Liquid Legal Institute e.V.’s Post

AI memorization, global status, and deepfakes Last Friday, Therkel Sand Therkelsen and I delivered a presentation on AI at the Nordic CMO Meeting 2024. Not a day without an AI discussion! We covered many of the most central questions about AI. Some of the key points: 1) Memorization in LLMs  We discussed how Large Language Models (LLMs) can sometimes store popular material within their systems, allowing for precise reproduction. This is important when considering the copyright implications of these systems. Feel free to comment below if you’d like a copy of the technical reports on this topic, including those led by Danish researchers. 2) Global status on AI output protection  We examined the current global status on protecting AI output. Decisions from various jurisdictions are starting to reveal whether and how AI outputs can be protected. If you want an overview of these rulings, just drop a message, and I'll send it over. Anders Kristian Rasch pointed out that it might make sense in the future to have a distinct (less robust) protection for AI outputs, similar to the protection known for photographs, where outputs often arise with minimal creative input. 3) Deepfakes  We showed how alarmingly easy it is to create deepfakes, illustrated with a deepfake of myself (see below). Using deepfakes in advertisements, films, etc., without consent is generally not allowed. However, the rules are less clear concerning deceased individuals. We discussed how long one must be deceased to be considered public domain, referencing cases involving Einstein, Marilyn Monroe, and others. These were just a few highlights from our session. A special thanks to Copydan Tekst & Node for the invitation and to Therkel Sand Therkelsen for his usual sharp focus on the latest AI developments. If you have any questions or need more information, feel free to ask in the comments. #AI #IP #copyright

🤖📚 Bluey is a cute little #AI model that can help you put your kids to sleep with some amazing bedtime stories. But, if you’re planning to use #AIgenerated content for your business, make sure you’re not infringing on anyone’s copyright. Bluey can generate children’s bedtime stories that are both entertaining and original. The creators of Bluey used #ChatGPT GPT-3, an AI language model, to train Bluey. Unfortunately, there are potential legal issues surrounding the use of AI-generated content, particularly in the context of #copyright #law. ❌ "copyright can be infringed if a character is replicated in another context in a way that reproduces enough of these aspects" ❌ "character-specific GPTs create more #trademark problems than #copyright ones." So, if you’re planning to use AI-generated content for your business, it’s important to be aware of the legal implications. You also need to build and test many #ethical #safteguards when letting AI interact directly with children. Read the full article on WIRED: https://zurl.co/zzRm ------------------------------------------------- *Follow me for the latest Data Analytics & AI News as I run the AI R&D Lab Kavi Global , the Leading Data Analytics & AI Firm in the USA.

Copyright and AI, amiright? I'll be watching this case closely. This is one of the biggest concerns with the use of AI in creating content. It's not original content. Someone somewhere wrote that. This recent NPR article does a great job of explaining everything. TLDR: The Times is taking OpenAI to court over the question of whether artificial intelligence companies broke intellectual property law by training AI models with copyrighted material. #ai #contentstrategy #copyright https://lnkd.in/gRcSartN

__Daily AI Posts__ Dec 27, 2023 Three Trends in AI Regulation in 2023 — AI : The Washington Report | Mintz The year 2023 has been a significant one for artificial intelligence and AI policy. With the emergence of powerful and easily accessible generative AI ... https://lnkd.in/g4vCCfFk Conor Sen on X: "So how exactly are LLM's going to work when all the content providers block ... So how exactly are LLM's going to work when all the content providers block access to their content without permission? Or will this be a ... https://lnkd.in/gcjenw7E Online Sellers Use AI Images to Sell Products That Look Totally Different | PetaPixel researching how artificial intelligence generates images. Needless to say, synthetic images of this type are improving with Midjourney v6 the ... https://lnkd.in/gFzn3uHB New York Times sues OpenAI , Microsoft for copyright infringement - Spectrum News The New York Times announced Wednesday that it has sued OpenAI and Microsoft for copyright infringement. The suit alleges the companies used ... https://lnkd.in/g7EEdzJu Happy reading! #ai #ml #ainews #llm #genai

What would Shakespeare think if AI rewrote Hamlet?? The rise of Gen AI has raised significant ethical concerns, particularly about copyright and authorship. Gen AI models are trained on large datasets, which frequently contain copyrighted material. When these models create new content, they may unintentionally replicate or derive works that belong to others, raising concerns about intellectual property rights. The ethical quandary revolves around proper attribution of original creators whose work influenced the AI's outputs. Managing copyrights in the age of Gen AI requires the establishment of clear frameworks that distinguish between inspiration and reproduction. Even when AI systems use their creations, creators still need to have control over how their work is used. To guarantee that the original creators of datasets used to train AI models are paid for their contributions, one way to address this issue could be to include licensing agreements. This would encourage ethical use of creative works while also promoting innovation. When AI is used in the creative process, authorship also becomes more difficult. When an AI creates a work of art or text, who gets credit? The developer, the user or the model itself? To address this, clear distinctions must be made between AI-assisted creations and those produced solely by humans. Transparency about AI’s role in content creation is essential to maintaining integrity in authorship and protecting creative rights. #Openai #AIInnovation #TechPreview #AITrends #AICommunity #MachineLearning #AI #FutureofAI #AI2024 #AIResearch

Siri, transform me into a futuristic cyberpunk cartoon of a Marvel avenger superhero, dissolving into catwoman, dissolving into a Bratz doll 🫠🦸🏼♀️ Learn how to do this in the Ethical AI Workshop 😍👇 1️⃣ Ethical AI v Generative AI: What are they, and how do we apply them to our own business? - Essential definitions. - Text generation, GPT, and beyond. - Prompt Engineering 101. - WHAT IS LEGAL COPYRIGHT WITH AI CONTENT? - EU AI ACT - US Copyright Law 2️⃣ Image Generation AI - Models and breakdown of DALL-E, Midjourney and other competitors. - Pricing, benefits, and quality comparison of all image generative AI models. - Examples and case studies by Coach Megan. - Tutorials. 3️⃣ Video Generation AI - How to translate images to videos to maximise your time and increase your profitability. - Best Video Gen AI softwares. 4️⃣ Audio & Music Generation - How to create your own music legally with AI. - How to clone your voice and replicate it. - Case studies & softwares you can use to maximise your time and increase your efficiency in your business. 5️⃣ Sales & Lead Gen AI - Tools to maximise your lead quality. - Tips on how to humanise your ad copy. - How to personalise your ad copy for more sales. ✨️ BONUS ✨️ → Ethical AI policy templates → Employee master list guidelines → Extra courses you can optionally take to further your knowledge. Final Exam 75% passing grade. __________________________________________ EVMO News Premium Education prides itself in bringing back meritocracy in education, delivering the highest quality knowledge to the highest quality CEOs who do the work, implement the work, and get the results. Coach Megan is a 4x certified Gen AI expert, 2x MSc in Marketing and Innovation, 1x summa cum laude graduate in Political Science with course in International Law 🎓 In 2011, the 44th President of the USA awarded Coach Megan with the Presidential Award for Academic Excellence. You can trust Coach Megan with your educational excellence, too. __________________________________________ → Enroll today by messaging "AI" 📨 #ethicalai #aiforbusiness #generativeai #artificialintelligenceforbusiness #premiumbusiness

Tension between Generative AI and Media   - Tension has rose between transformative potential of generative AI and the concerns of media companies and individual content creators (E.g. journalist) regarding their rights, fair compensation, and the impact on their brand value. - Some rights owners are becoming more cautious about embracing generative AI, with some opting for legal action instead of collaboration to protect their rights and content, and seek fair compensation. For examples,   🔘 Getty Images sued Stability AI over use of its picture library. 🔘 Music publishers Universal Music, ABKCO and Concord sued artificial intelligence (AI) company Anthropic over use of copyrighted song lyrics. 🔘 New York Times lawsuit against OpenAI and Microsoft over use of copyrighted work.   - The legal landscape for generative AI draws parallels to early search engine cases, where courts ruled "fair use" for indexing copyrighted content, but key differences exist as AI-generated content poses greater legal risks as it directly answers queries, potentially substituting original sources.   - Legal risks prompt caution for AI companies defending fair use before a jury.   - Media companies can block AI from using their content but risk losing value due to AI's commodification of information.   - Negotiations likely to resume as generative AI offers new markets for media content.   - Media companies aim to benefit directly but risk ceding audience and brand value to AI platforms. - Uncertainty about AI's future services complicates negotiations for media companies. - Growing use of generative AI will heighten pressure for a resolution between media and tech industries. #aicommunity #ai #aichallenges #aiandbusiness #aiadoption #ai

The dawn of AI has made most of us dream of (or maybe dread) what the future might look like. As enticing as the benefits undoubtedly are though, are we so enamoured with the possibilities of Generative AI that we turn a blind eye to how they grow and learn? This was a great opportunity to work with Martin Soames and Maninder Gill on this article. (No AI was used in this post) #ai #copyright https://lnkd.in/em4bMnwk

Things We Don’t Need: Generative AI - The Illusion of More: The text discusses the author's contemplation of a blog category called "We Don’t Need This" focusing on non-innovative innovations. It addresses concerns about generative AI displacing human creativity and criticizes the FTC's approach to regulating AI's impact in creative industries. It also highlights tensions between creators and technology companies. The summary captures the essence of the text's critique of generative AI and its potential impact on creative industries. https://lnkd.in/dt6QbbDP - IP topics: Intellectual Property topics! #ip #intellectualproperty #copyright

Spawning AI, a company focused on empowering artists in the digital age, is tackling the controversial topic of AI training data with its new #project, Source. Plus. This platform strives to provide a curated collection of "non-infringing" media specifically for training generative AI models. ⬧ Source. Plus: A High-Quality, Ethically Sourced Dataset Source.Plus kicks things off with a nearly 40 million-image dataset built from public domain and Creative Commons CC0 licensed works. While smaller than some existing options, Spawning AI argues that their focus on quality over quantity ensures the data is sufficient to train state-of-the-art image-generating models. This platform functions as a universal "opt-in" system. #Creators can contribute their media to AI training on their own terms, while developers can seamlessly integrate this media into their workflows. ⬧ Addressing Ethical Concerns in AI Training The debate surrounding the ethics of training generative AI models, especially those for art creation, is ongoing. These models learn by ingesting massive amounts of data, and some developers believe "fair use" allows them to scrape data from public sources regardless of copyright. Spawning AI CEO, Jordan Meyer, highlights the lack of consensus on a best #approach. He emphasizes the current challenges: ⬪ Limited artist control over how their data is used for training. ⬪ Difficulty for developers to find high-quality, ethically sourced alternatives. Source. Plus builds upon Spawning AI's existing art provenance and rights management tools to address these issues. ⬧ Source. Plus: More Than Just a Data #Repository Source. Plus goes beyond simply offering a dataset. It provides an "enrichment platform" with tools to streamline the training pipeline. Their goal: to establish a high-quality, non-infringing CC0 dataset capable of supporting a powerful base AI model within a year. While some organizations claim to use ethically sourced data, Spawning #AI aims to set a higher bar with Source. Plus. Here's how: ⬪ Opt-out filtering: Ensures images respect artist training preferences. ⬪ Provenance information: Shows how and where images were sourced. ⬪ CC0 focus: Excludes non-CC0 licensed images, including those requiring attribution. ⬪ Copyright monitoring: Addresses potential copyright #challenges. www.tycoonstory.com #SpawningAI #SourcePlus #AITrainingData #GenerativeAI #CreativeCommons #PublicDomain #EthicalAI #AIArt #ArtistRights #DataProvenance #CreativeEconomy #AIForArtists #TechForGood #AIInnovation #OpenSourceAI #FairUse #AIethics #ArtistCompensation #AIandArt #DataRights #AITraining #ArtCommunity #AIDataManagement #EthicalAITraining #AIArtCommunity #DigitalArt #ProvenanceTechnology #FutureOfAI #AIandCreativity