Classifying and Labeling AI-Developed Works [Developed without AI]

Authoring
41

I just want to say, I don’t like “Human made” at all. It sounds like one of these vague corporate labels that don’t actually mean anything, like “environmentally conscious” or “We are committed to ensuring good working conditions for our employees”. Those things you see everywhere and that sound good, but are in fact not contractually and legally binding.

Also, “human made” (or some similar slogan) is already used here and there to, in fact, justify an “ethical” use of generative AI. For example, the Youtube channel Kings and Generals uses it, because humans are indeed making the videos, but they still use genAI for portraits and illustrations. The slogan is therefore deliberately misleading. (Note that the channel is very good, though.)

If we’re going in this direction, the slogan should be explicit, like in “No AI” or “No AI on visible materials”.

42

I think not, it could be of interest to people to are blocking the tag, at least I doubt it would annoy them.

43

Well given the newly added direction of once again litigating what counts as genAI…as someone who has the tag blocked, i’m annoyed. i’m going to add it. if people disagree, remove it again. i thought this convo would be about fun logo design and celebrating/promoting stuff where no genAI was used. now it’s another llm discourse thread.

44

@heasm66

After reading all the opinions here, I’m kind of leaning towards a more grassroots vibe. Like, “Crafted With Care” or “Made With Care”. Then you can explain the specifics of what that entails. Trying to make a nice label that’s completely self explanatory and accurate would start to feel robotic. There’s nuance to the details of your label so strike at the heart of your intent instead.

For me, “with care” means deliberate and conscientious so I think there’s some weight to that word. When you ask, “do you care about this?” it means are you significantly attached to it in some manner? It think the word goes both ways from writer to reader.

Anyway, my (hopefully) last two cents on the matter. I think it’s way better than my other idea of “No-AI Adjacent”. :wink:

45

This has gotten a bit sidetracked, yeah.

I think what will end up working best is if people with strong feelings on this—some people have been saying they don’t want to touch games coded with LLMs, so the label would matter a lot more to them than it does to me!—come up with a definition that works for them, get a logo, make a tiny website (carrd?) that explains what it all means, and then let authors include that logo with a link to the website if they’re onboard.

At which point I’d gladly put it in my games! But I don’t think I’m the right person to coordinate this.

46

Honestly, requiring every line of code be typed by a human sounds like a ridiculously high bar with no real benefit. I’m pretty oldschool, writing code in nano and invoking compilers directly at the command-line, but unless I’m coding helloWorld in a new to me language or writing a bash script, I never start with a blank file instead of a template(and the only reason I don’t have a template.sh is because it would be nothing but #! /bin/bash) or copying the source of a program I wrote previously that does something similar to what the new program is supposed to do and if I have several lines in a row or several branches of an if block that do very similar things but with slight differences, I’m probably going to write from scratch once, make copies, and edit the copies instead of typing each variant by hand… and I would imagine things like tab completion on existing variable/constant/function names, automatic indentation, or auto generating the code to close a freshly opened code block are baseline features for an IDE. Kind of feels like dinging an artist for using a drawing program instead of using paint/ink/pencil/etc on paper/canvas or dinging an artist for using compass and straight edge or line and curve tools in a drawing program instead of drawing all their lines and curves free hand.

Also, I’m sure there are people who would reject a cancer cure because of AI… how many would still reject it if it was for their own cancer or the cancer of a loved one is probably smaller than those who would reject it only as long as it was only a stranger’s cancer in question, but I suspect both groups are non-empty… Afterall, we still have a disturbingly large number who are anti-vaccine based primarily on one debunked study from like 20 years ago and you have people who will praise content right up to the moment its revealed to be AI, so don’t underestimate people’s ability to be stupid, anti-science, and/or willing to reject stuff for reasons that have nothing to do with its quality. I suspect the main reason you don’t find people getting out their metaphorical torches and pitchforks over AI-assisted medical research is that the AI is bad because AI is bad crowd are too preoccupied with the perceived cheapening of human creativity and the general modern medicine is a scam crowd are too visible and vocal to focus on those who are pro-medicine, but draw the line at AI.

47

Here’s my take on @heasm66’s idea:

A line-art figure, reminiscent of Auguste Rodin's "The Thinker" but more upright. Across the bottom, it says, "By humans, for humans."
A line-art figure, reminiscent of Auguste Rodin's "The Thinker" but more upright. Across the bottom, it says, "By humans, for humans."200×300 12.6 KB

48

I probably wouldn’t take a vaccine whose formula was developed via Inform or TADS, personally.

Maybe Adventuron.

49

Fair enough.

I do maintain that the original topic of this thread, a label and/or logo to distinguish AI-free works, is one that even the most ardent AI opponent might be interested in talking about.

50

There are lots of independent efforts to produce a ‘Not by AI’ logo. Here are links to a few of them:

There’s also a Wikipedia entry on the subject at No-AI label - Wikipedia

51

This is sort of what I did at my repo for Quendor. GitHub - jeffnyman/quendor: A Z-Machine Interpreter and Debugger. · GitHub

I have a badge that says “NO AI : 100% MADE BY HUMAN” but I point to a section called “Crafted by Humans, Optimized for Everyone” and I do that because I have AI helper files in the project, and someone might wonder at that disconnect.

Personally, and I speak of this from a hobbyist and a professional context, I find that the amount of time people worry about this is pretty staggering. What matters is the experience. If I get code that works like crap, I don’t really care if an AI or a human did it. If I read a story that I don’t like at all, it matters little to me whether a human or an AI wrote it. After all, I still don’t like it. So disclaimers of this sort really don’t matter to me, but I figured I would try it out on a project or two.

Mind you, I’m not saying my standard applies to everyone and I get the distinction in the community. If something was written by AI or AI assisted, there might be an expectation that it’s going to be bad and people don’t want to waste time. But does usage of AI extend to, say, “I had a story idea. I wrote it down. I periodically used AI to help me tighten up some plot points and suggest puzzle situations in areas that I felt were too dull. But then with its suggestions, I coded it all up on my own and tweaked its puzzles.”

Now, if I have to put “AI Used” or whatever, someone might jump to conclusions that what I produced will be crap. But that might be totally wrong. AI usage is a spectrum.

The current sentiment in the professional world is rapidly this: “Assume AI was involved in anything I create. In fact, assume I use every tool available to me. If we’re going to require people to disclose AI use, then let’s also require disclosures for spellcheckers, grammar tools, editors, copywriters, researchers, photographers, assistants, and staff. At some point, the distinction becomes absurd. What matters isn’t whether a tool was used. What matters is who directed the work, evaluated the output, made the decisions, and accepted responsibility for the result.”

In the creative fields that I work in, the sentiment is becoming more this: “A photographer doesn’t build the camera. An editor doesn’t write every word. A film director doesn’t operate every piece of equipment. Yet we still recognize their contribution as creative work. The real question isn’t, ‘Did you use AI?’ The real question is, ‘Are you willing to stand behind what you published?’”

I’m not saying I agree or disagree with those stances in whole or in part. Just putting them out there to add some flavor.

52

Everyone has their own truth. In a manner of speaking a persons perception becomes their reality. Truth is stranger than fiction, so the phrase goes. My truth is that I’ve spent a lifetime practicing the craft of computer programming from the perspective of an artisan. I did not spend my life learning how to write stories, plots, narratives, and all the things I truly enjoy when I read or listen to a good story. Now I’m faced with two untenable choices. To AI or not to AI, that is the question. As a programmer what I object to more than anything else is the cheese pizza problem. I just want people to be honest about it. When it comes to writing IF, I’m held to even a higher standard. That story I write better 100% be written by me because if it wasn’t then the performance floor kicks in. It is 100% out before it is in at all. It is not authentic. So my truth is that I have to learn how to write now, which is not a bad thing, but it does mean that my writing is probably mediocre to bad at best. If I could just use AI to help out, and that is the crux of the problem. If it is OK with computer code then it needs to be OK with fiction. But it isn’t or is it? Many people spent their life learning how to write stories, become masters of words and grammar and their untenable position is very clear on the matter but they are fine with using AI to aid them in making some automation on the computer; after all it is just a new medium to publish to. In computer code words, syntax, and structure are just as important to the programmer as the word salad is to the author so should the standards be the same? I think that is the real question here. For myself I just think about a phrase that is very important to me and comes from some word salad that AI had to explain to me and that is when in Hamlet Polonius says to Laertes “Neither a borrower nor a lender be; For loan oft loses both itself and friend, And borrowing dulls the edge of husbandry. This above all- to thine own self be true”

53

To a lot of people, that’s the question (and the dilemma) these days. And, in my case, I’m never arguing against disclosure. I’m pretty much always arguing against the reflex where “AI was involved” auto-translates to “so it’s probably bad.” Those are different claims. I think the IF community’s version of the second one is doing something specific: when the whole contract with a player is “a person built this trap for you to out-think,” swapping out who built it isn’t a smaller version of the same experience. It’s a different experience wearing the same label. (Granted, that’s an arguable statement.) In my view, that’s not the community being harsher or more suspicious than programmers; it’s them being right about what their form actually promises somebody.

That being said, it’s a side note next to what you’re actually describing, which isn’t really a policy question, as I read it. What I’m hearing is “I’ve spent a life getting good at one craft and I’m now being asked to be a beginner at an adjacent one, in public, under a standard I didn’t get to set.” That’s just a hard place to stand, and I don’t think my “AI usage is a spectrum” statement fixes it, because the spectrum isn’t the thing that’s bothering you. The thing that’s bothering you is the gap between the craftsman you already are in code and the craftsman you aren’t yet in prose.

For what it’s worth, the Hamlet line you reached for kind of already answers this and, I would argue, in a way separate from the AI question entirely. “To thine own self be true” isn’t “never use tools.” It’s “know what you’re actually claiming as yours.” If you used AI to get the reference, decided it mattered, and put it to work, saying something true about your own situation … well, is that so bad? If your puzzles end up built the same way (you deciding, a tool occasionally explaining), that’s not a violation of the standard you’re describing. That is the standard.

54

I’d recommend moving the “is AI use good or bad” debate to another thread and keeping this one for discussing the label itself.

55

Some of you people use AI and don’t feel wrong about it. They asked not for permission. So why should people who are against AI ask for permission. I wish we could talk about a no-AI label without the folks for AI debating.

56

I’m unconvinced that a standardised “no AI” label is going to work unless someone wanted to actually trademark a design and license it out, with all of the attendant expense and legal rigmarole that that involves. If you do it wholly on the honour system, it only takes one person deciding “well, my use of AI doesn’t count so I still get to put this badge on my project” to undermine the whole thing.

57

Exactly. Which is why the conversation drifted because, after all, why wouldn’t it? If someone wants to use a label, create one and use it. That was my point with my Quendor example. I don’t really care what people think about it one way or the other. I just did it. Then I called out some nuances around it so people would understand exactly what my stance was on the repo.

A label is also not something you can really prove against one way or the other. Case in point: I have routinely seen developers scrub “co-authored by” commits so it didn’t look like they used Claude Code in their commit history. So there is still a bit of an honor system with any badge someone chooses to use.

58

This is a weighty sounding sentence, but in a world where a clever prompt produces seemingly polished work, the standards of what someone will stand behind will be less and less as time goes on. People will stand behind anything they are invested in (for better or worse).

I think it’s more useful to ask, “Is this your authentic voice and thoughts being communicated to the audience?” This contends with your dilemma of AI usage equating to slop. The slop is the undesirable aspect of AI content, which I think can be boiled down to something that doesn’t jive with human sensibilities.

You mentioned AI usage and acceptance is on a spectrum (that’s two spectra combined). A slightly ambiguous question, that requires a second more of thought to decipher, will uncover deeper truths.

59

Sure! There is. If there wasn’t why a label if anyone can spot AI usage?

But sometimes there is evidence that a programming piece or art piece claiming to be AI-free actually used AI. You can use even AI to check for AI suspect.

61

[!warning]- Ironic AI Created Example

OIG4.gU9Pr.
OIG4.gU9Pr.1248×832 91.9 KB

Or this might be a bit less negative.