← back

Useful General Intelligence — Danielle Perszyk, Amazon AGI

8.2K views · Aug 02, 2025 · 19:58 min · Watch on YouTube ↗
Takeaway

Amazon's bet is reliable browser-based agents (NovaAct) that augment humans rather than replace them — automation has to lead to augmentation, not echo-chamber lock-in.

Summary

  • Cognitive scientist at Amazon AGI SF Lab argues 'general intelligence' is not in a thinking machine but in the human-plus-AI loop; Engelbart's augmentation framing vs the McCarthy thinking-machine framing.
  • Hallucinations are features not bugs — needed to go beyond the data — but must be controlled; aligns with how human perception works as 'controlled hallucination'.
  • Introduces NovaAct: an Amazon-Nova model fine-tuned for high-reliability UI interaction plus an SDK; demos finding a 2BR/1BA in Redwood City and computing bike-time to Caltrain via Python integrations in three lines.
  • Frames computer-use as the atomic unit of digital interaction since most sites lack APIs — even icon interpretation is hard, so the model has to be UI-trained, not just multimodal.
agentscomputer-useamazon-nova
Original description
We’re all hearing that AI agents will enable AGI, but they can’t yet reliably perform even basic computer tasks. It turns out that getting AI to click, type, and scroll is more challenging than getting it to generate code. How can we build general-purpose agents that can do anything we can do on a computer?

This is our goal at the Amazon AGI SF Lab. In this talk, I’ll propose a new approach to agents that we call Useful General Intelligence. After describing how we’re solving the biggest challenges in computer use while enabling developers to access our tech in it’s earliest developmental stages, I’ll show real workflows that developers have built with Nova Act, our agentic model and SDK.

About Danielle
Danielle is a cognitive scientist at the new Amazon AGI SF Lab. She received her PhD from Northwestern, where she studied the evolution and development of language. Previously, she was at Google and Adept.

Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter