NeurIPS 2020 Social on Open Collaboration in ML Research

Making AI research more inviting, inclusive, and accessible is a difficult task, but the movement to do so is close to many researchers' hearts. Progress toward democratizing AI research has been driven by open-access resources: by MOOCs, by accessible code and software, papers, blogs, youtube channels and podcasts dissecting research papers.

However, open, online resources are only part of the equation. Growth as a researcher requires not only learning by consuming accessible information, but hands-on practice whiteboarding, coding, plotting, debugging, and writing. One's growth along these dimensions is often critically enabled by working with and directly learning from collaborators, whether mentors or peers. We learn the most important lessons from each other. Of course, making "collaborators" more universally accessible is fundamentally more difficult than, say, ensuring all can access arXiv papers, because scaling people and research groups is much harder than scaling websites.

Can we nevertheless make access to collaboration itself more open? Can we flatten access to peers and mentors so the opportunities available to those at the best industrial and academic labs are more broadly available to all entrants to our burgeoning field? This social is for those who would like to talk about or to help answer these questions.

Schedule

The social will be Tuesday, 8 December 2020, 2pm - 4pm PT / 5pm - 7pm ET / 11pm - 1am EU / 6am - 8am Taipei. Here's a calendar event to help with timezones.

The below times are in the Pacific timezone.

2:00 - 2:45pm: Panel discussion (link to join; NeurIPS registration needed)

  1. Topic: “Making ML research more open”
  2. Panelists:
    1. Subutai Ahmad, Numenta
    2. Eric Jang, Google
    3. Sara Hooker, Google Brain
    4. Jeremy Howard, fast.ai
    5. Feryal Behbahani, DeepMind
    6. Devin Guillory, UC Berkeley
    7. Olga Russakovsky, Princeton & AI4ALL
  3. Moderated by:
    1. Rosanne Liu
    2. Ashley Edwards

2:45 - 3:15pm: Meet a random researcher! (link to join; open to public. Looking for someone you met in the Icebreaker? Check here to see everyone you matched with!)

  1. We will use a fun platform, Icebreaker.video, to randomly pair participants up for a fixed-time icebreaker chat. Come to meet and greet your potential future collaborators!
  2. Note: Icebreaker.video requires a Google account to sign in.

3:15 - 4:00pm: Discussion tables and casual hangout (link to join; open to public)

  1. We will move over to Gather.town where invited researchers will be staying at tables leading a discussion on their chosen topic. Participants interested in that topic are free to swing by their table. There is also a large space for casual hangouts.
Table Discussion lead Topic
#0 Feryal Behbahani "What objective functions to use to train your agents and/or yourself?"
#1 Eric Jang "How to get better at research before writing your first paper"
#2 Sara Hooker "Trustworthy ML: what it is, why is it important, and what are possible research topics?"
#3 Devin Guillory "Risk of open collaboration and how to protect your diverse teams!
Or Data Augmentations"
#4 Kyunghyun Cho "Being grumpy about (or critical of, if a bit more polite) one's own and others' research"
#5 Subutai Ahmad "Hmm... are your results really correct?? Reproducible ML research in an open, collaborative world"
#6 Ken Stanley & Joel Lehman "Open-endedness: what is it, why is it important to AI, and what are important open research questions within it?"
#7 Laurent Dinh "1/ Deep generative models; 2/ Nurturing your research; 3/ Open discussion :)"
#8 Rewon Child "Unsupervised learning at scale: what’s the right objective? The right model? What are open problems and what does success look like?"
#9 Jeremy Howard "Is deep learning research stuck in a rut, with more following than leading? If so, what can we do about it?"
  1. Read more about each discussion lead and their table's topic here.

Organizers

  1. Rosanne Liu, ML Collective
  2. Ashley Edwards, DeepMind
  3. Jason Yosinski, ML Collective

Resources

If you're looking for collaborators or just want to talk to someone about research, feel free to stop by the Open Collaborators! Discord. Of course, it's open to anyone.