What are your thoughts on using AI for web design?

[schlime9k]

What I argue is: The more complex your tool/media is, the less is your ability to take influence, while the influx of the medium increases

If you're saying "the more complex the medium is, the less creative control you have" then I disagree hard. Are the Beatles less creative than Beethoven because they had access to Multi-track recording, microphones, synths, mixing consoles and audio editing technologies? No, in fact, all those expanded the range of ways they can be creative because they had more control over their artistry.

The same can't be said for generative AI because it's more simple. I think the real question is not "does complexity reduce creative freedom?" but instead "How much of a tool is being delegated creative freedom, and how much is too much when it comes to art?". Now, you are saying that, and I agree that is a topic that needs to be discussed further in order to come to a proper answer. But I think you're just too hung up on the "complex tool" idea.

This is the last I'll comment here, I think we've both said our pieces here nicely, so I'll leave it here. Also I've searched up the digital art stuff, it only came up with post-AI topics so I'll have to refine my search to like 2022-2021. Thanks.

[ThunderPerfectWitchcraft]

If you're saying "the more complex the medium is, the less creative control you have" then I disagree hard. Are the Beatles less creative than Beethoven because they had access to Multi-track recording, microphones, synths, mixing consoles and audio editing technologies? No, in fact, all those expanded the range of ways they can be creative because they had more control over their artistry.

I'd argue that Beethoven had more creative freedom - especially if you consider that he only used paper and pen to compose. In fact, he had so much freedom that some of his compositions weren't able to exist in the real world for quite some time. The medium is so free that it would have - theoretically - allowed him to compose like Avo Pärt as well as Penny Lane or Within You Without You; however, since he missed the instruments and tools, there is a low chance that he would have been directed to doing something like it.

The Beatles were highly dependent of their technique, and their individual skills - many songs were written with one of their members as singer in mind, to give an example. They used what they had masterfully, but it locked them out of even sounding like a band with a slightly different setup (lets say: The Velvet Underground) - the sound of a rock band is formed a lot by their technical equipment and their production, and many things can't be done with it, while a certain sound is rather easy to produce with it.

We can stop here - however, I want to say that this approach to see media works out in the end, while I believe that most others will entangle in discrepancies.

[Lyonid]

I have tried using AI on my work, and I found it dull and unappealing – especially for anything more expressive. People are using it though, and they may not be as informed about it as some may think. I am happy to see people knowing about the catastrophe that is any LLM, and taking action, no points. If someone attacks people over non-malicious AI use, then they are making the fight more difficult for anyone, I believe. That is all.

Really fun theory that I have been thinking about in this context tho!! David Pye in 1968 was thinking about how we attribute value to pieces of creative media based on the medium. He described "workmanship of risk" as the art people would be most impressed by, basically, the more things can go wrong, the more impressive it is that they made it work. A dull photograph becomes impressive when an artist has drawn it realistically with a pencil, basically. Or, you manage to translate whatever was on your mind to an oil painting, and people thought "damn that is neat." The opposite, the "workmanship of certainty" to him was stuff like any printing technique. It was pretty easy to be certain about the print turning out the way it did. Both ways of creating are legitimate as they all have purpose for various reasons, obviously, but now this is being put to the test very heavily. A printer is obviously not on the same level as the most demanding technology in the world (even though it's probably more useful and quite a bit less predatory ). If you give the chatbot zero to no guidelines, you also are probably very much uncertain about the outcome, rendering the artistic process very questionable. So taking it from an artistic perspective, I think that employing AI for your web design, in my eyes, just leaves you as someone who wanted to create something that you know will probably work out just fine. You didn't fail at any stage, and that is all there is to say about it! If people ask you about your process, you will probably not be able to say more than one sentence about it. If you're fine with people thinking about your work that way, no one can really stop you from using it. I personally click away an obvious AI-generated website very quickly.

[Computatrum]

I'm a bit reluctant to write these because of the issue that I'll mention in the third paragraph, but hear me out.

I can understand the (mostly justified) hate for AI. Most popular tools are held by corporations that have the concern of so-called "shareholders" and most of their CEOs or chairmen is bunch of weirdos that constantly having dreams about world dominatioooon  .

However, AI tools' itself can commit to a greater good to yourself, if you use it accordingly (and responsibly). Based on a saying that goes by "Don't catch fish for me, teach me how to catch it", or something like that, I think that AI can be really useful when it comes to stuff like teaching, or redirecting to other resources that you cannot regularly stumble upon, or not indexed in a search engine. Regular folks would prompt them as "Hey, make me an X something." and the AI would give that "X something" right away. However, instead of wanting that directly you could ask like "Hey, how can I do this X something, do you have a proper resource for it?" and it would share the resources, methodology  etc. Unlike receiving an artificial, soulless product, you could make one with a soul, and that skill that you've learned by making the product would remain inside yourself forever.

Yet, there are some issues that we should address before using an LLM. The morality. We hear stuff about AI companies like they're using rare books for resource and burn them afterwards, data centers and the issues they bring, the whole AI """art""" situation, easiness of creating something (slop, misinformation, you name it) with AI and spread across Internet and this whole thing contributes AI corporations' expansion over the tech industry. They're taking advantage of it while our society gets forced to deal with these.

That's why, an open-source LLM would solve the corporate part of these issues and could be more humane than corporations' proprietary LLMs. I don't know any projects like this but I would favor them if someone decides to do it   .

[kieraoona]

As someone who's been on the web since yee olde days of dial up, I think AI is inherently terrible in situations if the person using it doesn't know how to read code.

Especially code that could be hallucinated by the AI into the work, or even worse, some sort of hijacking code that the language model may eventually be either forced to, or forces onto those who are first learning it, or trying to get a website up as fast as possible to start a business, causing key other infrastructure to crash, or to start scraping other peoples' data without their consent, and the "programmer" has no idea the site is doing it.

We also seem to loose something whenever a computer does 100% of the thinking for us, and its already been proven that the continual reliance and use of AI causes cognitive decline, and a reduction in critical thinking skills.

The only time AI should be used is in a diagnostic tool, and only with human oversight, with the knowledge to know if something has been made malicious or hallucinated.

[Lyonid]

The only time AI should be used is in a diagnostic tool, and only with human oversight, with the knowledge to know if something has been made malicious or hallucinated.

I kind of agree, but LLMs aren't really diagnostic tools as opposed to simple regression models and image recognition software. The human oversight is important, and the more traditional tools are incredibly clear about the high risks of false positive and false negatives, which an LLM usually is not, as it is designed to be frictionless. I remember people being all clever about how useful AI is for identifying breast cancer, but there is no LLM doing this reliably as it can only employed for reading textual clinical data.

That's why, an open-source LLM would solve the corporate part of these issues and could be more humane than corporations' proprietary LLMs. I don't know any projects like this but I would favor them if someone decides to do it   .

Love this. I have always been wishing for more people to train their own models on their own data to generate things with real intent. Recently, I saw a talk by Hito Steyerl on Friction, and how any system requires it so we can stay vigilant and resistant to it. They trained their own model on open-source material, and it obviously was utterly useless (which is good; and good to know). After all, it's in the interest of corpos for us to never get a hold of sufficient open source data to create a truly open-source LLM. It is a tedious uphill battle, which requires an unsurmountable amount of labor. I don't think that's what we need, though. I feel like what we really need is finding ways to value the data we already have and give it meaning for our communities. For us here, any weird and obscure gif that someone created 20 years ago is incredibly valuable, and we attribute it far more value than any LLM output could. I'd love to see this translated to other communities.

There are SO many cool models for generating anything, which didn't require data theft in the first place. These things are hidden, however, and we need to make an effort in finding, curating and making them accessible. I feel like a messy generated bouqet of flowers is far more useful on a personal website when it was created using some shitty ancient model as opposed to one generated in an LLM.

[Kallistero]

if you have time, check out my site https://meridianos.zip. .  It's not just a site but a site builder. completely free. runs locally on your machine.  let me know your thoughts and how it compares to some of the sites that used AI

Other sites that used AI often don't have the vision that you display, so I can see that while you have the idea, you're just unsure of your skill. Albeit, this vision of a website that looks like a retro OS isn't the most uncommon theme among people on the independent Web:

• Webamp - Winamp 2 in your browser
• s0vie's n0de
• Sidows 95
• Swiss Portrait (musician)

I see that you can at least appreciate a Web aesthetic that's outside of the mainstream schools of Web design.

I would never have been able to create the website I created without it, and thats my dilemma.

I've been tinkering with no-code generators and meta-coding widgets since I made the Neocities page on my profile some years back. I haven't touched it in years. I'm better with coding now than I was back when I was actively working on it, but your basic features are already features of my old site there. It has element draggability with embedded content, albeit I clamped it so it can't be dragged outside the window and I put it on a postmessage listener so it opens & closes in a complicated way that lets me take the draggable window to websites that are outside of my page. It has GUI-based visual effects customization on my effects maker prototype page. It's just plain HTML/CSS/JS, so it just runs out of the box without the build & deployment stages that come with a Web framework or filesystem editor.

I hand-coded every bit of it, from not having done much with HTML beforehand, with my primary inspiration resource being the Neocities website browser and my primary learning/coding resources being MDN Web Docs and W3Schools. I began without the skills to make what I made. My answer was to learn how to make it. Your answer was to get something else to do it.

That isn't an indictment of you. I would dare to say that everyone here who has their own Web page has copy-pasted code at some point, and that was before they went on to make it their own. Nearly every artist has a "tracing" phase. Every musician starts by covering songs that other people made. You are not wrong JUST for using training wheels. Just, they are training wheels.

I've been made to learn about AI, and the way AI works is that it's a massive interpolator; it regresses to the mean. That's to say, it can only reliably operate in spots that are in between pieces of data that already exist. If the thing you want doesn't exist, it'll have a hard time making it. If you wanted to vibecode my website without (implicitly or otherwise) just having it reference the code from my website, it would take a pretty long series of very specific instructions just to get the sparks from the bottom of the page working the way that they do. However, if you want a draggable window, the AI can easily see that there's a "react-draggable" dependency from NPM or even grab the code from within react-draggable to apply it to your page. The pages within the frames, the rotating globe with meridians on canvas, the menu behaviors, they're all things that an AI can get straight out of some box or other. You've put together templates. Which tells me that maybe you're ready to start digging into those component templates yourself, modifying things, experimenting! 

But let me try explain you my perspective by asking the question: why do you want to make a website? What are you going to get out of it? If you end up deciding to actively deprive yourself of the processes of artistic creation and learning, what's in it for you?

I can answer these for myself. I want to make a website to showcase my work. I understand that on forums like this, the website itself is the art itself, but for me, I don't want to be spending so much time editing and building a website. I still want it to be artistic but I would rather spend my time creating other things. With that being said, has that deprived myself of artistic creation/expression? Not at all. It's still my ideas being created.

Now, with all due respect, when you first open your meridianos.zip website, you're hit with a page encouraging the user to use your AI-generated site templater. If you wanted to make a website to showcase your work, you don't want the website itself to BE the work, and I see that the repo is a few months old, where is the work? The other things you've shown are other frontends using this template generator, and every window is one of the sub-templates from the generator.

Knowing what I'm looking at in the repo, and this may be an honest mistake resulting from vibecode: I would not run this on my computer.   While there isn't obvious exfiltration, the updater.html file gets privileged filesystem access that I do not grant to a locally running Javascript application that is actively & automatically changing the files on my system, unless it's a purpose-built WebDAV that is sandboxed into the folder that it's in. This itself is a learning moment: a vibecoded application may include things that have potentially unsafe or excessive/unnecessary behaviors that you probably don't want to encourage others to risk. Plus, ALL of your posts on the forum, from your months-long stay here, are about getting other people to use this. It comes on kind of strong 

And beyond this: it's far too early to be putting up a donations page whenever this entirely new vibecoded application is the only thing on your GitHub. It's okay to be hopeful, but there's a culture around vibecode that makes a person's intentions look especially dicey whenever they do this. Again, it isn't just to indict you, there's something to be said about the company you keep & how the subcultures whose practices you adopt will affect your priorities. It's okay to be coming from the types of people who would be thinking about monetization before most other things, but it's the exact type of person who SHOULD be reached with our native ideas that while it might feel like the default way people engage with sharing creations online ("influencing"), it's not so early on the list for creatives who spend time away from that type of culture    And if the AI put you up to this by suggesting "oh, and this spot can be used for your donation link," that itself gives a message that the use of AI for design can have some cultural baggage beyond just your own ideas being created.

So, putting it all back to the question at hand, using AI for Web design is:

• Tracing. You're tracing someone else's work. You can trace things. Just, until you make it your own, it isn't yours (not even legally  ).
• A way to skip the design step just so you can get the thing you want to feature on the website out there. Again, you said you want the website to showcase your work, as opposed to a situation where "the website itself is the art itself." That's how businesses operate & a reason they don't have a need to care much for a human touch; they first want users to have a way to get to the actual product. So, an AI can generate a placeholder design, but it's:
• Limiting in ways that you may not realize until you take off the training wheels. The AI stuck a bunch of templates together to output this template generator. Even the concept of an old desktop, that itself is a template. You as the human, with the universe & its connections as your dataset, can imagine things that are beyond the template, if you dig into the building blocks and add your inspiration.
• Potentially unsafe. If you're doing it because you don't understand, it will do things that you don't understand! 
• Culturally precarious. If people are given the option between an AI-generated Website and an obviously human-made one that delivers the same, they will choose the human-made one. There are lots of reasons for this, outlined by other replies, but just know that it makes your designs replaceable by anyone else with the AI and your intentions easier to question than if you made it by hand. The Patreon would be easier to match with your stated intent if you had hand-made this (badly), iterated (so it got better), sought feedback, questioned if people might want to use it before opening donations... instead of vibecoding and well, y'know, if you ever wanna give me money for it...

I dare guess that relying on AI as we know it (as in, large-dataset predictive mass interpolators) for design can't really revive the Web. If everyone started using your AI template or just designing as you've done, all the things about it wouldn't be flags for Web Revival anymore; they themselves would be consumed and copied by AI again. Web Revival is a creative movement, and creativity is largely an anti-metric, a deviation away from norms that change when they're added to. The further you get from the norm, the worse an AI is at operating at those reaches because its dataset isn't geared toward that.

Now, while "It's still my ideas being created," you can tell an AI your ideas & hope it sufficiently approximates them, but it will default in the direction of patterns that it's already seen, moving your creation away from its creative potential. Unless you're dictating the features element by element (at which point you'd might as well write them yourself), you'll lose the "Oh, but can I do THIS?" moments that happen mid-stroke when you're working with any creative medium by hand. Great strides of creativity live in those moments, and you don't get them when you're offloading that granularity of the process to AI, because you don't even know what the medium is capable of. Now, you CAN ask it "What if X?" but what I'm saying is that the question itself doesn't even arise because you aren't thinking about the element, only the output.

So, AI for Web Design? Tracing, training wheels, placeholder content, something you might keep for your own use, something people will treat as inferior to the handmade equivalent. AI for Web Revival? It can create an undead internet, but not a full revival