AIR#125 - July 10, 2024
Good morning, AI aficionados! As you sip your morning coffee, get ready to dive into today's edition of AIR: The AI Recon. Leading the headlines is the dramatic turn in the GitHub Copilot lawsuit, where a judge has dismissed a key DMCA claim. This leaves GitHub, Microsoft, and OpenAI with only two allegations to face, marking a significant moment in the ongoing debate over AI's use of copyrighted material. If you're intrigued by the legal battles shaping the future of AI, this is a must-read.
But that's not all! Meta's MobileLLM is making waves with its optimized sub-billion parameter language models designed for on-device use. Achieving state-of-the-art results at ICML 2024, this development promises to revolutionize mobile AI applications, making them more efficient and accessible. Meanwhile, War and Peas have taken a bold stance against AI collaboration, calling the AI hype a threat to creatives and urging for more artist-supportive tools. This story is sure to resonate with anyone passionate about the intersection of technology and creativity.
And for a touch of the uncanny, the Canadian Centre for Cyber Security has issued a warning about Russia's use of AI "Meliorator" software for foreign influence campaigns. This alarming development underscores the growing role of AI in geopolitical strategies and the urgent need for robust cybersecurity measures. Whether you're here for groundbreaking tech updates, ethical debates, or the latest industry buzz, today's edition is packed with stories that will both intrigue and challenge you. So, sit back, sip your coffee, and let's delve into the dynamic world of artificial intelligence together!
Business
🔥 Why AI Hype is a Slap in the Face for Creatives: War and Peas Respond
War and Peas reject AI collab, calling AI hype a threat to creatives and storytelling. They urge focus on artist-supportive tools.
Russia Uses AI "Meliorator" for Foreign Influence Campaigns
Russia uses AI "Meliorator" to create fake personas for foreign influence campaigns, warns Canadian Centre for Cyber Security.
Humane Execs Launch AI Fact-Checking Startup
Former Humane execs launch Infactory, an AI-driven fact-checking search engine focused on data accuracy for enterprises.
The AI Summer: Benedict Evans on the Hype and Reality of LLMs
Benedict Evans argues LLMs, like ChatGPT, are overhyped and not yet practical products; real adoption will take time and refinement.
OpenAI: "We Can't Make Money Without Free Use of Copyrighted Material"
OpenAI argues it can't profitably train AI without free use of copyrighted material, sparking legal battles with major publishers and authors.
Justice Department Disrupts Russian AI Bot Farm
Justice Department disrupts Russian AI bot farm spreading disinformation in the US, seizing domains and searching 968 social media accounts.
Chinese Developers Scramble as OpenAI Blocks China Access
OpenAI blocks China access, sparking rush for local AI alternatives like SenseTime, Baidu, and Tencent, offering free tokens to lure users.
Turkish Student Arrested for Using AI to Cheat in University Exam
Turkish student arrested for using AI to cheat on university exam with hidden devices; accomplice also detained.
He created Oculus, now makes AI weapons for Ukraine
Oculus creator Palmer Luckey now builds AI weapons for Ukraine, revolutionizing warfare with drones and autonomous systems.
[Paper] Governing in the Age of AI: Transforming the State with a New Model
Governments must embrace AI to modernize, save costs, and improve services, potentially saving the UK up to £40 billion annually.
Engineering
🔥 Judge Dismisses DMCA Claim in GitHub Copilot Lawsuit
Judge dismisses key DMCA claim in GitHub Copilot lawsuit, leaving only two allegations against GitHub, Microsoft, and OpenAI.
[GitHub] LightRAG: SylphAI's PyTorch Library for LLM Applications
SylphAI's LightRAG: A modular, robust PyTorch library for building and optimizing LLM applications. 100% readable codebase.
🔥 [GitHub] MobileLLM: Optimizing Sub-Billion Parameter Language Models for On-Device Use (Meta, ICML 2024)
Meta's MobileLLM optimizes sub-billion parameter language models for on-device use, achieving SoTA results in ICML 2024.
🔥 Dynamic Translation of Smalltalk to WebAssembly
Smalltalk now compiles directly to WebAssembly, optimizing performance for web browsers via Catalyst VM. 🚀
[GitHub] CopilotKit v1.0: Build Custom AI Copilots for React Apps
GitHub's CopilotKit v1.0 lets you build custom AI chatbots and AI-powered text areas for React apps with ease. 🚀
LLM Honeypot with Beelzebub Framework: Revolutionizing Cybersecurity
New LLM Honeypot with Beelzebub framework detects cyber threats without supervision, mimicking real environments for better security insights.
[Paper] Mixture of a Million Experts: PEER Layer for Efficient Transformer Scaling
PEER layer enables efficient transformer scaling with over a million tiny experts, outperforming dense FFWs and coarse-grained MoEs.
[GitHub] Libchatty: Ultra-Fast ChatGPT Wrapper in Pure C
Libchatty: A new ultra-fast ChatGPT wrapper in pure C, boasting speeds 212x faster than competitors. Perfect for AI engineers. 🚀
TerifAI: The AI That Steals Your Voice 😱
TerifAI mimics your voice in a minute, highlighting AI's role in voice phishing. Talk to it and hear your cloned voice! 😱
How to Build Privacy-Protecting AI | Proton
Building privacy-protecting AI involves running LLMs locally or securely server-side to ensure user data remains private and safe.
Fine-Tuning CLIP Models: A Comprehensive Guide
Learn how to fine-tune CLIP models for image classification with our step-by-step guide, boosting accuracy and performance! 🚀
Academic
Claude 3.5 Sonnet Reproduces BIG-Bench Canary String
Claude 3.5 Sonnet can reproduce the BIG-Bench canary string, revealing potential training data leaks. GPT-4o can't replicate it.
AI Revolutionizing Medical Diagnosis
AI is transforming medical diagnosis, identifying patterns in data to detect diseases faster and more accurately than humans.
We Cannot Cede Control of Weapons to Artificial Intelligence
AI weapons must be controlled by humans to prevent catastrophic errors. Urgent global legislation needed to avoid a "Terminator" future.
Can AI Be Superhuman? Flaws in Top Gaming Bot Cast Doubt
Flaws in top Go-playing AI bots reveal vulnerabilities, questioning the reliability and safety of so-called "superhuman" AI systems.