The latter phrasing is used in a published motion (pdf warning) to dismiss the lawsuit filed by the New York Times against Microsoft and OpenAI.
In its initial claim, the Times states that it had been working for "several months" to reach an agreement with the defendants regarding the terms of OpenAI's contribution to the training of a large language model, but is now instead pursuing compensation through the courts.
OpenAI itself places particular emphasis on the NYT articles on which it trained its models, claiming that "by OpenAI's own admission, the high-quality content, including the Times content, was more important and valuable in training the GPT models than content obtained from other low-quality sources. . was more important and valuable." [Microsoft, however, likens the NYT's copyright claim to the entertainment industry's attempts to stave off the rise of VCRs, citing copyright infringement as working against Hollywood. The application states, "The court ultimately rejected these misgivings in favor of innovation and consumer choice."
The motion to dismiss then describes LLMs and the machine learning algorithms (transformers) that have driven the recent explosion of more effective chatbots, and makes claims about the data sets used to train these LLMs. This is the essence of what copyright law considers to be transformative and fair use"
.
The main thrust of Microsoft's motion, however, is not about the training itself, but rather to address the NYT's allegations that public use of LLM, particularly GPT-based products, is harming the publication.
The Times claims that it used unrealistic prompts to make LLM output text that matched the NYT article, stating that it was not really based on how ordinary people, or reasonable people, would use these tools.
"Nowhere does the Times claim that anyone other than its legal team would actually do this," it said, "certainly not on the scale that would merit the doomsday futurism that it is pushing before this court and promoting to its readers."
This could be an important point in this argument. The idea of a 'reasonable person' is an important point in law, and the NYT's argument that a person could be harmed by LLMs could be used to determine whether a reasonable person should be measured by what he or she does with the technology. Similarly, Microsoft and OpenAI might follow what a reasonable person might expect with respect to privacy in an era that is digital and always online.
Because the second trial (pdf warning) is a class action lawsuit that apparently "spends 198 pages on apocalyptic hyperbole about AI being a threat to civilization," 13 plaintiffs have filed a lawsuit against the defendants (OpenAI and Microsoft) for "continued violation of millions of privacy and property rights and . are being sought by the Court to take immediate action to enjoin the continued violation of the privacy and property rights of millions and to put in place appropriate safeguards and regulations for the products, their users, and society as a whole."
The Court's decision is "not a matter for the courts to decide, but for the courts to decide.
Among other things, the original application states that the scraping of the plaintiffs' digital footprint means that the defendants are "exploiting our skills and encouraging our own professional obsolescence. This would erase what we know about our privacy. "
To which Microsoft responds (pdf alert): plaintiffs have not alleged any facts that plausibly show that they were affected by the "scraping," "interception," or "eavesdropping" they claim. Nor can they identify what harm they personally suffered as a result of Microsoft's alleged conduct. Plaintiffs cannot state a claim based on the hypothetical experience of others. This defect alone requires dismissal of all of Plaintiff's claims."
OpenAI has taken a similar stance in its own motion to dismiss, and the plaintiffs in the privacy lawsuit have clearly refuted it, as The Register points out.
"OpenAI never informed the public that for years it has been secretly harvesting from the Internet everything that hundreds of millions of Americans have created and shared online.
"For more than a decade, every consumer's use of the Internet served as a free donation to OpenAI: our insights, our talents, our works of art, our personally identifiable information, our copyrighted works, our family and child photos, and anything else that expressed our humanity, For a product that concentrates the wealth of the country in the hands of a smaller number of giant corporations, takes away jobs on a massive scale, and jeopardizes the future of core industries like art, music, and journalism, while creating dangerous new industries like the high-speed creation of child pornography. "[T]here now appear to be as many lawsuits focused on AI as there are claims against former presidents. You may see these claims and counterclaims as nothing more than a paper forest filled with legalese and fighting big tech, but they will affect all of our lives. For better or worse. They will set a pivotal precedent that will inevitably change the development of both AI and the Internet.
And perhaps, if we stick to over-the-top themes, humanity itself will change as well.
.
Comments