All it would take for AI to completely collapse is a ruling in the US saying these companies have to licence the content they used to train these tools.
They simply would never reach a sustainable business model if they had to fairly compensate all the people who wrote, drew, edited, sang or just created the content they use.
Simply being forced to respect attribution and licenses would kill them. Will that ruling ever happen? Maybe not. Should it? I think so.
**Must** happen!!
There are plenty of English corpora that are either publicly available or for which the license holders would likely be happy to partner for limited use cases. This is not some doomsday scenario for ML, it's just a doomsday scenario for Big AI.
@thelinuxEXP @lepapierblanc wouldn't large tech companies just amend their standard terms of service to permit machine learning on uploaded material? Like what many tech companies have already done?
I'm sure it would hinder AI development, but all of FAANG seem very interested in this tech already - they'll be hurt, but anyone attempting to create a non-commercial model or compete with these huge firms would be killed.
They could use nearly everything that is older than 100 years (and therefore in the public domain). I think the resulting chatbots would really be interesting.
Big companies when they see someone using their 57 years old 2 second long sound effect: GO TO JAIL
Big companies stealing every bit of creative content from the internet without permission from the small creators: 
@thelinuxEXP Ironically, US copyright law punish people for making derivatives, but somehow AI companies are exempted.
"First, the derivative work has protection under the copyright of the original work. Copyright protection for the owner of the original copyright extends to derivative works. This means that the copyright owner of the original work also owns the rights to derivative works." - LegalZoom (22 MAR 2023)
@sysop408 These companies are mainly US-based, and I would argue the US is the biggest repository of works they use, so this would put a stop to most efforts.
I would also love to see rulings in other areas of the world, though. I live in the EU, and I would be very happy to see the European Commission making it illegal to use EU produced content to train AIs without licensing rights.
China is very active in AI. They're training video generators on all the TikTok content, and they've released large language models.
China has a huge incentive to develop AI in part because of it's demographics: male surpluss, and elderly surpluss.
Japan, Korea are in a similar boat
@thelinuxEXP To play the devil's advocate a bit here, but people also learn in a similar way. You have to read to learn how to write. You have to listen to music to learn how to make your own, etc.
I think there are at least 2 main differences. The first one is that a human can only produce so much work on their own, while AI can mass produce.
@thelinuxEXP The other one how derivative the work is. This is hard to tell. A lot of the work that humans produce is derivative too. It's just that we don't normally publish most of it.
With AI somebody can craft an elaborate prompt to make the AI generate very derivative work and then publish it or claim that it's bad.
I'm not on the side of big tech here, but I want to point out that the question is more complex and more nuanced than just "copying is bad".
@yukijoou @thelinuxEXP Creativity is hard to define and even harder to measure. I don't think it's suitable to base policy on that.
AI is one of the most dynamic fields right now. Things change in months. Basing policy on its current status (e.g. how it learns, or how it works) is also pointless.
My current thinking is that we should focus on what it produces and whether that is original, rather than on how it was trained.
@TapiocaPearl @thelinuxEXP I must be living in an alternative universe. AFAICT this is a big part of how people learn new skills.
I agree about the energy, but that's another topic. Also the amount of data that people need to learn something is generally smaller than what AI needs. And the process of learning is similar, but not the same. But this doesn't invalidate my points.
@mayadev @thelinuxEXP I think this is a bit narrow view. Memorization is certainly part of learning even for humans. AI is doing more than just that though. If that was all it needed to do, just saving the data would be enough and computers can do that in 1 go.
For example there's this task: develop a program that can recognize hand-written digits. It's practically impossible to write a good one in the old-fashioned way with rules and stuff. With ML it's trivial. It's the "hello world" of AI.
@mayadev @thelinuxEXP The learning is similar not the same. Both:
- need positive and negative examples
- need external or internal feedback
- extract patterns so they can apply the knowledge to new situations.
The main differences for me are:
- AI needs many more examples
- teaching AI is expensive. to make it acceptable AI is trained "in the lab". The trained models are then used, but they don't learn new stuff while they are used.
I'm sure experts are working on both.
@mayadev @thelinuxEXP I'd consider this "learning". You may disagree of course. But please answer then what is learning?
Creativity is another interesting topic. What if a painter just "drip" paint on the canvas? Would that be creative or just random?
True, AI still doesn't learn like a human. And maybe it never will. But you can't make policy based on its current state. Things change too fast for that.
@thelinuxEXP while I'd love to see such a ruling happen, it might be detrimental in the long run. It would just give nation like China or Russia, who don't care about licence or intellectual property a massive advantage in this field.
Although, as I musician myself who has a lot of friends in art, I think generative AI must be regulated but not destroyed
@osi @thelinuxEXP yes. But as I understand, AI training doesn't uses materials with paid/limited access – in this case there wouldn't be such a discussion that Nick started – it takes only things that can be accessed without barriers. And I think, that if some information is just lying in the web without paywalls or other access restrictions, it can be used freely as basis to create new information.
For me, the first and main sin – let's call it so – is taking some others' materials and then claiming it, without any modification, as your creation. But if you saw some art online and then drew similar – there's nothing wrong with it.
@thelinuxEXP I don't think it would...
In fact, it could make things worse...because those with power and money would be able to acquire (and perhaps even force exclusivity agreements) with those who own the data. Unfortunately, data ownership has become significantly centralized.
They would certainly lose access to some data, but ultimately I don't think it would stop these companies, it would simply provide them another way to limit competition.
@thelinuxEXP @vartak That’s definitely not the rule, Nick. If it’s in public, it’s legal to photograph and the photo belongs to whomever took it.
Barbra Streisand learned that rule the hard way.
@thelinuxEXP @bouncing @vartak In the US, you're generallyy free to take and publish pictures of anything visible in public including people and buildings, and you own the copyright to those photos. Where things get complicated is when you want to use those photos commercially. For example, you can't go around taking photos of people's faces and sell them as stock photography without their permission.
More/better info: https://jmpeltier.com/photographing-people-in-public-legal-ethical-considerations/
@thelinuxEXP @vartak Looks like that thing about the Eiffel Tower is only at night, only in France, and completely untested in court: https://www.snopes.com/fact-check/photographs-of-eiffel-tower-at-night/
See also, from @jimvernon’s link, https://en.wikipedia.org/wiki/Nussenzweig_v._DiCorcia
I know there are some stronger privacy protections in some countries in Europe, though. Eg, it isn’t always legal in the EU to photograph someone’s domicile, IIRC.
@thelinuxEXP
I feel like it's getting too late at this point. Many companies have started adding weird clauses where if you post anything on their website, they own all intellectual property to that content.
So while it be a bit more expensive, the AI companies will still get your data to by licensing companies for the data (this will still be cheaper than fair compensation). Of course the added expense will simply be passed on to the consumer and all blame will be placed on the regulations.
@thelinuxEXP This guy had a nice drawing about that
I am in complete agreement with this
Nick @ The Linux Experiment
visone
Ben Mezger
Duco
Le Papier Blanc
Nasado
Michael Bacon
Emmanuelle Germond
ಚಿರಾಗ್ 🌹✊🏾Ⓥ🌱🇵🇸 (he/him)
Carlos Solís
Binks
C.Suthorn 
Gabriel Pettier
Mahbub Hasan
Sheldon
walter4096
ivt
yuki - queen of the snow
Boba
maya
Lhyr
Gary "grim" Kramlich
ChibyX 
peter royal
Daniel Brotherston
Nachiket Vartak
Ken Kinder 
Jim Vernon
Ⓐvebury Rosetta 
🍉🏴🚩
Diego Pino
Itchi5 illustrations
Xucaen ☮️💚🌿