MIT-IBM Watson AI Lab

The GANpaint system developed at MIT can easily add features to an existing image. At left, the original photo of a kitchen; at right, the same kitchen with the addition of a window. Co-author Jun-Yan Zhu believes better understanding of GANs will help researchers be able to better stamp out fakery: “This understanding may potentially help us detect fake images more easily.”

Teaching artificial intelligence to create visuals with more common sense

An MIT/IBM system could help artists and designers make quick tweaks to visuals while also helping researchers identify “fake” images.

July 1, 2019

Users feed Gen relatively short code defining a target task, and the system automatically generates the results.

New AI programming language goes beyond deep learning

General-purpose language works for computer vision, robotics, statistics, and more.

June 26, 2019

Researchers have developed a flexible way of combining deep learning and symbolic reasoning to teach computers to write short computer programs. Here, Armando Solar-Lezama (left), a professor at CSAIL, speaks with graduate student Maxwell Nye.

Toward artificial intelligence that learns to write code

Researchers combine deep learning and symbolic reasoning for a more flexible way of teaching computers to program.

June 14, 2019

Phillip Isola is an expert in generative adversarial networks, a form of deep learning that is changing how we look at images and inspiring applications far beyond computer vision. Isola helped organize GANocracy, a day-long workshop at MIT on the art and science of GANs.

Q&A: Phillip Isola on the art and science of generative models

Image-translation pioneer discusses the past, present, and future of generative adversarial networks, or GANs.

May 30, 2019

In a pair of studies, researchers show that grammar-enriched deep learning models understand some key rules about language use. Peng Qian (left) and Ethan Wilcox, graduate students at MIT and Harvard University respectively, presented the work at a recent MIT-IBM Watson AI Lab poster session.

Teaching language models grammar really does make them smarter

Researchers submit deep learning models to a set of psychology tests to see which ones grasp key linguistic rules.

May 29, 2019

(L-R) MIT Assistant Professor Michael Carbin and PhD student Jonathan Frankle

Smarter training of neural networks

MIT CSAIL project shows the neural nets we typically train contain smaller “subnetworks” that can learn just as well, and often faster.

May 6, 2019

Three-part illustration: On left is a battery labeled "efficiency"; in center is a robot labeled "defensive quantization"; on right is a lock symbol labeled "security."

Improving security as artificial intelligence moves to smartphones

Researchers unveil a tool for making compressed deep learning models less vulnerable to attack.

April 23, 2019

A new “particle simulator” developed by MIT researchers improves robots’ abilities to mold materials into simulated target shapes and interact with solid objects and liquids. This could give robots a refined touch for industrial applications or for personal robotics— such as shaping clay or rolling sticky sushi rice.

Giving robots a better feel for object manipulation

Model improves a robot’s ability to mold materials into shapes and interact with liquids and solid objects.

April 16, 2019

$Researchers trained a hybrid AI model to answer questions like “Does the red object left of the green cube have the same shape as the purple matte thing?” by feeding it examples of object colors and shapes followed by more complex scenarios involving multi-object comparisons. The model could transfer this knowledge to new scenarios as well as or better than state-of-the-art models using a frac...$