one another’s realness. Every time you find a new favourite blog or Twitter account or TikTok personality online, you’ll have to ask: Is this really a whole human with a rich and complex life like mine? Is there a being on the other end of this web interface I can form a relationship with?
Maggie Appleton
I do not disagree with the above quote (other than noting that you should not have any favorite TikTok personalities because you should not be using TikTok, which should be banned) – but my concurrence is qualified. There have been many instances where I have seen a headline or snippet and suspected that it came from a blog spammer with limited command of English who may have been aided by a language model. However, in general – this suspicion arises because the content is unoriginal and obviously written for bots. People have claimed they can create novel language model-aided content, but I remain unconvinced for the time being. Ms. Appleton then leaves us with a query:
Before you continue, pause and consider: How would you prove you’re not a language model generating predictive text? What special human tricks can you do that a language model can’t?
Maggie Appleton
This is where I disagreed with how Ms. Appleton framed part of her essay, specifically “[h]ow would you prove you’re not a language model…” Before I get into the particulars, I will introduce my perspective as someone who writes a bit on the internet and does not use AI.
Digression: On internet people saying internet stuff
I use Koko Analytics – an entirely local page view tracker for WordPress – to gain a general idea of how many visits the site and individual posts receive. Koko Analytics sometimes, but not always, picks up on referrers. Occasionally through this process or through an occasional stroll through my Google Search Console links information (note I do not use Google Analytics and I have no Google scripts on site – Search Console has information picked up by Google’s web crawlers), I find when people have linked to my site. I will follow the links and see if I can find who linked to us and why they did so. I do not have too many external links and mentions in the grand scheme of things, but I have occasionally found some interesting thoughts or notes about my writing by keeping tabs on things.
Similarly, I have occasionally come across misinformed takes about my site – by which I do not mean disagreement but things that are outright wrong. My favorite example is from a small web forum where one random poster speculated that I could be trying to create a “blogging empire,” luring poor innocent internet lambs with references to “privacy” before slaughtering them for my nefarious commercial ambitions. If I recall, this line of logic was based on the fact that I post in more than one place and occasionally write about online privacy and alternatives to big tech, but I do not have the link on me.
There have also been rare examples of criticism. For example, when my essay on understanding doped cycling results made page one of Hacker News and generated some discussion, my friend pointed out that one commenter called my article on reconciling cycling’s Armstrong era boring and he noted humorously that the commenter sounded like a joy.
Some people who post stuff on the internet are sensitive to ill-informed takes or criticism. But I do not care. In the end, I run The New Leaf Journal and I write about what I want to write about it. If some random person on some forum somewhere thinks I am running a blogging empire based on an article about the issue with “alternative” search engines relying on Bing, this person clearly did not read the article carefully or take the time to read what my site is. While I personally do not understand why someone would want to advertise his or her paranoia on the internet, it ultimately has no effect on me or how I run my site. Conversely, if someone does not like my article – all the power to him or her. I come across many articles, including ones that become much more popular on Hacker News than my cycling essay, that I am not interested in for one reason or another (I have never taken the time to chime into a Hacker News thread by announcing that something is boring, but everyone has their own hobbies).
However, while I do not chase people with bad internet takes or different opinions, even the worst takes can inspire someone to do better. For example, while I am confident the blogging empire person did not take a serious look at my site, it played a role along with other notes in inspiring me to re-write our About Page more like I write our regular articles. I reflected on some criticism of my headline choice on another one of my Hacker News page 1 articles, a review of /e/ OS, that I had not thought about when I chose the headline. After conducting a bit of research and thinking about it, I concluded that my headline was fine and I disagreed with the critiques, but I thought that the issue was interesting enough to turn into a new article. I have occasionally criticized or disagreed with articles on the internet from The New Leaf Journal – see for example my critique of hiding behind one’s own political opinions behind amorphous experts (I doubt I caused a sea change at the Huffington Post, however).
Rejecting the reverse turing test
Let us return to Ms. Appleton’s comment. To the best of my knowledge, no one has yet accused me of using AI in furtherance of my hidden dark blogging empire schemes. But let us assume arguendo that someone, somewhere, has read one of my articles and thinks that I outsourced it to an AI chat-bot or an SEO farm in India. I would humbly suggest that someone who accuses me of hiding behind an AI language model after reading a few of my articles and studying our About Page is probably an idiot. But why do I need to prove my humanity to a random idiot on the internet? There are many idiots in the world, much less the internet. I have better ways to occupy my time than worrying about how to prove my humanity or my benign intentions to the small subset of idiots on the internet who happen to stumble upon The New Leaf Journal.
But let us be charitable and assume arguendo that the language models start producing decent high school newspaper-level writing on original topics and that it is possible for someone who is not a bona fide idiot to suspect that I am an AI language model.
Even in the charitable case, I still reject the premise in my own context. I personally pay for the server to run The New Leaf Journal and I spend uncompensated time working as site administrator in addition to writing the articles. I am not asking for anointing oil (bonus points to anyone who gets the reference without clicking) or applause. I run The New Leaf Journal because I want to. If you enjoy my writing or if it makes you think (whether you agree or disagree), I am glad that you found it valuable in some small way and I am always interested in constructive feedback, regardless of whether I happen to agree.
To the extent I write some articles that some people out there apparently enjoy or otherwise find informative, I know for a fact it is because I take time out of my day to draft these original articles in Ghostwriter, edit them, and transfer them to WordPress for formatting and publication. To the extent I owe anyone anything for my uncompensated time, all I owe to my readers is doing my utmost to make sure everything I publish is worth the reader’s time.
But beyond the foregoing – do I have the burden of proving that I am not a language model?
No – no more than I bear the burden of proving that I am the sole administrator and editor (not sole writer, see my friend Victor V. Gurbo’s articles too) of The New Leaf Journal. I let my writing, photos, and quality art speak for themselves. If someone chooses of his or her own free will to believe that my history of Evacuation Day, a report on a fallen stop sign, niche visual novel reviews, in-depth anime hair color coverage, or my essay complaining about formerly being blacklisted by Bing (which, unlike random internet people, did actually negatively affect my project), that is a choice they make of their own free will. I will continue to run the site as I see fit and the mis-coded NPCs can continue to yell “THIS IS AI” at every essay they open on the internet before going back to TikTok (which should be banned) or Instagram.
For anyone considering putting original writing on the internet – take a lesson from someone who is approaching year four. You cannot control what other people think. You can no more control whether people falsely impute bad motives to you than you can control whether they will enjoy your writing or other media content. You can certainly learn from criticism and praise and we should always look to improve, but in the end – try to simply focus more on doing producing your best work under whatever your circumstances are and less on trying to control what random people on the internet think. Your writing or other multimedia content should, for better or worse, be your entry into the random internet people debate. If you are sensitive about people thinking you are a language model, follow me in not using language models and producing original work. I will venture that more people than not will not assume you outsourced your work to ChatGPT or an SEO farm if you did not in fact do so.
Re-framing Original Writing Prompt
Having noted that I disagreed with how Ms. Appleton led into her tips, I offer an alternative way of thinking about the issue. Human beings writing original content or producing media should not think of needing to pass a reverse turing test. They should instead think about how to produce distinctly humane writing (or media) on the internet. By humane, I mean by humans, for humans. You, a human being, are creating something from your own perspective for other human beings. This concept is more useful than focusing solely on language models because it is entirely possible for human beings to create inhumane content for consumption by search engines (e.g., articles full of SEO keywords for the primary purpose of ranking on Google) or to prompt readers to action instead of to inform (e.g., review articles that exist primarily to convince readers to buy products through affiliate links instead of providing informed first-hand opinions of the quality of the products). Those who are interested in producing humane content should consider how to write articles and produce media that stand out in a good way in a sea of manipulative gobbledygook.
Framing aside, Ms. Appleton presented good ideas and points of emphasis for producing unmistakably humane – as in by humans and for humans – internet writing. These points are as applicable to my re-framed point of emphasis (producing distinctly humane writing and media) as they are to the original reverse turing test idea. Below, I work through Ms. Appleton’s ideas and provide my own spin from the perspective of someone who has been running The New Leaf Journal for almost four years.
Never forget that you are a human writing for humans
Ms. Appleton notes that language models cannot reason like a human. I agree entirely. They follow rules for producing “content” from specific data sets. While these rules are (probably) written by humans, the bot itself is not human. Ms. Appleton notes that the language models “cannot go outside and touch grass.”
(Touching grass is actually a great way to come up with new article ideas. Be sure to go outside and, if possible, take some long walks.)
Ms. Appleton then explains how we can use the limitations of language models to prove our humanity:
This leaves us with some low-hanging fruit for humanness. We can tell richly detailed stories grounded in our specific contexts and cultures: place names, sensual descriptions, local knowledge, and, well the je ne sais quoi of being alive. Language models can decently mimic this style of writing but most don’t without extensive prompt engineering. They stick to generics. They hedge. They leave out details. They have trouble maintaining a coherent sense of self over thousands of words.
Maggie Appleton
To be sure, these limitations of language models are shared by the SEO spam bloggers on the internet. I think I can put the moral here succinctly:
You are a human being. Write articles from the perspective of a human being for other human beings.
Nicholas A. Ferrell
This is not to say you must ignore search engines. I include meta information with all of my posts and sometimes write about things that people may be interested in searching for. But all the while, I have never forgotten that I am a human being. If a human being lands on my article while using Google Search of the like, I hope he or she comes away thinking the following:
Wow. This Nicholas A. Ferrell is a human being. As a human being, he wrote an article at The New Leaf Journal for human beings just like me. I will now use my agency as a human being to add the feed for The New Leaf Journal to my human-curated feed collection so I can make sure to receive human being Nicholas A. Ferrell’s humane articles as soon as they are published. Wow, there are also human feeds for other projects like The Newsletter Leaf Journal (surely this is a human-created newsletter for human beings) and The Emu Café Social? This inspires in me, a human being, a distinctly human-sense of wonder. Let me do the human thing and add these feeds of human-created writing to my human-centric feed reader.
Example of the organic thought a new visitor to The New Leaf Journal would have
Challenge a language model to come up with that organic sentiment.
The process
Ms. Appleton encourages us to “[b]e original, critical, and sophisticated,” opining that “[l]anguage models spit out text that sounds like a B+ college essay.” (I did not expect her to hit on the lowering of standards at our colleges – I will avoid that topic so we do not lose our way.) I agree with Ms. Appleton that being original, critical, and insightful are all desirable, and I also agree that this is not easy. There are only a few writers I follow who include a genuinely original thought in every article. I must concede I am probably not among the rare few. I do not claim to be a Lee Smith, albeit to be fair I write much more often than he does and I do not think he is going to expand into goose sightings or Openclipart for the foreseeable future.
I would summarize this point as being one that puts process over results. Learn, read, and touch grass. Be interested in the world around you and when it comes time to write, think about your topic carefully and try to approach it in an original way. If there is a popular topic about which you have nothing original to say, consider a different topic that you are interested in. When you write, think clearly about your topic: Clear writing flows from clear thinking.e
Quirks
Ms. Appleton encourages internet writers to use “neologisms, jargon, euphemistic emoji, unusual phrases, ingroup dialects, and memes-of-the-moment” to signal their humanity. She compares this to teenagers subverting their elders.
I am not entirely onboard here, at least as the point is presented, but there is merit. For example, I compared a hypothetical person who actually reads my articles and accuses me of outsourcing to a language model as being a “mis-coded NPC,” which probably qualifies as jargon, an unusual phrase, and ingroup dialect. But while I will occasionally do that, I do not prioritize it – perhaps not least because I am not big on internet socializing and am not part of any internet cultures, much less sub-cultures.
(With that being said, I have maintained for years that I could be a top-flight internet meme lord on 4chan or X if I applied myself to that great endeavor.)
I offer an alternative recommendation for people who write often. Develop your own original phrases and lexicon. For example, I regularly applied the term “survive and advance” to describe certain visual novel choice structures. While I borrowed survive and advance from other contexts, I think my application to visual novels is uncommon if not unprecedented. After using it enough, I decided to define it in its own article.
There is no problem with borrowing common phrases from elsewhere, and it may make more sense in the context of specific sites and styles than it does for me. I would only caution making sure that your memes and phrases are not so “of the moment” that your article will become unintelligible in the next moment. But even if you borrow more from the internet cultures and sub-cultures than I do, consider applying your creativity to developing your own original recurring concepts and phrases.
Verification
Ms. Appleton speculates about the possibility of having “institutional verification” of humanness before waving the idea away as being unlikely to solve the problem.
I agree that “institutional verification” of humanness is an unwieldy concept. However, there is an alternative solution. Human beings can vouch for other human beings using classic web methods such as blogrolls or webrings or by simply linking to other interesting articles in their own writing. For example, unless I explicitly say “I am linking to this article because it is terrible” (which I have done before), you can assume that I am probably referencing an article in my own post because I think it is informative or even worth discussing. Now Ms. Appleton certainly does not need me to give her site more attention – granting I am confident her site is much more well-known than mine – but if you regularly read me and are not familiar with her work, and you are reading this article, I assume you took the time to read her human-generated essay too. If you found it interesting, maybe you even decided to read more of her work.
I always make a point of citing my sources and inspiration, and I make an extra effort to highlight particularly interesting articles on independent websites. See, for example, my recent article on organizing RSS and ATOM feeds, which shares some terrific posts from small web sites, a couple of which I subscribe to. If you like and regularly follow my writing, you would probably be somewhat likely to take recommendations from me about other websites. At the very least, you can rest assured that I will not encourage anyone to follow SEO blog spam, whether it is generated by a human writing for bots or an actual bot.
Meet real people
Ms. Appleton notes that people can prove their humanity “by showing up IRL with our real human bodies” before noting some challenges in certain context.
This approach does not work for me because I am not a big fan of people. If you like my writing, consider this good news for you. How would I have time to write all of these articles if I spent my time with real people? I dare say one “party” and I would be out for at least half a week.
Do not let 3D people come between you, a human being, and your human-created website.
(Lest any human being readers or bots are confused, I submit for the record that this section is mostly a joke.)
Conclusion
Between human-created SEO spam and language model blandness, the internet, as seen through the lens of big tech search engines and proprietary social media, looks a bit bleak. But there is plenty of original writing and media on the internet for those who care to look and search intelligently for it. For those of us creating original writing, I agree with Ms. Appleton fully that it is important to consider the effect of a new class of inhumane content that is becoming more common, albeit I frame the issues differently than they are presented in Ms. Appletons‘s ssay. Having written from the perspective of someone who writes on the internet, I have reached conclusions from the perspective of someone who reads internet writing. I am a critical reader. When I read an author’s take in a news article, I consider all the perspectives that may have informed the take. When I read something that appears to having been written more for Google than for a human or more to call to action than to inform, I consider why this may be the case. When I see something that looks like a poor 10th grade essay, I consider whether a language model or an actual 10th grader is responsible. I fully encourage critical reading. But at the same time, there is a fine line between being critical and being paranoid. A critical reader considers the quality and source of what he or she is reading. A paranoid reader assumes that everything is SEO spam or a language model (for a good example of the distinction, look up some disputes on the true meaning of clickbait and reach your own conclusions). One way to make the internet a better place to read critically while keeping internet-induced paranoia in check.