Google launches three exploratory models, OpenAI seeks billions yet again
what to know for now
🧠 Google unveils trio of experimental AI models, enhancing capabilities. Gemini 1.5 Flash 8B excels at multimodal tasks and long-context summarization. Enhanced Gemini 1.5 Pro outperforms predecessors in complex prompts and coding. Improved Gemini 1.5 Flash boosts speed and efficiency across benchmarks. Read more
🏛️ California advances AI safety regulations for large models. Bill requires companies to test and disclose safety protocols for AI systems. Legislation faces tech industry opposition, while supporters argue it's crucial to prevent misuse of powerful AI. Read more
⚡ xAI's Grok-2 models double speed after inference stack rewrite. Developers used SGLang to optimize Grok-2 and Grok-2 mini, improving speed and accuracy. Grok-2 now ranks second on Lmsys Chatbot Arena leaderboard, tied with Google's Gemini-1.5 Pro, excelling in mathematical tasks. Read more
🔬 AI identifies cellular heterogeneity using nanoscale nuclear features. AINU, a deep learning method, analyzes super-resolution microscopy images to detect subtle nuclear changes. It distinguishes cell types, identifies cancer cells, and detects early-stage viral infections by examining chromatin structure and protein distributions. Read more
🎓 Yale commits $150 million to AI research and education. Investment supports GPU infrastructure, secure AI tools, faculty hires, and curriculum development. Yale aims to lead in AI innovation while addressing ethical and social challenges. Read more
🧪 AI Research of the Week
Writing in the Margins: Better Inference Pattern for Long Context Retrieval
From Writer, Inc.Jake’s Take: Large Language Models (LLMs) struggle with processing long texts, but the technique in this paper called "Writing in the Margins" (WiM) improves their performance by up to 30% on certain tasks. WiM works by having the LLM generate relevant notes for each chunk of text it processes, then using those notes to inform its final answer.
Techniques like this could help make LLMs much more effective at handling long documents and complex reasoning tasks, potentially expanding their usefulness in fields like research, analysis, and decision-making that involve processing large amounts of information.
what to know for later
🍎 Apple in talks to invest billions in OpenAI. Potential investment would value ChatGPT creator above $100 billion. Move solidifies Apple's AI strategy, leveraging OpenAI's technology for enhanced Siri and iOS features. Read more
🤝 NIST partners with Anthropic and OpenAI for AI safety research. The U.S. AI Safety Institute will access new AI models pre-release, collaborating on evaluation methods and risk mitigation strategies. This initiative aims to advance measurement science and foster responsible AI development, aligning with Biden's executive order. Read more
💰 OpenAI seeks billions, potentially valuing company over $100 billion. ChatGPT creator negotiates funding round led by Thrive Capital, with Microsoft expected to invest. OpenAI's valuation skyrockets, driven by ChatGPT's popularity and AI market momentum. Read more
👂 Google trains AI to diagnose diseases through sound signals. Partnering with Salcit Technologies, Google's model analyzes coughs and breathing patterns. This smartphone-based tech aims to improve healthcare access in underserved areas, expanding Google's efforts in digitizing human senses. Read more