2024-05-09 Hoping for the next AI winter ======================================== @taichara@chirp.enworld.org pointed me to a blog post that Creative Commons (CC) wrote in 2023 about the court cases against GitHub, Stable Diffusion, Midjourney and the claim that their use of copyrighted works to train the machines is copyright infringement. CC does not agree. > As Creative Commons has argued elsewhere, and others agree, I > believe that this type of use should be protected by copyright’s > fair use doctrine. In this blog post I will discuss what fair use > is, what purpose it serves, and why I believe that using copyrighted > works to train generative AI models should be permitted under this > law. – Fair Use: Training Generative AI I guess I don't mind that blog post. For one, Fair Use is a very USA thing and I don't live there. Furthermore, the CC licenses don't say "this material is fine for AI training" – so whether this is fine or not will be decided by the courts. I would only worry if CC turns out to be working on a version 5 of their licenses with an explicit exception for AI training. And even then, we can keep on using version 4 of the license. There is no automatic upgrade option. This is why I feel fine. Imagine the Dude from The Big Lebowski saying, "Yeah, well, that's just, like, their opinion…" It doesn't look as if CC will be changing the license. The courts still get to decide. If the courts decide in favour of free-for-all AI training, then rich people win for the moment. But even then, we can still nationalise their businesses, or prohibit AI training, or raise prices on electricity and water. The fight isn't over. Not by a long shot. Maybe this is a thing with lawyers. We trust people who share our values and we’d trust somebody when they write something along the lines of “AI training is terrible and AI use is dubious but what can you do, it’s in the hand of the courts, we certainly don’t support it.” But lawyers trust other lawyers who say “we’ll see what the courts say but our understanding is that nobody should stop AI training because that’s the trade-off copyright makes and we support people and corporations exercising their rights.” The apparent moral support on copyright grounds is revolting, but this approach seems typical to me. They say: The point of the law is this or that. We say: The result of these actions is bad for society and bad for the climate. They are externalising all the costs. The artists and authors who want to produce new works are losing their jobs. The people double-checking the computers and their training are burning out. The communities where these corporations build their computing centres suffer from water scarcity because the water is used to cool the computers. The countries where these corporations build their computing centres need to build more power plants because so much electricity is required to train their models. The move to renewable energies is delayed. Phasing out nuclear power is delayed and more radioactive waste ends up on our lands. > Sending this message was important to us. We considered ourselves to > be a powerful culture. This place is not a place of honor... no > highly esteemed deed is commemorated here... nothing valued is here. > – Long-term nuclear waste warning messages, Wikipedia Or we can believe in nuclear fusion: > AI models made up of billions of parameters require huge amounts of > energy to train. OpenAI's old GPT-3 system reportedly consumed 936 > megawatt hours … the average household consumes about 10.5 MWh per > year. That means training GPT-3 consumed as much energy as about 90 > households consume in a year. – Energy breakthrough needed to build > AGI, says OpenAI boss Altman, The Register Sadly, nuclear fusion only seems to be working inside the sun: > 1955: At the first Atoms for Peace meeting in Geneva, Homi J. Bhabha > predicts that fusion will be in commercial use within two decades. – > Timeline of nuclear fusion, Wikipedia And what for? The people who are looking for answers get served bullshit slop that's wrong and hurtful and racist and misogynist. We see this in customer service degrading, in search engines degrading, in websites degrading. The corporations don't mind that they're feeding us slop as long as they are making money. We are enabling this because these corporations are powerful and rich. They buy our politicians with money and promises of jobs, they don't pay the taxes they ought to pay and then they don't deliver on the jobs, and the jobs they offer are menial and subservient when we'd prefer to be creative and free. We thought we'd be writing and painting all day while the machines serve us but we end up serving them while they do the writing and painting, badly! It's hard to defend against these corporations because the knowledge required to defeat them is not known to us all. We don't read about it in the media. Instead, they volunteer for all the posts, offer their experts, pay for their own research, write their own glowing press releases and commission promising reports. This is how regulatory capture works. So yes, we should nationalise them, regulate them, force them to pay for the true cost of resources they use, outlaw their energy waste, force them to label their output, shun them. We should kick them from our company. Like a creeping disease their rot spreads, destroying the technology we built, the search engines, databases, websites, turning it all into AI slop. It is revolting. And maybe copyright is not the right tool to stop corporations from training their machines on our works. But in the big picture, they are still wrong. @eloquence@social.coop says: > The folks who are calling such training "theft" might regret what > they seem to be implicitly asking for, i.e. much stricter copyright. > Copyright law won't prevent Microsoft, Google, OpenAI or Adobe from > making shady licensing deals, but they'll prevent the free/open > community from keeping up. It's true, I don't want to ask for stricter copyright. But I still oppose all the things that are wrong with the current AI bubble. @mcc@mastodon.social recently wrote about a similar thing: We use free software licenses in order to distribute our software for the benefit of our fellow humans but not necessarily to train the machines of corporations that are responsible for all the slop. > Like, heck, how am I supposed to rely on my code getting preserved > after I lose interest, I die, BitBucket deletes every bit of > Mercurial-hosted content it ever hosted, etc? Am I supposed to rely > on Microsoft to responsibly preserve my work? Holy crud no. We want > people to want their code widely mirrored and distributed. That was > the reason for the licenses. That was the social contract. But if > machine learning means the social contract is dead, why would people > want their code mirrored? – mmc We're using copyright to keep the software free. If AI regurgitating our software from its impenetrable memories circumvents copyright it also circumvents our licenses and therefore pulls the claws and teeth free software has to fight for its freedom. This is why opposing the current AI bubble is important. No matter how I look at it, the big picture shows that this kind of AI is our enemy. It's just that copyright is the wrong weapon. This is why we continue fighting all the negatives these corporations bring to the table. And if it turns out that their business model doesn't work unless they bring all these negatives to the table, then I guess that's just that. We are under no obligation to support their business model. In fact, it is our right and our duty to regulate the world in order to improve it for all of us. We've been doing it for a very long time. Here's to hoping for the next AI winter. > In the history of artificial intelligence, an AI winter is a period > of reduced funding and interest in artificial intelligence research. > The field has experienced several hype cycles, followed by > disappointment and criticism, followed by funding cuts, followed by > renewed interest years or even decades later. – AI winter, > Wikipedia #Copyright #Artificial_Intelligence #Artificial_Intelligence 2024-05-16. Cory Doctorow also thinks that copyright is the wrong weapon: > Creative workers who hope that they can use copyright law to prevent > AI from changing the creative labor market are likely to be very > disappointed in court … Individual creative workers rarely have any > bargaining leverage over the corporations that license our > copyrights. … Rather than going to Congress alongside our bosses > from the studios and labels and publishers to demand more copyright, > we could go to Congress alongside every kind of worker, from > fast-food cashiers to publishing assistants to truck drivers to > demand the right to sectoral bargaining. – AI "art" and uncanniness