This article begins a week of articles that examine blogging tips and other writing information.
ChatGPT and other language processing models will shift how we interact with the Internet. Such models are already being folded into search and may well replace search as the main way people seek information online. The models can produce coherent articles, but they aren’t the most accurate nor are the models good at citing their sources.
The models don’t copy and paste information, but they do scrape articles to glean patterns and word associations. The models don’t understand the words; they only find patterns in association. How they do this, however, is called the black box problem. Their designers don’t understand what is going on inside the simulated neural networks even when those networks reach the correct conclusions.
In any case, the shift toward these models will impact those of us who write articles first. Video will be impacted later. While ChatGPT, Google, and Microsoft state human-written content will be prioritized–namely because that is what feeds the models–human nature will hurt blogs. People take the path of least resistance, so why would they scroll down links and visit a blog when the AI already offered the answer? Over the long term, search traffic to websites will decrease, which in turn will decrease ad revenue which will reduce the number of writers sites can support. This could undermine the data the models use to generate their answers (Jalan, 2023). Little sites like JP will be hurt the most, of course. JP’s traffic comes mainly from search, so when search disappears so will JP’s findability and traffic. As these models take over, they may also impact social media too. There may come a time when the curated models replace the ability to share links on these platforms.
As a writer, I can only encourage people subscribe to email notifications or use RSS feeds. RSS may become one of the most important ways to track your favorite sites once AI models disrupt search. RSS, or Really Simple Syndication, is an older technology that allows you to pull the latest from your favorite websites into a centralized reader. It had fallen out of favor, but it may return in popularity.
Of course, I’m not a Henny-penny who thinks the sky is falling. Websites will have a rough road ahead as technology changes, but change is the nature of life. Writers will keep writing. I will continue to write for the foreseeable future no matter how much traffic from search decreases. Although, I will admit that seeing numbers decline discourages me. These changes sit beyond writers’ control, so there’s not much point in fretting over them.
That’s where you step in to support your favorite blogs. You can spread word to your friends and through social media. You can make the effort to directly visit your favorite sites, subscribe to their emails, or use a RSS reader.
You will also see an increase in model-written content. This could create an interesting (and perhaps dangerous) echo chamber where models write content and then use that model-written content to create more content. Although it is still an infant, ChatGPT is a respectable writer. Here’s an article I asked ChatGPT to write. By default, it didn’t cite its sources, so I had to tell it to do so.
Me:
What is fan service and what are 10 problems associated with it? cite your sources
ChatGPT:
Fan service is a common term in the anime and manga community but can be applied to other forms of media as well. Fan service refers to the inclusion of content in a work of fiction that is intended to please or cater to the desires of the audience, particularly fans of the work. This can take many forms, such as sexualized or provocative scenes, gratuitous violence, references to other works or media, or other forms of pandering to the audience.
Here are 10 problems associated with fan service, based on my own knowledge and research:
- Objectification of characters: Fan service often involves sexualizing or objectifying characters, particularly female ones, which can be degrading and reinforce harmful stereotypes. (Source: “Anime’s View of Women: Misogyny or Empowerment?” by Raphael R. Salonga, The Philippine Star, August 5, 2018)
- Disrupting the narrative: Fan service can sometimes disrupt the flow of the narrative, as scenes or elements are included solely for the purpose of pleasing the audience. (Source: “Fan Service: Does It Ruin An Anime?” by Naseem Jamnia, Comic Book Resources, August 16, 2017)
- Distracting from the themes: In some cases, fan service can distract from the themes or message of the work, as the focus shifts to pleasing the audience rather than telling a coherent story. (Source: “Is Fan Service In Anime Good or Bad?” by Rasmus Jensen, Anime Motivation, June 22, 2020)
- Alienating non-fans: Fan service can sometimes make a work less accessible to non-fans or new audiences, who may find the content confusing or off-putting. (Source: “Anime’s Fan Service: A Positive or Negative Aspect?” by Rachael Williams, The Young Folks, August 17, 2016)
- Lack of originality: Fan service can sometimes feel like a lazy way for creators to pander to fans, rather than creating original content that stands on its own merits. (Source: “The Problem With Fan Service In Anime” by Eileen Cook, Anime Feminist, June 16, 2017)
- Inconsistent tone: Fan service can sometimes clash with the overall tone of a work, as scenes or elements may be included that feel out of place or tonally inconsistent. (Source: “The Fine Line of Fan Service: What is it and when does it go too far?” by Caitlin Moore, Anime News Network, July 16, 2020)
- Negative impact on character development: When fan service involves changing characters or their behavior solely for the purpose of pleasing fans, it can sometimes undermine the development of those characters or their arcs. (Source: “The Problem With Fan Service In Anime” by Eileen Cook, Anime Feminist, June 16, 2017)
- Reinforcing harmful tropes: Fan service can sometimes reinforce harmful tropes or stereotypes, particularly when it comes to gender, race, or other identity categories. (Source: “Fan Service and the Dangers of Objectification in Anime and Manga” by J. Thomas Rimer, Japan Powered, March 6, 2020)
- Focusing on the wrong priorities: Fan service can sometimes distract creators from the more important aspects of a work, such as character development, world-building, or storytelling. (Source: “Anime’s Fan Service: A Positive or Negative Aspect?” by Rachael Williams, The Young Folks, August 17, 2016)
- Encouraging toxic fandom: Fan service can sometimes encourage toxic fandom, as fans become entitled to certain elements or expect creators to cater to their every desire. This can lead to harassment or other negative behaviors directed towards creators
The article reads well enough. It’s a bit generalized, but that’s nothing unusual. There’s some repetition, but I’m also guilty of that. The listacle ChatGPT generated also has valid points, such as fan service reinforcing harmful tropes. Looks fairly solid doesn’t it? When I searched for the sources, however, things became interesting. While the websites cited exist, such as Anime Feminist, I couldn’t find the articles. I could find the authors (sometimes) but not the articles. The most interesting citation (did you see it?) is from JP:
(Source: “Fan Service and the Dangers of Objectification in Anime and Manga” by J. Thomas Rimer, Japan Powered, March 6, 2020)
Now, I don’t have an article of that name (nor could I find one by J. Thomas Rimer), but I do have articles with elements of the ChatGPT-generated title inside them: Objectification of Women in Anime, Undressing Anime Fan Service, and Can Anime Show Sexual Assault Ethiically? . Those articles also cover the points ChatGPT writes. It did a decent job summarizing my points. As for the author the model claims wrote the article: J. Thomas Rimer is a professor of Japanese literature at the University of Pittsburgh. He doesn’t write about anime. So it seems, at least with the version of ChatGPT I tested, the model makes up its citations instead of actually citing. It seems to use several sources for each point if JP’s citation provides any guide and then combines that into a single citation. I’m uncertain why it chose J. Thomas Rimer as the author.
Based on my experience as an American librarian, most people don’t care about the source of their information other than finding a source that aligns with their preconceived ideas. This habit is called confirmation bias. Language models could challenge this siloing behavior, but, honestly, I’m not optimistic about this. Information siloing has proven profitable and search engines already silo your search results based on your previous behavior. Also, the backfire effect comes into play when anything challenges these information silos–people cling tighter to their views when information challenges those views. I’m guilty of doing this too. These cognitive habits are part of the way the human mind works, and its difficult to counter them even with practice.
As these models improve, it will become harder to discern articles written by people, written by the models, and written by models and revised by people. In fact, ChatGPT may be helpful for writers on strict schedule by providing bones the writer can build upon. However, the citation problem remains.ChatGPT may also bring ideas the writer hasn’t considered. It could be a useful way to think and explore a topic you don’t know much about–with caution in mind–or look at a familiar topic from a different angle. These models, because they don’t understand language, can create interesting, even off-the-wall, word mashups. These mashups can be creative since the models don’t have understanding as a barrier. These models can be a useful tool for writers wanting to break their usual ways of thinking or to speed up their writing pipeline. I suspect these models will become both content generators and writing tools. There will always be someone who wants to take an easy path. For other writers, these models will become an assistant.
Of course, we will see people attempt to create manga using drawing models like MidJourney AI and these writing models.
As models become more common, we may all drown in the content that these models produce, not that we aren’t already drowning in content. RSS and curating your information flow to just a few websites you enjoy will be the path to sanity. It already is!
I’ll leave you with another article written by ChatGPT:
Me:
Discuss the differences between the anime “Cowboy Bebop” and the anime “Samurai Champloo”.
ChatGPT:
“Cowboy Bebop” and “Samurai Champloo” are two popular anime series that share a few similarities but are different in many ways. Here are some of the main differences between the two:
- Setting and Genre: “Cowboy Bebop” takes place in a futuristic space Western setting, whereas “Samurai Champloo” is set in an alternate version of Japan during the Edo period. While both series contain elements of action and adventure, “Cowboy Bebop” leans more towards sci-fi and space operas, while “Samurai Champloo” is more of a samurai epic with hip-hop influences.
- Protagonists: “Cowboy Bebop” follows a group of bounty hunters, Spike Spiegel, Jet Black, Faye Valentine, and Edward, as they travel through space and take on dangerous missions. On the other hand, “Samurai Champloo” focuses on a trio of characters: the skilled samurai Mugen, the reserved ronin Jin, and the young girl Fuu, who are on a journey to find the “samurai who smells of sunflowers.”
- Art and Animation Style: Both anime have unique art styles that differentiate them from each other. “Cowboy Bebop” has a more realistic and detailed art style, with an emphasis on smooth and fluid animation during action scenes. Meanwhile, “Samurai Champloo” has a more stylized and colorful art style, with a mix of traditional Japanese and modern hip-hop influences.
- Soundtrack: The music in both series is a significant aspect of their appeal, but they have different styles. “Cowboy Bebop” features a jazz and blues-inspired soundtrack composed by Yoko Kanno, with each episode featuring a different music genre. In contrast, “Samurai Champloo” has a more hip-hop and rap-inspired soundtrack composed by Nujabes and Fat Jon, which blends traditional Japanese sounds with modern beats.
Overall, while both anime share some commonalities, such as the episodic format, action, and exploration of complex characters, they have distinct differences in their setting, style, and tone, making them unique and beloved by fans for different reasons.
References
Jalan, Ayush (2023) 6 Ways AI-Powered Search Engines Will Change the Internet Forever. Make Use Of. https://www.makeuseof.com/ways-ai-search-engines-change-internet/
Articles generated used ChatGPT Feb 13 version: 2023.
This is a bit of a tangent, but I did watch a YouTube video where a video-essayist used an AI to generate the script for one of her videos. I have watched and enjoyed many of her previous videos, so it was a bizarre experience, to say the least.
The most bizarre part of it was how the AI would insert many of her mannerisms and turns of phrases into the script. As you can probably guess, it would include them without developing proper context or in really random, weird spots.
I wouldn’t worry too much about AIs replacing internet content creators such as yourself.
Here’s the link to her video if you are interested:
https://youtu.be/8ZSMJNee2p4
Here’s the link to her follow-up discussion of the AI-Scripted one:
https://youtu.be/XKrfCgWM3Tk
Interesting videos. I wonder exactly what inputs she fed the models. Good output requires precise feeding of keywords and phrases, based on my own experiments with the models. You have to provide the context; otherwise, the models go off in amusing tangents and circles.
“The tyranny of an object, he thought. It doesn’t know I exist.”
— Philip K. Dick, Do Androids Dream of Electric Sheep?
Aside from some occasional glaring technical errors, ChatGPT seems to do a fair job at structuring academic papers, where the technical structure is very formulaic. But there’s a sort of Chinese Room effect that becomes apparent, especially with regard to drawing conclusions. Affective responses tend to be airless, or sometimes just bizarre. The limit to an algorithm is that it doesn’t know *why* it works.
Sports writing also seems to be the forte for these models from what I’ve read. Again, these articles are formulaic too. It makes me wonder how long before templated best-selling books fall under these models. I’ve seen experiments, but they aren’t ready yet.