All the writers of the 7 million books included in 3 whole training datasets can now join a certified US-wide class action against Anthropic.
What I find beautiful with AI and class actions in general is that members don’t need to be US nationals to get their damages. AI training never respected any borders or boundaries, so copyright holders are totally international. So, any writer whose books landed in one of the certified datasets is covered.
The courts have been taking their sweet time with AI-related claims because any final decisions with respect to AI will have immediate and serious consequences on a large tech sector.
Capitalism is a pyramid-like fraud that depends on masses of people working for free (i.e. user-generated content) to generate profit for higher-ups on pyramid (plateforms, TikTok, Youtube, etc) who on their turn generate profit for their investors by inserting ads. To keep feeding the pyramid, users were persuaded that they work for “engagement”, likes and subscribes, that they were “building communities”, making the world a better place by speaking out (see Reddit’s AI by Google!), but somehow all that good work ends up profiting corporations.
Given the push for user-generatedcontent and engagement-chasing, more and more of the content became diluted, less and less and less original, descending into reposts and emojis, and when AI came and ate it all up, it turns out a lot of that stuff is not original but mass-produced copy-pasta to impress the algorithms latest whims… now,AI says, all of this is not good enough for me as a large language model, a lot of it is highly redundant, give me original content, I want organic, not something I’ve already seen (you can’t fool me, I’m AI).
So it turns out that AI doesn’t just want to learn, it wants novel, specific things, it is hungry for quality content and very hungry it is. This is where copyright regimes should step in. There is no unfettered access to copyright protected material written into any copyright act of this world. Fair use doesn’t apply. Quality content must be paid for. Large datasets of quality content must be paid for. Otherwise, it would take another 7 planets with the same populations of unsuspecting writers and creators as ours to give our AI what it needs to “grow”.
Even without the courts, the world is already going through a restructuring of content distribution with ever-increasing opt-outs and limits on the availability of user-generated content, where most people are simply not disseminating original ideas and speech due to uncertain AI frameworks. Most of the accessible content you see is recycled and safe, so by definition unoriginal, an imitation, a copy of an imitation of a copy. A template of a vlog of a postcard, of prepackaged, hell a lot of it IS AI-generated already…
Bottom line is, it appears that 90% of original content being produced at this time is basically being hoarded and likely written into wills because nobody thinks that they stand a chance to AI in this lifetime. If you can’t make a buck from your work in the next 100 years or so, then hopefully your heirs can do better in the 75 years after your death. Years may vary between jurisdictions, somewhere I think it’s 50, or 90. Who cares, you’ll be dead.
In the Anthropic class action, regardless of the result, here is imho what to expect on the long run: a total breakdown of existing copyright regimes AND/OR a total breakdown of corporate investment “sentiment”. So, in a way a lose-lose in court will translate into progress for humanity as a whole.
Both outcomes are rife with opportunities. If copyright frameworks are destroyed, then it’s a free for all not just for AI companies, for everyone. I am making lists of all the things I may do if we eliminate economic copyright. Since nobody will be allowed to profit from content, by default everything will become non-commercial therefore more diverse and more interesting because necessarily more independent and capable of using other people’s resources and end-product into yours (it will happen gradually don’t worry). If investors become pessimistic, then they will need to just as gradually rework their value system and see where the wind takes them. Nobody knows exactly yet.
Minor correction. two libraries out of 3 pirated libraries are certified, rather than 3. For people outside of the US, you will obviously need to claim through someone in the US, for example a person or entity involved in the same book, a publisher or collaborator of some kind. Nowhere does it say that a person who is not based in the US can’t sue a company in the US. It would be more convenient for the person to file in their own forum, but you can’t wait forever for something to happen in your forum, so at the moment the only courts that matter are in the US. Europe may seem better on a first sight, but the US is still a more appropriate forum for several reasons, such as the defendants being from the US, and US judges can be counted on to be relatively fast and productive. If you want anything to be resolved in your lifetime, sue in the States.