View this email in your browser
Dear colleagues,

We hope this note finds you well and in good spirits. Like you, we've been looking for ways to contribute in this challenging time. We released a call last month for proposals for AI-related projects to address the Covid-19 pandemic. We recently followed that with our regular annual call for proposals, for 2020-2021. If you have an idea for a project and are looking for MIT or IBM collaborators, we invite you to attend our virtual networking and poster session on Friday, May 8.

In the meantime, we'd like to share with you our new MIT-IBM Watson AI Lab website and visual identity. You may recognize the play on the horizontal and vertical lines of our respective logos, with the converging 5-bar and 8-bar lines representing our mutual collaboration. We hope this spirit of teamwork comes across throughout the site, including in our expanded news and research pages.

This spring, the Lab had another strong showing at the Association for the Advancement of Artificial Intelligence conference and the International Conference on Learning Representations. We highlight some of that work below and on our new website.  

We hope to see you Friday at the poster session!

Please take good care, 

Aude Oliva, MIT Director, MIT-IBM Watson AI Lab
David Cox, IBM Director, MIT-IBM Watson AI Lab


A Foolproof Way to Shrink Deep Neural Networks

As more AI applications move to smartphones, deep learning models are getting smaller to allow apps to run faster and save battery. Now, MIT researchers have a new and better compression technique. “It’s the pruning algorithm from the ‘Book,’” says MIT's Jonathan Frankle. “It’s clear, generic and drop dead simple.” Read the story in MIT News and on our blog.
Reducing AI's Carbon Footprint

In related work, MIT and IBM researchers have developed a new automated technique for designing efficient deep learning models. “The upside of developing methods to make AI models smaller and more efficient is that the models may also perform better,” says IBM's John Cohn. Read the story in MIT News and VentureBeat.

Using Music and AI to Crack the Coronavirus 

MIT's Markus Buehler recently translated the structure of the novel coronavirus into music to visualize its vibrational properties, which could help to pinpoint sites on the protein for antibodies or drugs to target. In this Q&A with MIT News, Buehler discusses his broader work developing AI models to design new proteins for sustainable, non-toxic applications. 

Uncovering the Hidden Power of Vitamin A  

In a new study in Cell Reports, researchers discover that vitamin A palmitate, a common supplement, and gum resin, a popular glazing agent for pills and chewing gum, could make hundreds of drugs more effective, from blood-clotting agents to anti-cancer drugs. “Machine learning gives you a way to narrow down the search space,” MIT's Giovanni Traverso tells MIT News.


Bringing Deep Learning to Life

For a third year in a row, PhD students Alexander Amini and Ava Soleimany taught their crash course in deep learning at MIT. More than 350 MIT students enroll each year, and another million people from around the world have watched their lectures online. This year, students heard from IBM's David Cox on neuro-symbolic AI, or merging deep nets with symbolic programs.

How Green Is Your Deep Learning Model?

Training a big AI model is energy-intensive, but doesn’t have to be. MIT’s first Green AI Hackathon, co-sponsored by the MIT Research Computing Project and our own MIT-IBM Watson AI Lab, drew several dozen students and generated a list of promising ideas for shrinking the carbon footprint of modern AI models. Check out some of the energy-saving hacks.


For AI to advance it must understand the cause and effect relationships that underlie daily life. A new Lab-created dataset aims to train computers in basic causal reasoning. Read the story in Wired, and on our blog.
Automation may not necessarily kill jobs — but employees and managers will need to work differently. Writing in Harvard Business Review, IBM chief economist Martin Fleming discusses the future of work as artificial intelligence upends business-as-usual.
The future of AI depends on systems that learn on their own, with little to no human supervision. IBM's David Cox offers ideas for how to get there in this New York Times piece on the state of artificial intelligence.
MIT and IBM joined a national consortium to donate supercomputing time to Covid-19-related projects. "Ultimately we need a cure," IBM's Dario Gil told the Wall Street Journal. "To be able to tackle that, we need to accelerate science."


MIT-IBM Virtual Poster Session 

4 pm - 5:30 pm, Friday, May 8, 2020

If you have an active project with the Lab or are looking for collaborators, please join our annual poster session. Aude and David will open with a Q&A, and then hand the stage to participants to view and discuss research posted on a secure website. RSVP by 5 pm May 6.
MIT-IBM Watson AI Lab, 75 Binney Street
Cambridge, Mass., 02142

This email was sent to *|EMAIL|*
why did I get this?    unsubscribe from this list    update subscription preferences