William Agnew

I want to be able to make the world a better place and I believe I can best do this by developing and applying AI.

I’m currently a doctoral student at the University of Washington studying reinforcement learning, planning, and robotics. I'm advised by Pedro Domingos and Sidd Srinivasa and supported by an NDSEG Fellowship. I run marathons, rock climb, backpack, read, and cook in my spare time.

CV Google Scholar Twitter

Contact: wagnew3[at]cs[dot]washington[dot]edu

For anything queer related, please contact me at william dot agnew at ostem dot org for privacy reasons.

News

November 2020: I'm organizing the Queer in AI @ CoRL Social

October 2020: My paper Amodal 3D Reconstruction for Robotic Manipulation via Stability and Connectivity was accepted to CoRL 2020 as an oral (20/~485)!

October 2020: New preprint Relevance-Guided Modeling of Object Dynamics for Reinforcement Learning on arxiv

August 2020: I'm organizing the Resistance AI Workshop and the Object Representations for Learning and Reasoning Workshop at NeurIPS 2020

July 2020: My paper Amodal 3D Reconstruction for Robotic Manipulation via Stability and Connectivity was accepted for a spotlight the the ICML Object-Oriented Learning (OOL): Perception, Representation, and Reasoning workshop

July 2020: I co-organized the Queer in AI ICML 2020 workshop and socials

May 2020: I gave a talk on amodal 3D reconstruction in cluttered environments to MIT CoCoSci



Research

My current research focus is making AI that perceives, learns, and plans like humans. Humans learn much faster and generalize much better than AI. To close these gaps, I'm working on a variety of projects on object oriented representations and learning:

  1. How should the world be represented as objects?

  2. How can we perceive objects?

    1. How can we develop and refine and object representation with no supervision?

    2. How can we quickly create 3D object reconstructions in cluttered environments?

  3. How can we use an object representation to make learning and planning sample efficient, robust, transferable, and explainable?

I'm just as interested in understanding and working to solve the many problems AI creates and amplifies, and envisioning AI research, communities, institutions, and companies that use AI and the power it creates to radically change the world for the better. I explore these issues as an organizer for the wonderful Radical AI community.


Service

I chose to research AI because AI will transform the world, and I want to help make sure AI transforms the world for the better.

Everyone will be affected by AI, so everyone should have an equitable voice in its development:

  • I founded Queer in AI in 2018 to make the AI community more accepting of queer researchers and raise awareness of problems AI can cause for queer people

  • I chaired the first Queer in AI @ NeurIPS satellite workshop, featuring talks by queer AI researchers and a panel on AI ethics for the queer community. I helped organize the Queer in AI ICML 2019, NeurIPS 2019, and ICML 2020 workshops, in addition to many Queer in AI socials

  • I served on the Diversity and Inclusion Advisory Board for the NeurIPS 2018 D&I co-chairs, helping make NeurIPS more welcoming to queer people

  • I serve as the Vice President of External STEM Partnerships of oSTEM

  • I helped start UW CSE's queer prospectives reception at our prospective grad student visit days

Robotics competitions gave me the opportunity to learn many of the skills I use as a researcher. Giving students the same opportunities is incredibly meaningful (and fun!) for me.

  • I've volunteered in 25+ FLL, FTC, FRC, and BEST robotics competitions as a design judge, pit boss, or inspector

  • I've served on the planning committee of Georgia BEST

As an undergraduate at Georgia Tech, I helped start and lead the Undergraduate Research Ambassadors Program and the Big O Theoretical Computer Science Club, both of which have provided incredible opportunities, mentorship, and community for countless undergrads interested in research. I also proposed and organized the first Home Depot Deep Learning Competition which gives Georgia Tech students at a hands-on introduction to deep learning each year.

Students

Mentoring brilliant students is one of the most impactful and meaningful things I do. If you are interested in working with me, please email me your resume and 2-3 paragraphs describing your interests.

Undergraduate

Zhichao Lei

John Barcellos

Jize Cao

Christopher Kang

Ryan Pachauri

Joshua Brockschmidt

Caelen Wang