AI & the small Web

Webweaving
1

This post was catalysed by me coming across Webring Studio, a new webring hosting site that appears to be mostly or entirely written with an LLM, a Large Language Model, the type of technology used for Claude, ChatGPT and the like. AI-generated imagery is liberally used and the text has some easy tells in it, such as that whole “it’s not x, it’s y” business - I counted eight on the ‘about’ page alone. They’re doing what they’re doing, it’s their site and their business, but it did get me thinking.

The small Web (is that what we’re calling it? I guess that’s what I’ll call it) is an explicitly handmade project so the presence of AI-generated stuff on it is worthy of discussion. We’re hobbyists here, enjoying the process of making like you’d enjoy making a bowl at a ceramics course. Letting AI write your site for you is like bringing a bowl from Ikea to the last day of that ceramics course - yes you’ve got a bowl like everyone else, but we’re not learning how to have a bowl, we’re learning how to make one - that’s the whole end goal. I don’t think it’s gatekeeping to not call that person a ceramicist.

The site mentioned above is an extreme example, of course. Plenty of people having bits of code generated for them, have their avatars generated, have some example layouts made, that sort of thing. Enough’s been written about the impacts on environment, labour, plagiarism and the mental states of users and I do not believe we can leave those considerations out when trying to answer the question: how do we feel about this stuff in the small Web project? Billions have been spent to work this stuff into our daily lives, so now what? Yes, no, maybe? No more than 10% as the oddly named Not By AI proposes? I am curious how everyone feels about this stuff.

I’ll put in my two cents to get us started: I’m definitely not a neutral party and am unable to pretend to be even for a second: I hate this AI stuff. I feel a little sick whenever I find out I’m reading machine-generated text; AI images make my skin crawl; I administer a whole webring on the topic; I’ve been spending time at my job at a cultural institution trying to get the use of all AI banned and have refused to put my name under anything that’s been ‘run through ChatGPT’. People have accepted the presence of this malicious, plagiaristic, environmentally disastrous technology far too easily and I’ve been making a real nuisance of myself in response. I’m feeling some solace as at least the corporations behind it are now starting to circle the drain, like their crypto and NFT forebears. Hopefully we won’t ever have to talk about AI again a year from now.

Still, I’m a big believer in finding out what you hate to help define what you love; the proliferation of AI has certainly played some role in making me realise how much I love the small humanmade Web. I get to look at new websites every day and some of them are made by people taking their first steps into webmasterdom - those websites are beautiful, I love surfing them and watching webmasters come up with new stuff. Messy and beautiful, or simple and barebones, or anything in between or something completely different and unexpected. The beauty of building a website is that it takes so little to get going and anyone can go from knowing nothing about building a website to getting something going on Neocities within the hour. A text editor is all it takes, that’s the beauty of it.

15 Likes
2

the chatgpt voice is so so obvious when youve read a few of the texts it generates its crazy these kind of people think we wont notice

2 Likes
3

Of all the things that could be targeted by this kind of scam, webrings and the indie web in general is such a weird choice. Offering a supposed shiny new solution to a problem that was solved decades ago and pitching it to a group of people who specifically want to go back to old ways of doing things makes no sense. The reviews/testimonials section is so weird it’s kind of creepy. The header text in the first box of “reviews” basically admits that they’re fake.

6 Likes
4

hi! i am the one who flagged it. since you seem confused, here are the community guidelines i think this is violating:

  • We do not tolerate trolling, elitism, gatekeeping, or disrespect toward anyone.
  • Give everyone the benefit of the doubt first.
  • Our community is not a social media feed.
  • Unsolicited feedback will be removed.

posting a link to somebody’s project for the purposes of complaining about how you don’t like the vibes and find it annoying seems pretty obviously inappropriate for the cafe to me!

i don’t think that the “links around the web” subsection should be used for linking to stuff you dislike or for inviting other forum members to post in a thread complaining about that website
i think that posting this type of website critique about any other website would be obviously inappropriate, and i do not think that the use of AI or LLMs carves out an exception where this is no longer rude as fuck. the rules do not conclude with “…but if you suspect someone has used chatGPT or has bad vibes, the rules no longer apply and you can do whatever you want”.

it’s not twitter, posting links to stuff for the sake of going “get a load of this guy!!” is not currently the norm and i would not like to see it become the norm. that’s my two cents!

7 Likes
5

I’m not quite sure what to think about this project, to be honest. From what I can tell by reading through it, it seems like a well-intentioned thing made by some one who isn’t really part of the indie-web. The webmaster has been making websites commercially for a while, and has their own blog that uses AI in various ways.

I obviously don’t like the use of AI at all, but I also don’t like being gatekeep-y. And while I don’t like the obvious ties to monetization, I do like the idea of creating more webrings, and directories of webrings. Hopefully this is just an attempt at creating something genuine by someone who has embraced AI for filling the gaps in their skillset, but I guess we’ll only know after time has passed.

6

I’d actually like to push back on this a little bit, if that’s alright;

I do agree that this post fits generally into the “damn guys this dude SUCKS look at that!!” genre that we try to avoid here, but also I think the question of “how do we feel about companies marketing to the indie web” is a very valid one to talk about!

Of the four community guidelines you cited, I’m unsure if it fits into the “community is not a social media feed” bulletin; rmf posted this to start a discussion, with a long post. They could have been nicer, sure, but at the same time I’m not sure an LLM-created website requires the same sort of kindness and benefit-of-the-doubt that a human-made site does. That’s certainly up to discussion, though! Perhaps in another thread.

“Links around the web” isn’t for talking about things we love only. We have a handful of posts that are primarily focused on discussing issues that are relevant to our corner of the web, like this one about the future of online creators, this one linking to a page that just makes fun of every episode of Full House, this one that laments the state of the job market and work and such… the list goes on. A lot of threads end up as people discussing how much they dislike something. That’s okay!

Thank you for flagging the post, I do agree that the tone could be less gatekeep-y, and you’re right that this toes the border of what we want our culture to be here. But, also, it’s not enough of a violation for the post to warrant removal or anything. I think some great discussions could be done with that linked website as a jumping-off point.

6 Likes
7

okie doke! you are the mod and not me. perhaps the guideline language should be adjusted, since “we do not tolerate […] gatekeeping” and “it’s fine to be a little gatekeepy sometimes, depending on the context” are very different sentiments!

2 Likes
8

i agree with a lot of what you’re saying here. i don’t want the Links Around the Web category to be used for sending people to a site to essentially hate on it or go after. that feels inappropriate. but i do see the value in what @Jay is bringing up, the greater discussion around AI in indie web spaces.

@rmf - would it be OK if we adjusted the topic so that it’s less of a “look at this ONE SITE” and instead talks about the greater issue you’re bringing up: does AI-assisted code have a place in the indie web?

4 Likes
9

This is a good idea, I think! Definitely agree that maybe just rewording to the broader discussion could help :thinking2:

3 Likes
10

Withholding specific comments for now but putting forward an endorsement of this question as a discussion topic.

2 Likes
11

Fair points. I’m a big believer in being a hater but I wouldn’t want to set a precedent to turn this subforum into a hatefest. I will rewrite the OP more thoroughly. I think the site in question is a useful example, if that’s alright, but I’ll make clear that it’s just that, an example. I’ll get into it right now, please have a look in a bit and tell me if it’s appropriate.

–

Alright, hope this does the job.

3 Likes
12

I often find it easy to tell when text has been generated by ChatGPT, but I didn’t get this from that site. The writing feels more like typical corporate blurbs, piling on the rhetorical questions and the “it’s not x, it’s y” stuff.

Do we know how much AI this site uses or who’s behind it? I feel like I’m lacking some context here.

In terms of the site itself, I’m more weirded out by the overall tone and its unclear purpose. How exactly does it differ from existing ways of making webrings or from directories? It seems like it’s trying to come across as an exciting new product for the indie web revival but doesn’t really show us what’s exciting. I also rolled my eyes at the “the web has become boring” claims because surely anyone already interested in the indie web is already looking at more personal/niche stuff instead of the same few major sites?

To answer this thread’s question… my one concern with AI use are the environmental issues. I’m not likely to stare at someone’s avatar long enough to tell if it’s AI-generated and it’s unlikely to be stealing someone’s job.

Ultimately, I really don’t think AI in its current state can properly replace humans, much less in these parts of the web. A lot of what has drawn me here is the creativity and self-expression and that’s something AI can’t do for you. If I come across a blog that reads like AI-generated slop, I’m not going to read it or engage with it because it’s just not interesting to me.

In terms of the small web, I would say that the mechanisms in place are fairly different from those of big social media sites. On Facebook, you can just mindlessly share a post; on the small web, you often need to put a bit more effort into drawing people’s attention to things you find interesting. There are few to no algorithms shoving content down your throat, so you can have a much more curated, personal experience.

That means that we don’t have to engage with mindless AI use on our own little corners of the web, we can simply ignore it and let spammers figure out that engagement works differently.

Personally, I don’t use AI for any of my content. If I didn’t want to put effort into something like writing a blog or coding a site I would just not do it. Why do something if you don’t feel like it’s worth time and effort, y’know?

P.S.: I could write a whole post about how a lot of anti-AI discourse really overhypes AI and sounds like free advertising, but I won’t. The bottom line is this: if you’re doing something to express yourself and be creative, any AI is going to have a hard time doing it the way YOU would do it.

3 Likes
13

in my experience chatgpt takes every annoying quirk of advertising posts and turns it up to 11 and thats definitely how this feels. also the em dashes in every single paragraph

1 Like
14

The em dashes are not a good indicator that something was written by AI. Em dashes were created by humans, are used by humans, and have been a thing for a long time!

3 Likes
15

ChatGPT was a game changer for making coding feel more accessible to me, personally. Most documentation feels incredibly opaque and is generally set up in a “wiki” format that makes it difficult to figure out where to start or if there is a starting guide it makes it difficult to figure out where to go from the absolute basics.

“But just ask for help”, you might say. Fair point. I love asking people for help. The problem is that I seem to run into this problem where a large portion of the coding community doesn’t like answering the question you ask. I really like javascript as a concept, and wanted to learn it but I didn’t know where to start. So, I asked and one person gave me a guide, but what everyone who responded said was that I was better off learning a different language. Which might be fair if I wasn’t explicitly a hobbyist who specifically wanted to learn javascript.

That’s what made ChatGPT appealing. If I asked it a question it would answer the question. I didn’t have to wade through forums of people arguing over why x, y, or z is actually better than a, b, or c. It’s also really good at laying things out in plain English and quickly clarifying anything specific.

You might notice, however, the past tense. Everything I laid out is why I think that AI has a really powerful use in being able to parse opaque documentation, if it could be done more sustainably and ethically. But, other than a question of sustainability and ethics there’s one glaring problem with ChatGPT: it is really fucking stupid.

At this point, I’ve almost completely kicked the habit of using it for anything more than giving definitions that Google’s gotten too trash to find, because it is just… So… So… Stupid. I have had to beg it to pretty please only reference the current version of a codebase and repeatedly remind it of what we were talking about just one message previously. Maybe the paid tier has a memory longer than ten bytes, but I’m not about to give them money to find out.

That’s not including the fact that it will literally just make things up or lie just so it can avoid having to tell you something isn’t possible. It will run you in circles telling you that “you’ve” done something wrong when all you’ve done is taken the code it just gave you and found that it doesn’t work. At that point, I’d rather deal with a hundred people telling me I’m stupid for not using this program instead of that program than have to waste my time separating the wheat from the chafe by talking to ChatGPT.

There’s always going to be a little shame associated with it, and that was a motivator to more properly learn to code. None of the projects I made with ChatGPT will ever see the light of day, because I can’t in good conscious say I actually made them. But, sifting through ChatGPT’s bad code did give me a better idea of how it works and more confidence in being able to parse resources I find on the web. So, there’s that, at least.

8 Likes
16

the majority of people dont put them in every paragraph tho. and its obviously not the only tell its also combined with the ultra promotional language and the constant repetition of specific sentence structures… chatgpt didnt even use to put em dashes so often its just gotten worse with it over time

edit: and im saying this as someone who uses them a LOT in my creative writing like there ARE times where i use them in every paragraph but ive very rarely come across other people who the same without any of the other chatgpt tells

1 Like
17

@beetlegooze I can see what you mean about that particularly use case. I think some communities (like this one) are really nice to beginners but some are just condescending and hostile. I don’t join them but it’s disheartening when you see a post from someone asking what you want to know and the replies are rude.

This is one of the things that could be good about AI, but like you said, there’s unfortunately too many environmental issues, plus how unethical these companies can be.

You’re spot on about LLMs making up answers. People forget that their main purpose is just replicating human language, not necessarily checking information for accuracy.

@juette you’re probably right about the style, I guess AI slop and corporate blabber sound the same to me in how hollow and generic they are aha.

2 Likes
18

ai slop was inspired by corporate slop so thats not surprising :pensive_face: i actually found the personal website of the person who created webring studio and while she does have a few very corporate-sounding posts (apparently she used to work in SEO so that makes sense) most of them are written normally thankfully (though none use as many em dashes as webring studio’s website does. its a bit of a shame to use ai to generate the entire text of a website for an utility that aims to uplift small human-created websites but oh well)

2 Likes
19

my big annoyance with chatgpt (aside from environmental concerns) is that it tries to be an all-purpose tool and ends up being mid at a lot of things. like making smaller ai models that specialise in specific topics would be infinitely more useful than having a single one that tries to be an expert in everything like what theyre aiming for with chatgpt. i really wish theyd program it to admit when it doesnt know something instead of constantly lying or making shit up tho idk if thats even possible lol

5 Likes
20

While I appreciate you sharing your experience on the positives and negatives, I’d like to question this particular framing. Note I want to preface this by some acknowledgements: I’m glad you’ve moved on from using ChatGPT this way, and I’m also sorry to hear about your bad experiences with asking for help from real people. It’s shameful that anyone would botch the process of helping others so badly as to discourage them from asking questions at all.

On the subject of what chatbots can or can’t do for hobbyist coders, I’m skeptical that chatbots can parse opaque documentation reliably, for two reasons:

  1. Given the nature of chatbot programming and the nature of composition (where good composition = easy-to-read-ness), the kind of text that a chatbot can summarize the most reliably is the kind of text that’s already pretty easy to parse the first place – because these things don’t have perception, just programming. All the programming can reasonably do is reformulate text based on certain standard compositional rules, which it can do best for text that bothers to follow those rules. On this basis I think that the bot-generated summaries will be the most reliably accurate for the cases where they’re least needed. For cases where a helpful summary is most needed, chatbots have a higher chance to mislead you by generating a “summary” that’s off the mark. And on that note…

  2. If you don’t already know the ins and outs of what you’re looking at, a bot-generated explanation or piece of code could contain problems that you aren’t prepared to recognize. This includes security hazards arising from slopsquatting. For more on this whole set of issues, see Baldur Bjarnason’s Why I Think Using Language Models for Programming is a Bad Idea.

The reason I wouldn’t count on that is because chatbots don’t “know” anything. They’re text generators. They generate text. Ask one to give a botanical description of Fakeus planticus and they’ll do just that.

4 Likes