“No AI” Statements Are Much More Than Mere Statements
33 points by FedericoSchonborn
33 points by FedericoSchonborn
There is such a huge available spectrum now between “no AI used at all” and “completely generated by AI” that this sort of analysis is rapidly becoming outmoded for artifacts beyond the individual Reddit message level.
“Handmade” on ceramics doesn’t typically mean the artist invented their own glazes, or used a wood fire kiln with no thermostat. Of course, those are absolutely viable options and may make the work more interesting in a way, but lack of them wouldn’t disqualify it from being “handmade”.
I sometimes compose classical music as a leisure pursuit, and the music I compose is much harder than I can play. When the quartet plays it, I still get credit for composing it, even though I didn’t do any of the “sound generation”.
To me the question is: “does this work embody a human’s point of view, purpose, creative direction, taste?” Not “was this work produced directly by the human claiming credit for it, down to a certain arbitrary level decided on by the viewer?”, or “did a human do the very last step in the production process for this work?”.
And of course that is also a sliding scale, not a binary. You don’t get much credit for prompting a music generator with one sentence and hitting “go”.
I spent this most recent weekend building a thing that I've wanted to make for like 6 years. it's not a wholly unique idea, but it is two existing ideas smushed together with (afaik) a novel idea about how to glue them together nicely. it involved a lot of text parsing and some other things that I could probably do by hand if I committed months to the effort, but that I am currently not any good at. instead I used [insert frontier model here] for most of the actual coding. the end result is something that I think is genuinely very good, cool, unique and even broadly useful that I'm really excited about! I definitely believe that it embodies bits of myself. it took all weekend, dozens of prompts and rounds of finding bugs and tweaking the behavior, I made a little logo myself, I'm in the process of writing documentation for it by hand.
but I also dread the idea of trying to share it anywhere because I don't wanna get caught in the crosshairs of either the "would just call it slop and dismiss it entirely" or the "time to drown a stranger in slop pull requests" crowds. it feels like no one from either side would actually care about any of those details. I'm sure there are more people in the middle than I think, but the people on the extreme's are definitely the loudest, and it leads to a perception that is really is either "no ai" or "ai slop". I want to believe that there's room for a middle where a thing can be made using ai in the process and that it doesn't mean that it's compromised to its core, but when I look around me I don't see much evidence that it's true.
I wouldn't worry about slop PRs. Most projects in the world still attract zero PRs and yours is likely to be the same.
I want to believe that there's room for a middle where a thing can be made using ai in the process and that it doesn't mean that it's compromised to its core, but when I look around me I don't see much evidence that it's true.
I think this hits on where I'm at:
I want to believe that a thing can be made using ai in the process and that it doesn't mean that it's compromised to its core, but when I look around me I don't see much evidence that it's true.
It's a bit like CGI - you'll see lots of CGI that looks bad and is too distracting, you'll see lots of things where the CGI is a big part of the thing existing. But the CGI that fills the right place, is good quality and perfectly blends in is largely ignored and people don't realise it's there.
For the same reason, I don't think you'll find the evidence you're interested in if you just look around to check projects using AI. It's not like I explicitly tell a customer "I ran this project through LLMs that found and fixed 5 issues and fixed some UI inconsistencies" - they just care about my app working correctly. I'm still explicit about mostly-vibed public projects though - even if there's a significant amount of design and testing that went into them, I'd rather people weren't surprised by the fact.
I [built] a thing that I've wanted to make . . . but I also dread the idea of trying to share it anywhere because I don't wanna get caught in the crosshairs
You’re not alone
it took all weekend
I don't want to sound harsh but that is not a lot of time for a side project, no?
for someone with a full time job, and three kids, who values a full night’s sleep, it is an absolute whack ton of time. and also it was a pretty small project.
my point in bringing up the amount of time definitely wasn’t to paint myself as some hero, but it wasn’t trivial either.
well many open source mantainers in similar situations have spent countless weekends on their projects only for their code to be used thanklessly and without permission to train LLMs. getting your weekend side project to be called slop by leveraging these same tools is a small price to pay imo
There absolutely is room in the middle, and I think people will have to recognize that once the dust settles. Just ignore the extremists and enjoy building your stuff.
There are other arguments (ethics, energy, copyright, etc.) that you should pay some attention to, but this particular "one molecule of AI usage makes the whole thing slop" thing is just nonsense.
In the same boat here, I had a whole bunch of half baked projects and ideas I've collected over the years that I simply didn't have the energy to get to the finish line. LLMs allowed me to get a lot of these things to the point where I could see where the idea is even worth pursuing, and satisfy my own curiosity. Some of these things ended up being genuinely useful to me, and I now use them on daily basis.
All the people who keep telling you that you shouldn't use these tools and you gotta make everything artisanally by hand seem to have unlimited time and energy on their hands. Some of us have day jobs, and you can only do so many thing in a day. LLMs allow me to do side projects I enjoy and make things I find interesting without getting burned out.
To me, the question you present is for my readers to decide. The only thing I can do in my writing is to detail my process.
I wrote in my last article:
100% of this article has been handcrafted without using generative AI tooling, if somewhat hastily. Take my typoes and broken sentences as proof of that.
a bit snarkily, but I guess it gets the point across. Then I let my readers decide if the result "embod[ies] a human’s point of view, purpose, creative direction, taste"
I mean, you have a point in principle, but I'm really not seeing that kind of nuance in real life. The vast majority (99%+) of AI content in my social media feed isn't in the category of "someone had ChatGPT suggest a couple of edits and now they're being unfairly ostracized". It's wholesale slop generated for clout / karma / ad views.
In fact, I think that sharp delineation between categories is sort of guaranteed in the long haul. Writing is hard. If you use AI such that it's still your work and still your voice, then the process is still hard and you're not gaining a whole lot - so relatively few authors will bother. In the meantime, vibe-writers ship a thousand pieces of blog-slop and book-slop.
AI art is similar. Yes, you can in theory use AI to enhance your own artistic expression, and there's a small number of people who do. But rest assured, if you come across an AI-generated image, an AI-generated song, or an AI-generated video, it's almost certainly not gonna be the pinnacle of human creativity.
And yes, there are rare occasions when someone gets unfairly accused because they like em-dashes. It happened to me, but I think think the backlash here is misplaced. People go on these witch hunts because everyone, left and right, is trying to sneak 100% AI-generated content into their feeds against their will.
You're going binary again. My point is it's not 0% vs. 100%. And "AI-generated" is no longer a binary term either. Unless you just want to go with the "vegan" interpretation, which is perfectly fine but is more about the ethics of the practice than the result. Some people don't like using horsehair bows on their cellos, but that doesn't change the music.
I hate to tell you, but relatively few authors put in more than minimal effort before AI. I mean, have you been on LinkedIn in the past 15 years?
I guess my bumper sticker version is: Human creativity and taste is a separate dimension to tool use. Judging results by what tools were used is only useful during the period when the two measures are correlated. But that period may be brief, so we'll need to advance our judgement framework.
Indeed, there were times when digital art wasn't really considered art, as a principle. It's not a real painting if you use your tablet with an app that gives you undo and editing capabilities. The same was very much true for music, people couldn't phantom that electronic music could be actually artistic, not just kids making noise on the computer. Much like AI, these tools lowered the barrier to entry, so you have masses making low quality stuff, but that doesn't mean art doesn't exist.
There was a time when photography wasn't considered art either; it was just a soulless mechanical documentation, or a cheat for people who couldn't draw.
“Handmade” on ceramics doesn’t typically mean the artist invented their own glazes, or used a wood fire kiln with no thermostat. Of course, those are absolutely viable options and may make the work more interesting in a way, but lack of them wouldn’t disqualify it from being “handmade”.
"Handmade" can also mean it's made on a factory line in China by underpaid workers. One has to remember that all it means is that some part of manufacturing that you expect to be automated, isn't.
In a way, any type of "No AI" statement is the same. It generally means that I can expect that time was devoted to it, but it doesn't say anything about the quality. I know I'm perfectly capable of writing dog-shit code without AI.
but how is “quality” even defined?
This caught me. I happen to be re-reading 'Zen and the Art of Motorcycle Maintenance' which more or less revolves around the insight that quality cannot be defined (spoiler alert: because it is determined at the moment of observing, preceding the intellectual interpretation of a fact, even preceding the distinction between object and subject, meaning scientific methods like a definition cannot be applied to it. But please read the book).
If quality can't be defined and science can't be applied to it, does that mean that at the most fundamental level, AI will never be able to purposefully assess or produce quality?
Meta: could we get like pro-ai and anti-ai tags, so every side could just ignore people's copium/hypium?
Note: this comment was create by a human.
I think I'd rather see ai-use and ai-discussion, instead. Sometimes views should be challenged, both ways? But yeah not knowing when AI is used in an article is tiring. So many articles closed at the first mention of "I prompted," sorry guys.
(This comment is also entirely human-keyboard-written.)
I really still don't like AI. But I hate being in an echo chamber more. I want to read both sides, though I suppose if you did that I could just continue to read pro-ai.
I feel the problem with submissions like this one, and specifically the comments in response to them, they are manufacturing consent, so the "I usually don't like AI," and the "I was a skeptic but" type comments are often just advocates using rhetorical tricks. I don't know what labeling will do since the entirety of the LLM advocate community has decided that what is not even wrong must be right.