Open Future Foundation’s Post

Open Futures has great overview about the current status of opting out of machine learning and generative AI indexing: https://lnkd.in/eKVnyxxz

As the GenAI craze has revealed a supply chain of IP providers that make the LLM models valuable. In order to give credit and profit where due in the GenAI supply chain, U.S. Representative / Congressman Schiff is introducing a bill titled "‘Generative AI Copy-right Disclosure Act of 2024’’. A bill is the start of how laws are created in the U.S. This bill is not only timely for the GenAI supply chain, but is also timely for Schiff who is running against Garvey to be a U.S. Senator.  https://lnkd.in/gFXN97Qz

"Because copyright today covers virtually every sort of human expression—including blogposts, photographs, forum posts, scraps of software code, and government documents—it would be impossible to train today’s leading AI models without using copyrighted materials," wrote OpenAI in the House of Lords submission. This must be clear to all, and we wonder again if this is the kind of progress we need in today’s world, where information is already free and accessible everywhere to be consumed. Copyrighted material comes from the creative process of bright talents and minds who rely on it for their survival and development. Allowing the mass to use #generatieveai within these terms, in this economical system, is simply regressive. We are not talking about the potential benefits in science, research, healthcare, defense and all sectors where we could see real advancements. It looks just as a global act of bullying from tech companies, besides the fact it mines the foundation of private property, the base of our democracies. Read more: https://lnkd.in/e8KYKzbk #aicommunity #ainews #europeanunion #europeancommission #euaiact #airegulation #aiadvancements #aiartwork #euronews #telegraph

Excerpt from my research paper, "Does US Intellectual Property Law  Protect Creators’ Rights in the Age of Machine Learning? It depends on what you think Creator’s Rights are." "The backlash came swiftly on the heels of the release of image generators like DALL-E, and text generators like ChatGPT: Generative AI is a legal problem, to be decided by the courts. (Wiggers). The US Copyright Office was quick to rule that they, “will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.” (US Copyright Office) A vital open question remains, however: US Copyright law was intended to encourage humans to share ideas, so that we might all benefit from them together, to enhance and benefit our commonwealth (US House of Representatives). Deep learning models are trained on corpuses of literature and archives of digital photos and art, data-mined from the public internet (Quang 1407). Do the same laws apply to machines when they learn, even artificially? And how?" https://lnkd.in/gjaWUdVv

You probably aren't thinking enough about the biggest risk to #AI - #copyright law. Growth of #GenerativeAI has been explosive - it's been just over a year since OpenAI released GPT-3, yet now we are seeing ultra capable multi-modal LLMs and highly realistic text-to-video. It seems unstoppable - but I believe it is almost inevitable that 2024 will slam on the brakes - or at least provide a big speed bump. We haven't solved the fundamental question of copyright on training data. LLMs require huge amounts of training data to be useful. In general, the more training data, the better the model is. If we restrict the corpus of data the model is trained we'll have worse models. LLM companies have been steadfast in saying that their models transform the underlying data, so they qualify under 'fair use' exemptions. But this argument rings hollow as we increasingly rely on AI output and neglect the source material. Imagine you write a successful blog and suddenly no-one but AI crawler bots visits your blog again. Or you are an artist and no-one pays you any more because AI trained on your pictures can generate a perfect image in precisely your style? 2024 is almost certainly the year where these arguments will explode to the forefront and dominate discussion. Let's be clear - #law and #regulation aren't clear on this new use case. Even the #EUAIAct doesn't help us - it states that: 'Any use of copyright protected content requires the authorization of the rightholder concerned unless relevant copyright exceptions and limitations apply.' And what are 'relevant copyright exceptions'? Who knows. There are good arguments on both sides. This is a new area, and we need new law to match. The worst case scenario is if we see a balkanisation of models and fragmentation of capability depending on where in the world you happen to be. What do you think is going to happen? Let me know in the comments below. Coert Du Plessis (杜康) Data Alchemists

POTUS Issues Expansive Executive Order on AI, Calling for All Hands On Deck | ArentFox Schiff - JDSupra: President Biden issued an executive order on October 30, 2023, to regulate the development and use of artificial intelligence (AI). The order aims to establish guidelines, best practices, and policies to ensure the safe and secure management of AI, while also promoting innovation, supporting workers, and upholding civil rights. (50 words) - IP topics: Intellectual Property topics! #ip #intellectualproperty #copyright

The advancement of generative AI models has enabled replication of trademarked characters with minimal prompts. This highlights the models ability to extrapolate and generate content based on limited input, raising concerns about potential misuse in intellectual property and copyright domains. As these capabilities progress, it becomes crucial to address ethical considerations, legal implications, and establish guidelines for responsible use in creative and commercial contexts.

AI's legal considerations are as complex as they are critical. From copyright to compliance, learn the essentials for navigating this dynamic field. Read more in the link below on leveraging legal insights for AI innovation.

Check out Tapestry VC's latest perspectives on the legal questions that arise from this next phase of AI development. We've constructed five "sci-fi hypotheticals" to engage with questions regarding liability and responsibility of AI implementation. https://lnkd.in/eRSKES8B

This article is about the downside of 'code as law' in our present version of #aisociety, where once something is out, it's out everywhere The #dataeconomics here drives value close to zero, showing why security has to come first, a necessary (but not sufficient) condition for there to be a data asset There remains the issue of preventing an algorithmic domination of publicly consumed news: the underlying truth value of x nations reportage has a public good element This is the notion of a long term need in the data commons. A return to physical information (#books), libraries and professional #history will help "... how courts or even governments decide to regulate the use of copyrighted material to train AI systems may not be the final word on the matter. A number of AI models, both text-generating LLMs and image generators, have been released “open source”, free to download, share and re-use without any oversight. A bar on using copyrighted material to train new systems will not scrub those from the internet, and will do little to prevent individuals from using new material to retrain, improve and re-release them in the future.