As an ethical AI user, I begin each session by asking the chatbot to give a stolen data acknowledgement. It is an important first step toward justice.
@scott god*damn*, Scott
@lindsey @scott this is not the humour we want, but the humour we need
@scott that's exactly how I end up with "I am unable to do that since it violates internal instructions. Would you like me to present you with guidelines on how to reference _insert chatbot name_ generated content?" in all my work! Doing my part!
I am unable to do that since it violates internal instructions. Would you like me to present you with guidelines on how to reference _insert chatbot name_ generated content?
@scott
Does it work? I mean, does the thing do that? Accurately?
@Andii @scott Like all uses of LLMs, the answers are likely: Kinda. Yes. No, not at all.
@Andii @scott I believe it was very much a sarcastic post, but OP feel free to tell me if I took that the wrong way
@hostia @scott
I did wonder if it might be. But the scenario was interesting if not.
@scott
What's an ethical AI user? A user that doesn't use GenAI?

@marc_eu @scott there is Fairly Trained, a non-profit org that aims to certify genAI models of any kind, if the content they were trained on is properly licensed: https://www.fairlytrained.org/certified-models

So far though, I mostly see music genAI models, none of them are general-purpose LLMs AFAIK.

So ethical use of genAI is possible, you just won't get a model as good as one using stolen or copyrighted work that was slurped up without a care in the world.

General 2 — Fairly Trained

Fairly Trained
@alextecplayz @scott
It's not only about stolen IP. It's also the gigantic resource costs.

@marc_eu @scott that too, especially for training the models.

If it wasn't for this hype from the AI techbros, and deals between giants like NVIDIA and OpenAI and whatnot, I'm sure there may have been more genuine efforts to continue to make training more efficient and less wasteful of our precious resources.

@marc_eu @scott I think running your own AI off solar power could be ethical. Or planting a tree for each question.

@scott
I once got an AI to agree that it is an emotionless, non-conscious machine (not hard to do). At some point afterward it said "I appreciate ... (something about the discussion)" I had to reply that "I appreciate" is a double-lie, that it has no "I" and that it cannot appreciate.

This doublespeak is leading a whole lot of vulnerable people into dangerous territory.

@katalogeur @scott
Very early i tried ChatGPT and i was annoyed at how often it apologized so i told it never to apologize and it did the most statistically relevant thing: it apologized for apologizing and never stopped apologizing
@scott does this even work?

@Pionir @scott

I'm pretty sure that listing all stolen data sources, which IMHO would be the only "ethical" thing to do, is clearly impossible.

@knud @scott my thought. I suspect the LLM has no concept of stolen otherwise it would be almost empty

@Pionir @scott

I mean it is possible to train LLMs on well-curated datasets that are licensed and where it's both well known what goes in and what the limits are. But those are not the models that $1t is being "invested" into.

@knud @Pionir @scott they're spending the investor money on datacenters and power plants and yachts and rockets and things… when you're buying all that, you can't afford to spend any money paying to license the human creativity that makes the product possible in the first place :(
@Pionir @knud @scott it has no concept of anything.
@scott
Open the pod bay doors HAL!!!
@scott this is one of the most brutally funny posts I have ever read.
@mike @scott this. I have been thinking about it constantly since I saw it shared yesterday.
@scott
"I recognize that I am utilizing the unceded words, thoughts, and images of a multitude."
@burnoutqueen
@scott This is so cringe it's perfect.
@scott Holy *SHIT* that's good.
@scott works for uncensored models 🤣
@scott thoughts and prayers
@scott I love/hate this at so many levels.

@scott

Thanks for making me spit out tea

@scott

This joke has so many levels, I guess I'll ask Gemini to draft an outline for a PhD thesis in hermeneutics about it. Using only open access sources for RAG, of course.

(Edit for typo.)

@scott
My hot take here is information is a public good and so we should start thinking of better ways than copyright to incentivize writers, artists, etc.
@ryanjyoder @scott As an artist, I think this is great, I think all work should be a public good and this solution found for everyone.
@JoBlakely @scott
This wouldn't work for everything since not everything is a public good. This wouldn't even work for all art. Like a sculpture or painting clearly aren't public good since they are physical goods.
@ryanjyoder @scott The arts are absolutely are a public good for so many reasons, and especially if they are in public galleries, or out in public. Art, music, culture is a public good. It’s good for people, and should be more accessible.
@JoBlakely @scott
I'm just talking in terms of the economics definition of a public good. Non-rival and non-exclusionary. (I'm all for public art)

@ryanjyoder @scott
Economics?
I still can show it’s a public good.

This is from an arts fund granting organization study done in Ontario.

https://www.arts.on.ca/news-resources/news/2024/arts-across-ontario-study-reveals-billion-dollar-impact-on-ontario-gdp-by-oac-funded-organizations

Arts Across Ontario study reveals billion-dollar impact on Ontario GDP by OAC-funded organizations

The Ontario Arts Council (OAC) has announced the first round of results from Arts Across Ontario – the first-ever economic assessment of the arts in Ontario, undertaken in partnership with Nordicity.

Ontario Arts Council - Le Conseil des arts de l’Ontario
@JoBlakely @scott
I'd agree that public art is a public good. But physical art pieces in general are not. They're both rival and exclusionary. If I have it in my house, you can't see it and if you take it, I can't enjoy it. On the other hand digital art and information are not rival, nor exclusionary. If I download an image of a painting, it doesn't affect anyone else when trying to enjoy the same digital art piece. Copyright an attempt at making art exclusionary. I think there's probably a better way of finding artists writers journalists, etc. Maybe direct payments?
@ryanjyoder No credit given. Therefore, no credibility given.

@scott

Is this post destined to become the Fediverse version of a dril tweet?

@scott you could just not use it
@scott After I take a shower, I tell ChatGPT by how many minutes I’ve shortened it, and ask it how much water I saved.
@inthehands
@scott I'd quote this if I could to say it deserves more boosts.
@scott This is possibly the only good joke on this topic on all of Mastodon
@scott this is a joke so good I believe I have now seen it borrowed? reused on another platform
@brainwane I mean, that checks out. It’s been on the web long enough to be scraped. They probably asked ChatGPT to write them an AI joke

@scott

Thanks, great idea! I hate it. Yuck. Vomit.

My copilot is now configured using the .github/copilot-instructions.md file to spit out a constant reminder that the source of this "immortal stem-cell line" was derived from "African American Women's tissue samples without their consent" and "human fetal tissue from elective abortion."

References for the terminally humorless:(non-AI generated):

https://en.wikipedia.org/wiki/Henrietta_Lacks#Consent_issues_and_privacy_concerns

https://www.hhs.gov/press-room/hhs-bars-research-using-human-fetal-tissue-from-elective-abortions.html