Empathy (draft)
tl;dr: if you're strapped for time, please just go read this comic by Cat and Girl. It's very short, and I promise it's worth these few minutes. Especially if you're not a comic person – the point is that it's a perspective from outside our tech bubble.
I was at a very impressionable age when Gemini became popular in my circles. A version of the Web so heavily focused on the actual content! Very lightweight1, and lacking all the "bullshit" I came to associate with the web – think the website obesity crisis2, intrusive ads, etc. At the time I was the sort of guy that used NoScript and Tor3 for my daily browsing.
Critics have later said there is no website obesity crisis.
The reasoning being that the more "normal" people use Tor, the more you will blend in if you do have important reasons to use it. Using a slow network also obviously aligned with my other views, in a way I was practicing what I preach.
One particular pet peeve of mine were hero images, the big images on top of many articles. They usually don't add anything to the article. For tech topics, at best they're only vaguely related to the matter at hand, and at worst they're a generic stock image. Despite that, they take up a lot of space on the page, and I have to scroll down past them to get to anything worthwhile. They also take up the bulk of the bandwidth4.
An image is worth a thousand words, but that's just a few kilobytes.
But still, despite them being useless, it was just a pet peeve of mine. They weren't actually harmful in any way, and they helped add a bit of personality to a site.
Exciting tech
In 2021, people started sharing some early image generation models. I was absolutely enthralled, and I spent two entire days just messing around with them. I was exchanging AI art with friends, and keeping up with the cutting edge developments. As someone who at the time was still drawing as a hobby, I was coming up with a lot of ways artists could use this to create amazing art in the future.
…four years later, I realize I was really fucking naive.
The attitude in artist circles is overwhelmingly negative. Generative AI is known to directly plagiarize training data, without attribution. It endangers many people's jobs5, further worsening the situation of the stereotypical starving artist.
Note that I don't think taking jobs is a good argument against technology. Ideally, the machines would automate all menial work, and then we wouldn't have to prove our worth to be able to eat and have a roof over our heads. Alas, we don't live under Fully Automated Luxury Gay Space Communism, and it isn't coming any time soon – so I do have solidarity with the people our industry has fucked over. Especially since the benefit for the rest of us is dubious.
Oh, and also, nobody consented to being in the training data. At first I didn't care – the output didn't resemble any work of any artists that I knew6, so I didn't really see any harm. Well, now half the techbros have avatars in the style of Studio Ghibli. The hard work of many great artists to create meaningful, moving art, reduced to a meaningless commodity. A novelty avatar for people who couldn't care less about the original art or artist7.
Note the fallacy. I need to know the original work to be able to spot plagiarism, and I only know a negligible fraction of the total human artistic output. The article about Midjourney I've linked before focuses on AI outputting frames from famous movies, or other well-known copyrighted media. Some people have taken offense to this – everyone knows Disney, who cares. But how would you know if AI has ripped off an artist whose works you don't know?
Not to mention being used for alt-right propaganda.
Think about how it must feel to be one of those artists right now.
Cat and Girl made a really good short comic showing how it feels to be on the receiving end of this. As I've said before, please read it. Preferably now – it's pretty short. It's a perspective many of us in tech aren't that familiar with.
Stock images
So…
Remember when I complained about people adding meaningless stock images to their articles? Not adding anything of value to the article, but also not really harming anything too much?
Well, now people started obviously AI-generated images. The good part of this is that now it's at least vaguely related to the article. Haha, it's an elephant, and this article is about PostgreSQL! Haha, it's a troll holding a bunch of documents, and the article is about a troll with some data!
The actual value added to the article is still just about zero. However, now it's also an implicit approval of the genAI industry and AI art. A casual disregard for the rights and well-being of artists. A statement that you don't care. In the name of what? An image that adds nothing to your words, and that I have to scroll past to get to your actual thoughts?
This is honestly so absurd to me. Why would you ever make this trade-off?
Blame
I don't actually blame the bloggers. I assume the issue might be lack of awareness?
AI is a very polarizing issue nowadays. IMO, current AI discourse sucks at conveying the actual flaws. It's easy to get only a small part of the whole picture, especially if you only hang out in tech bubbles.
And look, I used to be into this stuff too. I get it. I used to think the anti-AI crowd was misguided8, seeing harm that isn't there. In a way I do still think it's exciting technology. I did eventually realize, however, that the questionable benefits aren't at all worth the price. Do you really need that dumb image?
Seeing people I respect use AI art is thus really fucking depressing. Especially if it serves such a banal purpose.
I feel like I'm beating a dead horse – but this is still too common.
Anecdote: when 404media published their article on Mistral releasing an uncensored open-weight model, I thought it was dumb, so I started seeding the torrent myself out of spite. Now… I still think it's dumb, and a good example of how bad the mainstream AI discourse is. That's a topic for another day, though.