Understanding AI Agents and Why They Matter

This weekly update covers an exciting time for agentic AI, which is artificial intelligence that can think and make decisions on its own. Unlike regular AI chatbots that just answer questions, agentic systems can plan ahead, use different tools, and complete complicated tasks with very little help from humans. Think of it like the difference between a helper who answers your questions and a worker who can figure out what needs to be done and do it all by themselves. These AI agents are moving from test projects into real-world use this week, and that is a big deal for science, medicine, and business.

Duke University Creates "Virtual Scientists"

One of the biggest discoveries this week came from Duke University, where scientists built a team of AI agents that can solve hard design problems. The researchers created what they call an "agentic system"—which is really multiple AI helpers working together. Each AI agent has a special job. One agent makes sure all the data is organized correctly. Another agent writes computer code from scratch. A third agent checks the work to make sure it is right. A fourth agent tests the code to see if the answer is good. Finally, a leader AI agent helps all the other agents talk to each other and make decisions. When the scientists tested this "artificial scientist," it solved problems almost as well as real PhD students. The researchers say that AI agents like this could help scientists move much faster in many different fields like medicine, materials, and engineering.

Three Tech Giants Fighting to Lead the AI Agent Race

This week revealed that Google, OpenAI, and Anthropic—three of the world's biggest AI companies—are competing hard to create the best agentic AI systems. Each company has a different plan. OpenAI released AgentKit, which gives computer programmers the pieces they need to build AI agents quickly. OpenAI also showed a Computer-Using Agent, which means an AI that can see what is on a computer screen, move the mouse, and type on the keyboard—just like a person would. Google announced Gemini Enterprise this week, which helps big companies use AI agents safely while keeping everything under control through central policies. Google also uses a powerful AI model called Gemini 2.0 that can understand pictures, sound, and text all together. Anthropic is taking a different path by making sure humans stay in charge of what AI does. Their Computer Use system lets people check and approve what AI agents do before they do it. The main question is which company's approach will work best for helping people and companies use AI agents safely and well.

Learning How AI Agents Change Work and Science

On October 21, Michigan State University held a big meeting called the Data Science + AI Summit. More than 200 people came together—students, scientists, business leaders, and teachers—to talk about how agentic AI is changing the world of work. The main idea everyone agreed on was that AI agents should work with humans, not instead of humans. People learned that when AI agents help workers do their jobs better, instead of replacing them, people do better work and feel better about their jobs. The people at the meeting also talked about how schools need to teach students new skills so they are ready for jobs where they work alongside AI agents. Leaders from the university, companies, and the government all said that teamwork between universities, startups, and big companies will help create the best AI innovations that actually help people.

Real-World Uses in Medicine and Business

This week showed that agentic AI is not just an interesting idea anymore—it is actually helping people right now. In medicine, NVIDIA-powered agentic AI is being used with IQVIA, a big company that helps test new medicines. These AI agents help speed up every step of making and testing new drugs, which could help sick people get medicine faster. In supply chains and customer service, big companies are already using AI agents to handle repetitive work, which frees up humans to do more creative and important tasks. Scientists at Google DeepMind made a big breakthrough by teaching AI to understand the world using pictures, sounds, and movement information all at the same time. This kind of smart multimodal AI will help make robots smarter, create better tools for blind people, and enable many new uses we have not even thought of yet.

New Computer Chips and Better Safety

The week also brought progress in the hardware and safety sides of AI. New computer chips are being built that can run powerful AI systems on regular computers and phones, not just in huge data centers. This means more people can use advanced AI even without expensive equipment. At the same time, companies are taking safety seriously. OpenAI launched new safety layers to help organizations use AI agents in responsible ways. These safety tools help catch problems before they become big issues and make sure AI is helping people and not causing harm. The overall message is clear: agentic AI is ready for the real world, and it is already making a difference in science, medicine, and business. The next few years will show how these AI agents change everything we do.