OpenAI proposes framework to revolutionize U.S. AI infrastructure

Today’s Highlights:

📰 News: OpenAI proposes framework to revolutionize U.S. AI infrastructure + EU releases compliance guide for AI Act + Tons of updates from Anthropic, OpenAI, Google, Apple, Microsoft and more

💰 Funding: Pony AI the Chinese autonomous driving startup prepares for US IPO

⚡️ Top News Stories:

1. OpenAI proposes an ambitious framework to enhance U.S. AI infrastructure, which includes establishing AI Economic Zones, a National Transmission Highway Act, and increased private-public investment in AI-related projects to compete with China’s AI advancements.

• The plan proposes AI Economic Zones to streamline permits for AI infrastructure projects, with OpenAI envisioning the Midwest and Southwest as key regions for future data centers powered by renewable energy like solar, wind, and small modular nuclear reactors, leveraging expertise from the U.S. Navy.

• OpenAI proposes a “North American Compact for Artificial Intelligence” to foster a cooperative AI alliance among North American nations, with potential expansion to other democratic allies, aiming to establish an economic bloc capable of competing with China’s AI advancements.

• The blueprint includes discussions of a $100Bn, 5-gigawatt data center project—5x larger than any existing facility—to meet the growing computational demands of advanced AI systems.

• National Transmission Highway Act Proposal: The proposed act aims to modernize U.S. infrastructure, expanding energy, fiber, and natural gas to meet the demands of AI development, modeled after the 1956 Interstate and Defense Highways Act.

• OpenAI calls for governmental programs to prepare a workforce equipped for AI infrastructure roles, such as managing and operating AI-focused data centers and other infrastructure projects.

2. Anthropic has introduced new features in its developer console to enhance prompt engineering

• Prompt Improver that refines prompts with techniques like chain-of-thought reasoning, boosting accuracy by 30% in multilabel tasks

• Example Management for editing input/output pairs in the Workbench to improve consistency

• Prompt Evaluation tool with an "ideal output" column for benchmarking and refining prompts on a 5-point scale.

3. The EU has released a preliminary draft of the General Purpose AI (GPAI) Code of Practice, open for feedback until Nov 28, 2024. It will guide compliance with the EU AI Act for foundational AI model providers like OpenAI, Google, and Meta.

• Transparency and Copyright Rules: The draft emphasizes transparency, requiring GPAI providers to disclose training data sources, web crawlers used, and handling of copyrighted content. Providers must also establish contact points for grievances.

• Systemic Risk Mitigation: The Code identifies systemic risks (e.g., cybersecurity threats, disinformation, discrimination) and proposes measures for risk assessment, including advanced evaluation methods like red-teaming and forecasting.

• Focus on Safety and SMEs: A proposed Safety and Security Framework mandates policies for managing risks and forecasting potential systemic risks. The draft seeks proportional measures tailored to smaller companies, including start-ups and open-source developers.

4. OpenAI is set to launch “Operator,” a general-purpose AI agent capable of autonomously performing tasks like coding or booking travel within a web browser, with a research preview and developer API targeted for January, marking a shift from response generation to practical task execution.

5. The Department of Homeland Security (DHS) has released voluntary AI guidelines for critical infrastructure companies, including airports, water utilities, and energy providers, developed with input from a board of experts and industry leaders like the CEOs of OpenAI, Nvidia, and Alphabet, recommending measures to enhance safety, privacy, and transparency in AI use.

6. The Biden administration has issued AI privacy guidance for military and intelligence agencies, outlining rules for handling Americans' private data by balancing civil liberties with national security, clarifying that acquiring AI trained on U.S. data isn’t "collection" unless the original data is accessible, and requiring compliance with privacy laws like the Privacy Act for refining AI systems, while encouraging tracking, labeling, and documentation for oversight.

7. Anthropic is collaborating with the Department of Energy's National Nuclear Security Administration (NNSA) on a pilot program, using "red-teaming" exercises to test whether its AI model, could be misused to access or share sensitive nuclear information, marking the first use of a frontier AI model in a classified government setting.

8. At the Baidu World Conference, Baidu unveiled new AI tools, including I-RAG, a text-to-image generator that reduces hallucinations using Baidu’s search engine for improved accuracy; Miaoda, a no-code development tool enabling software creation without coding expertise; and an AI assistant integrated into smart glasses.

9. Apple is set to launch a wall-mounted smart home display that combines AI with HomeKit controls to manage appliances, enhance security, and enable video conferencing, with a release anticipated as early as March.

10. Researchers from Johns Hopkins and Stanford successfully trained a robot to perform surgical tasks by watching videos of experienced surgeons, bypassing the need for step-by-step programming.

11. Qwen2.5-Coder, a new open-source code model from Qwen, offers developers scalable tools for code generation, repair, and reasoning, achieving state-of-the-art performance across 40+ programming languages and positioning itself as a strong alternative to models like GPT-4o.

12. Formation Bio, in collaboration with Sanofi and OpenAI, has launched Muse, an AI-powered tool that streamlines patient recruitment for clinical trials by analyzing scientific literature and real-world data to create tailored strategies, identify optimal patient profiles, and produce diverse recruitment materials, reducing timelines from months to minutes and enhancing inclusiveness.

13. Leading AI firms like OpenAI, Meta, Microsoft, and Anthropic are overhauling model testing benchmarks to meet the demands of rapid advancements, shifting toward complex and realistic proprietary evaluations as traditional public tests prove insufficient, raising concerns about comparability without publicly accessible metrics.

14. Microsoft has launched tailored AI models for specific industries in collaboration with partners like Bayer, Cerence, and Siemens, addressing sector-specific needs—such as Bayer's crop protection in agriculture, Cerence's in-car automotive assistant, and Siemens' CAD copilot for design—through models pre-trained with industry data and accessible via Azure AI Studio.

15. The Wall Street Journal is testing AI-generated article summaries in a “Key Points” box at the top of articles to enhance engagement, with each summary reviewed by an editor and a note clarifying AI involvement, linking to information on WSJ’s AI tool usage.

16. OpenAI has created a student guide that recommends using ChatGPT for concept mastery and writing improvement, rather than as a shortcut to bypass learning, while advocating responsible citation for academic integrity.

17. Google has released its Gemini AI chatbot app for iPhone, providing a dedicated, straightforward chat interface accessible via text, voice, or camera.

18. AI pioneer François Chollet, known for creating the popular Keras library, is leaving Google after nearly a decade to start a new venture, though details are limited.

19. OpenAI co-founder Greg Brockman has returned from a three-month leave to rejoin the leadership team, where he will collaborate with CEO Sam Altman to define a new role focused on tackling significant technical challenges amid executive turnover.

20. Spotify is facing backlash over a surge of AI-generated fake albums appearing under verified artist names, which disrupt royalty distribution and mislead fans by compromising artist profiles.

21. Google DeepMind has introduced "Gemma Scope," a tool designed to improve understanding of AI processes by allowing researchers to look into how neural networks process information, aiming to identify and control potential flaws.

22. OpenAI’s ChatGPT desktop app for macOS now includes a "Work with Apps" feature that automatically reads and analyzes code from apps like VS Code, Xcode, TextEdit, Terminal, and iTerm2, eliminating the need to copy-paste by sending code sections as context with user prompts.

23. InVideo launched a new gen-AI feature allowing users to create live-action, animated, or anime videos from text prompts, marking a significant upgrade in video editing capabilities.

💰 Top Funding News:

1. Pony AI, the Chinese autonomous driving startup, aims to raise up to $224M through a U.S. IPO, targeting a $4.48B valuation.

2. Amazon announced a $110M "Build on Trainium" initiative providing universities access to AWS Trainium UltraClusters for AI research, focusing on generative AI to develop new architectures, machine learning libraries, and performance optimization tools, with all outcomes shared as open-source innovations.

3. CoreWeave, which operates a purpose-built AI hyperscale cloud platform optimized for high-performance computing, raised $650M through a secondary sale.

4. Writer, which focuses on enabling enterprise-grade generative AI with tools for automating workflows, providing structured AI applications, and creating domain-specific models for industries, raised a $200M Series C, le by Premji Invest, w/ Radical Ventures, Salesforce Ventures, Adobe Ventures, Citi Ventures, B Capital, IBM Ventures, Workday Ventures, and existing investors such as ICONIQ Growth, Insight Partners, and Balderton Capital.

5. Dexory, which combines autonomous robots with AI to optimize warehouse operation, raised a $80M Series B, led by DTCP, w/ Latitude Ventures, Wave-X, Bootstrap Europe, Atomico, Lakestar, and Capnamic. 

6. ScaleOps, which leverages AI to automate cloud resource management, optimizing application performance and minimizing cost, raised a $58M Series B, led by Lightspeed Venture Partners, w/ NFX, Glilot Capital Partners, and Picture Capital.

7. Tessl, which leverages AI to create and maintain software code autonomously, raised $125M  across a Seed and Series A, led led by Index Ventures, w/ Accel, GV (Google Ventures), and Boldstart. GV and Boldstart co-led the seed round.

8. Stepful, which leverages AI to support healthcare training by providing personalized engagement for students, raised a $31.5M Series B, led by Oak HC/FT, w/ Y Combinator, Reach Capital, AlleyCorp, and others.

9. Team-GPT, which offers a suite of AI tools, such as Claude and ChatGPT, that companies can integrate into workplace systems, raised a $4.5M Seed round, led by True Ventures.

That's all for today's email! If you want more please follow us at the social channels linked below, or check out our website!

Share our newsletter: If you like our work please share/forward this email with your friends, colleagues, and family. It's the best way to support us!

If this email was forwarded to you please sign up here to continue receiving them.

Want your content, product, jobs, or event featured in our newsletter? Reply to this email with the details, and our team will reach out to you.

Do you use AI for work? Tell us how, and you could be featured in our newsletter!

Check out our website for more resources, including a list of AI investors, products, events, and twitter follows.

For an archive of all our posts, click here.

We'd love to hear from you! You can always leave us comments or feedback by replying to this email!

Powered by AI. Curated and edited by Humans.