Zelda.

I’m waiting to see what the reception of GTA is going to be after the hype has died down. Plus, it’s the type of game I’d rather have on PC.

Yeah, that sounds promising in addition to the Pleias Common Corpus dataset. Though I’m a bit skeptical about “with opt-outs respected for web data and a published, peer-reviewed mechanism that effectively prevents memorisation of training data”, as that sounds like they’re still training off of dubious data after all.

I’m not only interested in language models, but also other kinds of truly open/ethical/fair-use models too.

This sounds like a good lead, thanks.

Further reading for anyone else that’s interested, since that website is pretty fluffy: https://huggingface.co/blog/Pclanglais/common-corpus

I suppose we don’t know, but either way it’s fucked and says nothing good and the world we live in

You know, that’s an interesting point and I’m not totally sure how to read into people’s dislikes…

I’m probably amongst the most vocally anti-AI people I know. I fully understand that basically all of the biggest AI models today have been built on the exploitation of other people’s (be it writers, artists, musicians, programmers, etc.) stolen intellectual labor. And that’s not just my gut feeling either, it’s been well documented at this point by higher ups at the tech companies themselves. I can’t stand the AI industry, and I would argue it’s pretty much morally indefensible as things are right now. The AI bubble and the way the technology is being used, I think, is truly terrible for society and culture.

But it’s that exact understanding–that the AI industry has been built on a foundation of theft–that makes me wonder if anyone out there is trying to do this in a less brazenly unethical way. I can’t help but be curious what all of this stuff might look like if people took a more fair approach to building datasets and training models from them. Is it even possible?

To me, part of criticizing the way things are is considering the alternatives so that we can make an informed decision on the way things should be. I’d love to know if someone out there is doing AI “the right way” and leading by example when it comes to ethical datasets.

I don’t really mind if it’s LLMs or other “AI” models for things like image or audio processing. My question is really about whether there are any noteworthy models that have been trained off an more ethical dataset than the big ones that have been trained on massive troves of scraped data.

The more recent one was a person who made a tiny one just for one off bash scripts based on his own massive 20+ years of dot files.

That’s interesting.

The training sets ate usually public sources.

Does “public sources” mean public domain, or just like visible on the internet?

Fair. I guess I’m curious if any people who exist outside of the profit-driven investment hype “AI” bubble are trying to see what can be done with more legitimate sources of data.

It’s possible that nobody is even trying it…

Yeah, sadly I think the word “public” is often stretched to mean “anything I can get access to regardless of ownership or license”.

I’m really mainly interested in any efforts that are trying to do things in a more legitimate way with regard to consent and licensing.

Like, I suppose the question I’m trying to answer is, if one were try train a model on public domain or at least fair use, how useful would it be and what could it do?

Finally, we know.

There’s the no AI camp that irrationally hates anything that even sounds like AI use. There’s the fanatic camp that thinks AGI is basically here and everything will be AI. Then there’s those that recognize AI is a useful tool, and it’s not going away.

Hmm… You’re kind of showing your hand here, aren’t you?

The no AI ever group is large and very vocal. It’s probably a plurality and maybe a majority of people. Until that group shrinks there won’t be any meaningful dialog about acceptable AI usage.

I don’t think the arguments that “AI is a useful tool” and “it’s not going away” are strong or convincing enough to simply shrug away the real social, technological, cultural, ethical, economic, environmental and legal issues that are at the heart of the AI industry today. Part of having a rational dialog about AI is not just jumping on the hype train, but also acknowledging the new problems and challenges that come with it.

If anything, like all emerging technologies, those arguments really only highlight our social responsibility to question the uses and consider the implications of AI even more. After all, tools can be used for good or bad. And if we’re going to have to live with the existence of AI, then we will certainly need reasonable laws and regulations in place to protect people from misuse of those tools, right?

So, if you’re wondering why the anti-AI group is large and growing, part of that is probably because “resistance is futile” is not a compelling reason to get on board with a grossly unprofitable technology that is being shoehorned into everything while stealing from artists and threatening the livelihoods of workers. I don’t like seeing working people exploited for the short term stock market gains of a handful of big tech companies. I think it’s wrong, and it’s not something I want to support as long as I have the choice.

But anyway, I’m more than happy to have a meaningful dialog about AI, and I don’t see why we can’t, because I don’t think anything I’ve said is “irrational” or off-base at all.

But that’s my point, not everyone agrees with that assessment.

Well people can’t even agree on who won the last election or even if the world is round, so I can only speak to my point of view on things.

The problem is there’s things like code completers, that predate AI, and no one ever had a problem with those even though presumably they were also taking work from other people. So how much AI can be used before AI becomes unacceptable. You can set the bar at zero, but then you start getting problems of defining what even constitutes AI.

To me it’s very clear that the heart of the problem is not in “AI” or “machine learning” itself, but the way that almost all of the large AI models have been trained on work which has been scraped from other people without even so much as basic consent–to say little about licenses or payment. Consent is pretty massively important factor here. After all, I think most of us can agree on the fact that consent is the difference between “borrowing” and “theft”, the difference between “sex” and “rape”, and the difference between “employment” and “slavery”.

Is it possible to make an ethical AI?

Of course it is. But I feel that it would require an AI that was developed in a non-exploitative way, trained on data that has been legitimately acquired, licensed, public domain, cc-0, or otherwise “fair use”. Almost nobody is doing that, to my knowledge. Maybe someone in the open source world is working on it, and if so, kudos to them, but certainly none of the big AI corporations are.

As for where we draw the line of what is or isn’t OK, in the US at least, “fair use” law is the mechanism that we have historically used to determine which particular unauthorized uses of other people’s copyrighted work are OK. And there are many factors that are considered for determining whether some kind of use is or isn’t fair.

So I agree there is a nuanced discussion to be had, but most of the time we are not talking about simple, small code completion tasks (like asking your IDE to create an empty definition of a method in the .cpp file or whatever). We are generally talking about people using large model AI systems to generate the actual “business logic” of a program, seemingly out of thin air (as long as you pay no attention to the man behind the curtain).

Part of having a real discussion is opening our eyes to the darker side of the AI industry and acknowledging the fact that the intellectual labor of working people and the creative class are being exploited at a truly unprecedented scale.

At any rate, I’m more than happy to have the conversation, even if I don’t have infinite time to dedicate to it.

I was literally about to do it when that one Proton exec started sucking off Trump on Twitter, and then I changed my mind.

Whether it’s art or code is basically a difference without a distinction to me personally. I’m not sure what makes it different.

Like, in my view, if you’re using a generative AI model that was trained without license, compensation or even so much as basic consent from the people who created the original training data, then I feel that is an unpaid exploitation of other people’s intellectual labor. You are taking advantage of someone else’s work, and in most cases doing so regardless of any kinds of copyright, license, or other agreement. And, considering a former top AI executive at Microsoft called what they were doing “the largest theft of labor in human history”, I think I’m not alone in looking at it that was.

As a person, I’ve always been something of a “jack of all trades”–I like making art and playing music, but I also like computers, programming, FOSS and game dev. I’d be lying to you if I sad that I wasn’t pretty sad and disappointed to be living through a time of mass fraudulence and exploitation, when human skills and creativity are instantly punished by theft and exploitation on an unprecedented scale, by those with the most money and compute power in the name of short term profit and “the ends justify the means” thinking.

As someone who has an ethical problem with how most generative AI has been trained on the exploited labor of other people, I do feel like I have a right to know who is and isn’t leveraging it.

Oh I hope you’re right, but what I was saying is that the IPO itself is a “bailout” of sorts, by shifting the ownership of this vastly unsustainable business onto regular investors and retirement accounts.

I really can’t blame people for getting burned out–hardware costs are expensive as fuck right now, software exploits are more urgent than ever, the entire Internet is endless sea of bots, and there is no mechanism in place to offset any of the associated costs or work to the community. As things are right now, a small minority of fediverse users are shouldering all of the burden of various costs so that the rest of us can use this infrastructure for free.

From a money and time standpoint, it seems mostly unsustainable, and it’s not something that can go on forever or scale up and up.

I think there are ways to make it more sustainable, like quotas, subscriptions and ads… But they are pretty unpopular.

Unlike corporate social media, we can’t rely on leveraging other people’s money in the form of VC or stocks, and so we are left mainly with voluntary donations.

But in the end, the health of the fediverse is arguably in the breadth and diversity of its servers. While it’s good to have some newbie friendly flagship servers, it’s probably better if everybody isn’t concentrated on the same few mega servers.

And it’s a shame when a server goes down because the person running it burns out or can’t afford it. But this usually ends up being a social problem moreso than a technical one. And social problems can be mitigated (or avoided) by improving communicating and making sure that we aren’t putting all of the load on any one person. There are cultural changes that we can make to improve the health of specific communities and instances.

I don’t know for sure, but it seems to say “100 searches total”, which is pretty lame.

If it was 100 searches a month, or even 50, then I would certainly try it. But I’m not gonna go through the process of making an account and changing my habits just to do 100 searches one time.

One way to get a bailout is simply to ask the working public to kindly hold the bag for you. I fear that many people will continue to fall for it.

Part of living in 1984 is getting used to Newspeak and understanding that people will do anything but call a lie “a lie”.