In recent years, the best-performing systems in artificial-intelligence research have come courtesy of neural networks, which look for patterns in training data that yield useful predictions or classifications. A neural net might, for instance, be trained to recognize certain objects in digital images or to infer the topics of texts. But neural nets are black boxes. After training, a network may be very good at classifying data, but even its creators will have no idea why. With visual data, it's sometimes possible to automate experiments that determine which visual features a neural net is responding to.

The information age has generated far more data than humanity can ever hope to manually process, but with the help of artificial intelligence, that avalanche of data is now revealing itself to be far more useful than we ever thought possible. The omnipresent wireless signals that keep us connected can now be used like X-rays to see and track the movements of people, even when hidden behind walls. Though they're invisible to the human eye, radio waves still bounce off of human bodies as the wireless signals emanate out from broadcast antennas. How those radio signals bounce and scatter can be measured, and researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) were able to train a neural network to extract the positions and movements of people as they interfere with radio frequency (RF) signals. Like babies, neural networks need to be trained on what to look for when analyzing the world.

Today the Association for Computer Machinery (ACM) announced its 2014 fellows, and among the awardees were five researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) -- more than any other academic institution in the world. Srini Devadas, Eric Grimson, Robert Morris, Ronitt Rubinfeld, and CSAIL Director Daniela Rus were among the 1 percent of ACM members to receive the distinction, according to the association's press release. The ACM fellows, chosen from universities, corporations, and research labs, were selected for contributions that have provided key knowledge to the computing field and generated multiple technology advances in industry, commerce, healthcare, entertainment, and education. "While it certainly isn't unprecedented for CSAIL researchers to receive this honor, it is quite remarkable that this year ACM has chosen to recognize five members of our community," said Daniel Jackson, associate director of CSAIL. "We are extremely proud of our PIs who have been selected to be part of such esteemed company."

In recent years, the best-performing systems in artificial-intelligence research have come courtesy of neural networks, which look for patterns in training data that yield useful predictions or classifications. A neural net might, for instance, be trained to recognize certain objects in digital images or to infer the topics of texts. But neural nets are black boxes. After training, a network may be very good at classifying data, but even its creators will have no idea why. With visual data, it's sometimes possible to automate experiments that determine which visual features a neural net is responding to.

With visual data, it's sometimes possible to automate experiments that determine which visual features a neural net is responding to. But text-processing systems tend to be more opaque. At the Association for Computational Linguistics' Conference on Empirical Methods in Natural Language Processing, researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) will present a new way to train neural networks so that they provide not only predictions and classifications but rationales for their decisions. "In real-world applications, sometimes people really want to know why the model makes the predictions it does," says Tao Lei, an MIT graduate student in electrical engineering and computer science and first author on the new paper. "One major reason that doctors don't trust machine-learning methods is that there's no evidence."