There is no such thing as "ethically using AI". AI is trained on stolen data. AI is destroying the environment and communities. AI is making us stupid. Use AI if you want, just don't pretend you're doing it in ethical way.
and make sure to remind AI proponents that using AI makes a person stupider. then, ask them how often they use AI.
no matter what they answer, just say 'well, that explains it, then.'
if they ask for clarification, tell them to figure it out. and laugh
@grumpygamer I disagree on the blanket condemnation. Frontier labs? Data centers? For-profit? Unquestionably unethical, no doubt.
But what if someone ran a model on their local machine to produce code that helps a community? Aids in volunteer work? Much of the data was stolen, but if in turn that data is used to help the commons, is that unethical?
I have hope such cases will be less rare when the for-profit industry is destroyed by its own hubris and greed.
@grumpygamer I came across this shirt a while back, and I think it defines the subject brilliantly.
@grumpygamer the word "hacker" is lost, and AI followed it.
What do you call the things now that were called AI before GenAI highjacked the term?
@grumpygamer I have been signing petitions against AI and data centers almost daily now. I joined a black out protest against meta and logged out of my accounts on instagram and Facebook for a week. And just generally voicing my distaste towards AI whenever the chance arises.
@grumpygamer There is plenty of "AI" that isn't LLMs.
And not all LLMs are "trained on stolen data". I used one that was trained on the document set for a particular software system, all owned by the same company, and was actually not bad at answering questions about it.
@grumpygamer I'm not sure how much AI is involved, but my cellphone carrier is using speech-to-text software to supplement normal voicemail. That saves them bandwidth & you can just look at the text, which means you don't have to get out of a noisy room to find out that someone is stuck in traffic and will be 10 minutes late. For anything important, you can double check the voice recording. A LLM might help a bit by not rendering phrases like "... is too big" as "... is two big".
@grumpygamer You don't have to train AI on stolen data. Cascadeur isn't.
I too am a strident critic of the generative models currently being hyped as the future of "AI";
https://disintermedia.net.nz/invasion-of-the-mole-trainers/
But to advocate for the demons for a minute ...
@grumpygamer
> AI is trained on stolen data.
What if the "AI" is not a generative model? Or what if it is, but trained entirely on data in the commons, with full respect for its licenses?
> AI is destroying the environment and communities
Maybe no more than streaming video, or online gaming;
https://medium.com/@meghannbf/your-screen-your-footprint-dd9cd535685d
Or any other use of the net that depends on hyperscale datacentres, instead of using renewably-powered, 'local-first' or 'edge' computing. Downloading an entire Office suite every time you edit a GoggleDoc in your web browser must use a fair whack of resources.
> AI is making us stupid.
Writing things down instead of using our memories is making us forgetful. A good argument for making sure we exercise our memories (eg learning songs, or acting in community theatre). But not necessarily an argument against writing as a technology.
My point is, slamming "AI" here is shooting fish in a barrel. Coming up with arguments against generative models that will convince institutional decision-makers is much more challenging.
@grumpygamer
Almost all data - everywhere - is from "stolen data."
I can’t help but feel there is always greyscale and nuance with practically everything. I’ve come across really cool research like using machine learning to detect deforestation in the Amazon in real time. The more powerful approach is to appropriate AI, rather than reject it outright which is pointless, the genie is out the bottle. We must demand ethical development of the industry, with plenty of guardrails, constraints and excessive caution. The bigger task: end the corporatocracy.
@grumpygamer I agree.
For people who want a longer discussion (whether for themselves or to send to others) I like this piece by Emily M. Bender and Decca Muldowney -
Six Problems With Talking About the “Ethical” or “Responsible” Use of “AI”
https://buttondown.com/maiht3k/archive/six-problems-with-talking-about-the-ethical-or/
Exactly. I’m adopting a zero tolerance model now, and unfollowing anyone if I see them referring to using LLMs, regardless of how “ethically” they think they’re doing it, or how much I might otherwise see them as good people.
I’m especially dumbfounded at supposed creative types who “only” make use of it in ways not directly connected to their primary art form. Hypocrites happy to throw others under the bus.
Just remember, ai-ers, you steal from me and you also support CSAM and pedophiles each time you use that fucking shit.
@grumpygamer Look I don't agree with how most models are trained either but AI is not going anywhere. At this point the models have been trained, it's done. There is no putting the genie back in the bottle.
@grumpygamer
That's misinformation and #ethic-slop.
Nvidia Nemotron3, AllenAI Olmo, Swiss-AI Apertus, and Hugging Face SmolLM are trained with public data or synthetic ones, with intellectual property filtered out.
Also, inference can be run with green energy, just like what Rogolo AI, Infomaniak and some companies do.
If anyone wishes, they can run LLM locally with solar energy as well.
You can vent your anger everywhere, just don't pretend it's because of ethic.
@grumpygamer Lotta people having trouble with this one for some reason.
@themalikficent_ The same way one owns a slave ethically...you can't.
@frederic Said like a true cult member.
European AI (especially from The Union) like French Mistral IS NOT trained on stolen data, is powered by EU green energy and is not destroying anything because IT HAS TO OBEY OUR LAWS AND STANDARDS. Just like every other company in EU.
Don't blame technology and progress for flaws of your dysfunctional country. 😒
@grumpygamer Using AI is like using someone's stolen goods. It should be illegal.
@grumpygamer I guess it depends on whether you consider machine learning to be "AI", but that is being used to find or characterize type 1a supernovae, and the data used for "training" is astronomical data, not human-written text. Type 1a supernovae are useful as "standard candles" - we can use them to get accurate distance measurements for distant galaxies.
@grumpygamer niggas be on here bitching about anything bru
@grumpygamer
'AI' should be specified in general phrases like that. I would totally agree regarding LLMs and generative AIs. But there are a lot of usecases for mashine learning, which are not based on stolen data and improve processes like in medicine, weather forecast and environmental research.
@grumpygamer i think i found the perfect analogy https://mastodon.social/@bison/117517691955861827
@grumpygamer And profiting sociopathic, acquisitive, environmentally-damaging and paedophile individuals at the expense of the rest of us.
@grumpygamer it’s fine to be angry about AI (or LLMs) as a technology but saying it has no potential to be ethical is just untrue.
In fact, AI needs more people who are passionate about making it ethical — about fixing attribution in training data, about improving local inference.
The people working on this are out there and they could use your help. Current AI is a good starting point https://www.currentai.org
@grumpygamer there is no such thing as “ethically using computers”. Computers are built by corporations exploiting their employees. They built factories and supply chains that destroy the environment. Computers are making us stupid. Use computers if you want, just don’t pretend you’re doing it in an ethical way.
s/computers/smartphones/g
@grumpygamer I think the authorities need to fix IP laws around Generative AI.
That's upsetting a lot of people. And it's true that most LLM's would work very differently, if they were only restricted to content in the public domain. I am amazed by the tech, but if it disappeared overnight, it wouldn't bother me.
The environmental argument is different. I'll admit to not fully understanding it. Couldn't the same argument be levelled at computing, generally?
@grumpygamer This is the same argument many used to oppose certain vaccines that were tainted by a fetal cell line of an aborted baby. https://cogforlife.org/ #prolife #ai #ethics #technology
There would be if you built the AI from scratch on data that you gathered ethically, but good luck getting that done... 😞
- replies
- 0
- announces
- 0
- likes
- 0
@grumpygamer i guess the correct name is LLM. AI is much more than LLM. I'm my previous project we used machine learning trained with meteorological data. None of the criticism you mentioned is valid in that case
@aldonogueira @grumpygamer If you're calling it AI, you're participating in the game of confusion. The only reason to positively call something AI is to make it unclear what you're actually talking about. If you're talking about machine learning, call it that.
@ahltorp @grumpygamer when i was in college they taught me AI is the broad term which includes many techniques, but in recent years some companies stole the term for LLM. I'm not the one who created the confusion. But we used to call it machine learning in this project
@aldonogueira @grumpygamer "AI" has always been an almost meaningless term. It was invented as a meaningless term, and from then on it has only ever been used as a buzzword.
The only meaning the term had before the current craze was "what we haven't done already".
@ahltorp @aldonogueira @grumpygamer the most previous standarized use of AI before current world situation was to talk about NPC behaviour in videogames. AI manipulation, difficulty and adaptability are probably some of the most talked about topics on the internet before FUCKING BILLONAIRES started with that shit.
Sorry about the FUCKING BILLONAIRES I was sounding a bit too polite and started to feel nauseous about myself.
genai
@Nawer_Rapter @ahltorp @aldonogueira @grumpygamer very confusing to be in gamedev discourse right now :P
Inclined to call NPC behavior "real AI", and automated plagiarism "LLMs". But, not all my peers agree.
https://smartsideincome8822.blogspot.com check out the link for some good remote job guides and ideas. Hope it helps!
@grumpygamer and it's ridiculous how quickly we invented something that ruins the climate even more effectively than any single thing ever before
@grumpygamer Theoretically, I agree. But couldn’t the same be said of many other technologies? 🤔
Is there such a thing as “ethically using a smartphone”? Lithium mining destroys lands and communities. The "capitalism of attention" turns us into its slaves. And surveillance and social control are spreading to unimaginable levels thanks to the digitization of every aspect of our lives, of which the smartphone is both a symbol and a tool.
@grumpygamer There is a way to run this software ethically. That way is to use the open source software against your own database that you build with ethically sourced and well curated data. There is no ethical way to use the publicly available professional AI (Automated Incompetence) engines because they are built by thieves and run on data that thieves have stolen. Don't let yourself be fooled into thinking otherwise, or that it does not matter. It does matter.
A musician listening to the Beatles, and later performing "Let It Be" on stage. An artist painting Spiderman and selling the paintings. Both are breaking copyright unless they have a license agreement. It's not theft, it's infringement. The artist's training is not theft.
The AI models training is a grey area at best. Humans can also recite text from memory.
But AI is something new. AI can produce unlimited works of art in any style. We simply haven't had these tools before.
Anyone can vibe code anything. Software that needed years of dedication, and a team of developers, can be written by a single person over a weekend.
The world is changing, but it's not all bad. We need to embrace AI, and we need to democratize AI. AI must not fall into the hands of a few select corporations. We tried that with Google, facebook, Apple and the likes. We must not make the same mistake again.
@grumpygamer I kinda agree but I would argue using it for finding cures to currently incurable illnesses (like my spinal muscular atrophy), which also likely need highly specialized models, could be considered ethical regardless of whether they used public datasets for training or not.