NVIDIA Powers Cal Poly AI Factory Launch

Key Takeaways

• <ul><li>Cal Poly launched an NVIDIA-powered AI Factory on October 22, with NVIDIA VP Charlie Wuischpard and President Jeffrey D. Armstrong attending the ribbon-cutting.</li></ul>
• <ul><li>UC Berkeley's Ken Goldberg and NVIDIA's Spencer Huang are developing agentic robotics where AI agents write and test robot code in simulation, reducing real-world data needs.</li></ul>
• <ul><li>OpenAI's Astra model has improved the agentic robotics pipeline, and the team is releasing open-source code on NVIDIA's Isaac simulation framework.</li></ul>
• <ul><li>Researchers warn that AI-powered anesthesia systems cannot replace human clinical intuition, situational awareness, or ethical judgment in operating rooms.</li></ul>
• <ul><li>Recursive self-improvement in AI is now plausible but lacks clear safety evaluation frameworks, with game-theoretic analysis being explored to assess risks.</li></ul>
• <ul><li>Google DeepMind is combining human reviewers and AI agents to test Signal's secure messaging, though the system is still in early stages.</li></ul>
• <ul><li>A survey of 79 university students found 55 percent use AI for studying, but 67 percent worry about accuracy, bias, and privacy, calling for clearer university policies.</li></ul>
• <ul><li>Ayanna Howard, president of Spelman College, published Rebooting the Machines to encourage public engagement with AI, with proceeds funding Spelman scholarships.</li></ul>
• <ul><li>PSI, founded by CEO Matt Pines, uses AI to solve hard scientific problems and recently had MIT graduate David Turturean help solve the Aaronson-Ambainis conjecture in quantum computing.</li></ul>
• <ul><li>PSI currently earns revenue optimizing data center physics and aims to turn theoretical physics breakthroughs into real-world technologies.</li></ul>

AI agents test robot code in virtual worlds before real use

Ken Goldberg and his team at UC Berkeley, working with NVIDIA's Spencer Huang, developed a new method where AI agents write robot code and test it in simulations. The agents revise the code when something fails, reducing the need for massive real-world data. This approach, called agentic robotics, was presented at the Conference on Robot Learning. The team also released open-source code and an industry benchmark for evaluating robots on industrial tasks.

AI agents generate and refine robot control programs

Ken Goldberg and Spencer Huang from NVIDIA are leading research in agentic robotics, where AI teams create and improve robot programs through simulation. This method addresses the huge data requirements and engineering effort traditionally needed for robot training. The team uses a Graph-as-Policy framework that breaks tasks into individual skills like finding objects and planning grasps. OpenAI's Astra model has made the process more promising, and the team is releasing open-source code built on NVIDIA's Isaac simulation framework.

Automated anesthesia still needs human clinical judgment

Researchers Fredericus H. J. van Loon, Menno A. de Wilde, and Angelique T. M. Dierick-van Daele published a correspondence in Anesthesiology and Perioperative Science warning that AI-powered anesthesia systems cannot replace human intuition. While intelligent closed-loop platforms can automate drug delivery and reduce workload, they lack situational awareness, ethical responsibility, and the ability to detect environmental changes during surgery. The authors stressed that clinical intuition and transparency in AI models are essential for patient safety in operating rooms.

Cal Poly launches NVIDIA-powered AI Factory for education

Cal Poly University unveiled an NVIDIA-powered AI Factory to expand artificial intelligence education and research. The ribbon-cutting ceremony took place during Cal Poly AI Day on October 22, with Cal Poly President Jeffrey D. Armstrong and NVIDIA Vice President Charlie Wuischpard in attendance. The facility aims to give students and researchers advanced tools for AI learning and development.

Is recursive self-improvement in AI safe?

Recursive self-improvement describes an AI system that designs a better AI system, which then creates an even better one in a continuous cycle. Once an abstract concept, this idea now seems plausible because today's frontier AI has advanced abilities in programming, mathematics, and tool use. AI companies are beginning to pursue RSI, but many researchers are raising concerns about safety. The author argues that clear frameworks for evaluating RSI outcomes are still missing and presents a game-theoretic analysis for deciding whether to initiate such systems.

Google DeepMind teams humans and AI to test Signal messaging security

Google DeepMind created a new system that combines human reviewers and AI agents to check Signal's secure messaging techniques. The system uses machine learning and human oversight to find and flag suspicious signals. It is still in its early stages of development.

University students use AI more but many do not like it

A survey of 79 university students found that most use AI for studying but many are uncomfortable with it. About 55 percent use AI to speed up their work and overcome mental blocks. However 67 percent worry about AI accuracy and many have ethical concerns about bias and privacy. Students also said universities need clearer rules about AI use.

Spelman's Ayanna Howard urges people to reboot their relationship with AI

Ayanna Howard is a robotics researcher and president of Spelman College. In her new book Rebooting the Machines she says people should reboot their relationship with technology and AI. She believes anyone can help shape AI's future whether or not they know how to code. All book proceeds will support scholarships for Spelman students.

AI video creator speaks proudly about making AI slop

A writer shares his experience making AI-generated videos despite criticism. He says AI video production is fun and a powerful tool for creativity and self-expression. He started a blog and community to share tips and techniques with other AI video enthusiasts. He believes the creative possibilities of AI are special and worth exploring.

Asia's AI boom faces timing and infrastructure challenges

Asia's AI boom is boosting exports and investment across Asean+3 countries including China, Japan and South Korea. But investors and policymakers warn of three mismatches between technology investment and infrastructure. They worry about what happens when investment, power infrastructure, productivity and financial markets move at different speeds.

PSI builds a business from AI-driven scientific discovery

PSI is a startup founded by CEO Matt Pines that uses AI to solve difficult scientific problems. Recent MIT graduate David Turturean joined PSI and helped solve the Aaronson-Ambainis conjecture, a major problem in quantum computing. The company currently earns revenue by using AI tools to optimize data center physics like heat and power. PSI aims to turn theoretical physics breakthroughs into real technologies by combining AI research with experiments. The team focuses on validating computational results efficiently to build new devices and capabilities.

Sources

• AI agents test and improve robot code in virtual worlds before deployment
• AI Agents Generate and Refine Robot Control Programs
• Automated Anesthesia Systems Require Human Clinical Intuition Alongside Artificial Intelligence
• Cal Poly unveils NVIDIA-powered 'AI Factory' to expand AI education, research
• Is Recursive Self-Improvement Safe?
• Google DeepMind melds humans and AI agents to vet Signal's secure messaging techniques
• Most university students now use AI, but that doesn't mean they like it
• Spelman’s Ayanna Howard Looks Beyond AI’s Hype and Doom
• How I became a proud producer of slop
• Asia’s AI boom has a timing problem
• PSI is trying to make a business out of AI-driven scientific discovery