Meta grants Llama model access to U.S. government agencies
Plus New research and tools from NVIDIA
Today’s Highlights:
📰 News: Meta grants Llama model access to U.S. government agencies + New research and tools from NVIDIA
💰 Funding: Physical Intelligence raises $400M+ for its foundational ‘generalist brain’ for robots
⚡️ Top News Stories:
1. Meta now permits U.S. government agencies and contractors to use its Llama AI model for national security tasks, in collaboration with Amazon, Microsoft, IBM, Lockheed Martin, and Oracle.
Meta emphasizes that open-source AI models can enable the U.S. to lead in developing global AI standards based on transparency and ethical use, which could shape the future of AI infrastructure
While Meta’s use policy restricts Llama 3 from military and warfare applications, the AI model is approved for support in areas like logistics, cyber defense, and terrorist financing tracking.
Meta emphasizes the strategic importance of U.S. leadership in open-source AI, particularly amid China’s substantial investment in AI development, as demonstrated by Chinese researchers’ use of an earlier Llama model for military projects.
2. NVIDIA launches an AI-driven Blueprint for video search, enhancing productivity and security across sectors with real-time insights.
The new AI Blueprint enables AI agents to perform video search and summarization, making it easier for industries to analyze and manage vast amounts of visual data in real time.
It uses vision-language models (VLMs) integrated with NVIDIA’s Metropolis tools, allowing visual agents to provide summaries, create alerts, and deliver insights across various environments.
Targeting industries like warehousing, smart cities, and public infrastructure, NVIDIA’s solution aids in security, productivity, and operational efficiency by monitoring and interpreting live video feeds.
3. Microsoft leaders Satya Nadella and Brad Smith, along with a16z partners Marc Andreessen and Ben Horowitz, issued a joint statement urging minimal regulation for AI development, emphasizing "responsible, market-based" approaches instead.
The policy statement contends that AI models need free access to data to "learn like people," challenging copyright laws that restrict free use of proprietary data and intellectual property, a stance critiqued as a means to bypass paying for content.
Rather than preemptive regulation, the group proposes reactive oversight focused on misuse by “bad actors,” avoiding upfront compliance costs that proactive regulation would entail.
Favoring federal rather than state-level regulation, the companies suggest this approach could prevent a patchwork of state laws and advocate for initiatives like digital literacy, open data, and improved government procurement for startups.
4. Researchers from NVIDIA introduced HOVER, a neural whole-body robotic controller, that democratizes advanced robotics with low compute requirements and versatile control.
5. Meta’s FAIR team introduces new AI tools and human-robot benchmarks aim to elevate embodied AI and touch-based robotics for real-world applications, with specific AI advancements in touch perception, dexterity, and human-robot collaboration
6. Nvidia and Quantum Machines employ reinforcement learning for scalable quantum computing improvements, marking a step toward reliable, error-corrected systems.
7. MIT researchers have debuted a versatile robot-training model using LLM techniques to improve adaptability in real-world tasks.
8. Anthropic’s Claude model now has PDF support interpreting text and visuals in large documents, boosting its analysis potential for industries like finance and healthcare.
9. Oasis, a new AI model, generates interactive gaming environments in real-time, promising a shift in how virtual worlds are built.
10. Google’s Gemini API now supports ‘Grounding with Google Search’, providing responses with real-time data and links, enhancing accuracy and transparency.
11. Amazon Prime Video’s new “X-Ray Recaps”, powered by Amazon Bedrock, provides viewers with AI-generated summaries of entire seasons, single episodes, or even specific scenes, letting users quickly catch up on key plot points without interrupting the streaming experience.
12. Disney launched the Office of Technology Enablement (OTE) to guide AI and mixed reality adoption across all divisions in a responsible and innovative way.
13. Amazon’s AI-powered Alexa, intended to rival ChatGPT, has been delayed to 2025 due to ongoing challenges with accuracy, response speed, and AI “hallucinations” that Amazon is working to resolve.
14. Xbox launches Support Virtual Agent, an AI chatbot to help players resolve support issues efficiently via text or voice.
15. Chinese researchers have adapted Meta’s Llama model for military AI, sparking concerns about open-source access amid U.S. regulatory moves.
16. Regulatory concerns over local impact stall Amazon’s and Meta’s nuclear power agreements for AI-driven data centers.
17. While U.S. AI regulation sees progress at the state level, with Tennessee, Colorado, and California introducing laws on voice cloning protections, risk-based frameworks, and AI training disclosures, roadblocks and resistance, especially in tech-heavy states like California, highlight the need for unified national standards.
18. Caitlin Kalinowski, former head of Meta’s augmented reality (AR) glasses team, has joined OpenAI to lead its robotics and consumer hardware division, while Gabor Cselle, co-founder of Twitter alternative Pebble, also recently joined OpenAI for an undisclosed project.
💰 Top Funding News:
1. Physical Intelligence, which is developing a foundational, task-agnostic software called π0 (pi-zero) that enables robots to perform diverse tasks like folding laundry, bagging groceries, and retrieving toast, raised $400M in early-stage funding, from notable backers like Jeff Bezos, OpenAI, Thrive Capital, and Lux Capital.
2. Hedge fund Coatue Management is raising a $1Bn fund to invest in AI companies, including humanoid robotics, adding to its flagship fund with a focus on institutional and select high-net-worth individual investors.
3. Insider, which provides an AI-powered platform for creating personalized omnichannel marketing and customer engagement experiences, raised a $500M in a Series E, led by General Atlantic.
4. DevRev, which develops AI-driven software designed to connect developers directly with customer needs, raised a $100.8M Series A, led by Khosla Ventures w/ Mayfield Fund, Param Hansa Values, U First Capital, and various accelerators, family offices, and angel investors.
5. Spot AI, which develops Video AI Agents that turn video cameras from passive observers into proactive agents capable of monitoring and taking action in physical spaces, raised $31M in recent funding, led by Qualcomm Ventures, w/ Scale Venture Partners, Redpoint Ventures, Bessemer Venture Partners, StepStone Group, and other investors.