Google is developing a groundbreaking multi-agent system within Gemini for Enterprise, capable of generating and evaluating ideas through a tournament-style approach. Users can input a topic and evaluation criteria, allowing the system to work on a single problem for approximately 40 minutes—a significant continuous run time. This innovative tool can produce around 100 ideas that are ranked, accompanied by detailed descriptions and performance reports. Two pre-made agents, “Idea Generation” and “Co-scientist,” facilitate idea generation and scientific research, respectively. The Co-scientist agent focuses on producing valuable insights for research teams. Additionally, there’s a “chat with your docs” feature enabling users to analyze and interact with PDF documents, enhancing data extraction capabilities. This multi-agent tournament system represents a substantial leap in AI-driven enterprise tools, positioning Google at the forefront of Level 3 AI. While still in development, this feature could revolutionize how businesses and research organizations approach idea generation and problem-solving.
Source link
