😵 Meta's internal AI training debates EXPOSED

 

 

 

 

A global community of 450,000 members mastering AI. Subscribe | Sponsor Hello RSS! How are you? I am currently in Bangalore for an exciting project and I am excited for you to see it soon. But until it’s released, let’s focus on today' s newsletter. Here is what's in store for this edition: Google’s AI solves a 10-year-old problem in 48 hours. Meta's internal AI training debates exposed. A comprehensive roundup of last week's AI news. Let’s go! < ![endif]--> 🔦 Last Week In AI🔦 Google’s AI Solves Decade-Old Problem in 48 Hours Google's "Co-Scientist" AI tool has demonstrated remarkable capabilities by solving a complex microbiology puzzle in 48 hours that took scientists at Imperial College London a decade to unravel, marking a significant milestone in AI-assisted scientific research. Technical Context: Successfully identified how superbugs develop antibiotic resistance through viral-tail borrowing mechanisms. Generated multiple viable hypotheses, including one novel approach not previously considered by researchers. Reached conclusions without access to unpublished research data. Demonstrated ability to replicate complex scientific reasoning independently. Why It Matters: This breakthrough demonstrates AI's potential to dramatically accelerate scientific discovery, serving not just as a validation tool but as an innovative research partner. As Professor Penadés notes, this could fundamentally change how scientific research is conducted, compressing years of investigation into days. Meta's Copyright Controversy Newly unsealed court documents reveal Meta employees discussed using copyrighted materials, including books from questionable sources like Libgen, to train their AI models, highlighting the complex intersection of AI development and intellectual property rights. Technical Context: Internal discussions focused on training AI models using copyrighted books without explicit permissions. Considered using Libgen, a controversial platform known for copyright infringement. Implemented systems to remove clearly marked pirated content and avoid IP-risky prompts. Explored alternatives to first-party training data (Facebook, Instagram, Meta for Business). Discussed overriding previous restrictions on content sources to remain competitive. Why It Matters: This revelation exposes the tension between AI advancement and copyright protection, raising critical questions about intellectual property rights in AI development. The case could establish important precedents for future AI training practices. Source: TechCrunch     News📰 1. Unique raises $30 million for AI in finance 💰 [Link] 2. iPhone 16e offers Apple Intelligence at affordable price 💡 [Link] 3. Nvidia CEO confident despite DeepSeek concerns 📈 [Link] 4. Inception introduces new diffusion-based AI model 🚀 [Link] 5. ElevenLabs launches Scribe: a new speech-to-text model 🎤 [Link] 6. Alexa+ can read and recall lengthy documents easily 📚 [Link] 7. Amazon launches Alexa+ with new features for kids 🎉 [Link] 8. Tech giants battle for leadership in AI chips 💻 [Link] 9. Phi-4-multimodal boosts AI capabilities significantly 🌟 [Link] 10. Gemini Code Assist offers 180,000 code completions monthly 📜 [Link] Get Latest AI News Daily Join The Fastest Growing AI Community In The World Join our exclusive Future & AI community to learn & network with the Top 1% AI leaders & early adopters.  Never miss out on a crazy AI hack, a powerful AI tool, or weekly AI workshops! Oh, and we NEVER SPAM!  Click to Join The Largest AI Community I’ll ensure you stay ahead in your life and career by leveraging AI.  It’s a promise! 🌟 Your AI Companion & Guide, Nivedan! I'm Nivedan Rathi, an IIT Bombay alumnus and ex-founding member of several tech & AI startups. I started Future & AI with LLA - India’s #1 Finance Influencer (6M Subs). I’m on a mission to help 1 Million people master the future with AI, and not be afraid of it. Learn more here..🙂⚡ Unsubscribe