Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the challenges and solutions for building AI agents capable of performing computer tasks in this conference talk by cognitive scientist Danielle Perszyk from Amazon AGI SF Lab. Learn about the concept of Useful General Intelligence as a new approach to creating general-purpose agents that can reliably click, type, scroll, and perform any task humans can do on computers. Discover why enabling AI to execute basic computer operations proves more challenging than code generation, and examine the specific obstacles that must be overcome to achieve truly functional AI agents. Gain insights into Amazon's work at their AGI SF Lab, including their development of Nova Act, an agentic model and SDK that allows developers early access to cutting-edge agent technology. See real-world examples of workflows that developers have successfully built using these tools, and understand the practical applications of this emerging technology. The presentation draws from Perszyk's background in cognitive science and her research on language evolution and development, providing a scientific foundation for understanding how to bridge the gap between current AI capabilities and true general intelligence for computer use.
