Presented By O’Reilly and Intel AI
Put AI to Work
April 29-30, 2018: Training
April 30-May 2, 2018: Tutorials & Conference
New York, NY

Collaborative machine intelligence: Accelerating human knowledge

Emily Pavlini (Diffeo), Max Kleiman-Weiner (Diffeo)
4:00pm–4:40pm Wednesday, May 2, 2018

Who is this presentation for?

  • Researchers, machine learning engineers, and venture capitalists who are interested in where AI is going

What you'll learn

  • Understand the challenges and promises of building AI agents that deeply collaborate with people on unstructured discovery tasks
  • Explore the current landscape of ML techniques for collaborative machine intelligence (e.g., active learning, machine in the loop, imitation learning, and theory of mind)
  • Discover examples of the practical applications of collaborative intelligence in a semisupervised context of helping users discover novel information over an extended amount of time


Recent advances have made machines more autonomous, but much work remains for AI to collaborate with people. Emily Pavlini and Max Kleiman-Weiner share new insights inspired by the way humans accumulate knowledge and naturally work together that enable machines and people to work and learn as a team, discovering new knowledge in unstructured natural language content together.

Collaborative machine intelligence will enable new products and services that help knowledge workers discover new knowledge answering known unknowns and also helping uncover unknown unknowns. Emily and Max discuss emerging research in this area, such as active learning, dynamic knowledge graphs, imitation learning, and theory-of-mind-based approaches to collaboratively creating cumulative knowledge, and explain how semisupervised hierarchical clustering models generate intuitive interactive structures for uncovering connections and reacting to user actions. Emily and Max also dive into how collaborative agents are transforming the way we work and interact with technology. They demo an AI agent that dynamically infers a knowledge worker’s interests and intents from natural language and finds text passages and entity relations that are both relevant and novel to the user’s current state of knowledge, instead of requiring users to constantly craft queries across many different search tools. The simple act of interacting with this recommended content provides active feedback that simultaneously sharpens the model’s representation of the user’s interests and intents and broadens the system’s ability to traverse farther into the data to find new elements that the user did not yet know.

Photo of Emily Pavlini

Emily Pavlini


Emily Pavlini leads user experience at Diffeo. Emily has a long-standing interest in how people perceive and digest complex information. Previously, she cofounded Meta, a search engine for your personal files, which is now part of the Diffeo platform. Emily has won best pitch and demo awards.

Photo of Max Kleiman-Weiner

Max Kleiman-Weiner


Max Kleiman-Weiner is a cofounder and chief scientist of Diffeo as well as a PhD student in computational cognitive science at MIT, funded by the NSF and the Hertz Foundation. He won best paper at RLDM 2017 for models of human cooperation and the William James Award at SPP for computational work on moral learning. Previously, he was a Fulbright fellow in Beijing. Max holds an MSc in statistics from Oxford, where he was a Marshall scholar, and an undergraduate degree from Stanford, where he was a Goldwater scholar.