Top Science Video Features UW Research

Views
Former UW College of Engineering and Applied Science student Anh Nguyen, center, worked with Laramie Robotics Club member Joseph Kilpatrick on robotic coding. (UW Photo)

LARAMIE– The work of researchers at the University of Wyoming was featured in a year-end review of content produced by a prestigious international journal.

In 2017, the video team for the journal Science created nearly 180 videos on various topics. The most-viewed entry for the entire year accompanied a special package on artificial intelligence and featured Science staff writer Paul Voosen.

UW Associate Professor and Graduate Student Work on the Team of Researchers

The basis for the video, “A.I. detectives are cracking open the black box of deep learning,” came largely from the work of a team of researchers, two of whom are associated with UW: Department of Computer Science Associate Professor Jeff Clune and graduate student Anh Nguyen, who now is an assistant professor at Auburn University.

Advertisement - Story continues below...

Science’s video explores information from a video, titled “Deep Visualization Toolbox” (www.youtube.com/watch?v=AgkfIQ4IGaM) and a paper, “Understanding neural networks through deep visualization,” both including contributions from Clune and Nguyen, along with Cornell University’s Jason Yosinski and Hod Lipson, and California Institute of Technology’s Tom Fuchs.

The information was presented at the Deep Learning Workshop of the International Conference on Machine Learning in 2015.

The Negatives of Using Machine Learning in Science

Voosen explores the pitfalls of using machine learning to conduct science, and what comes next. It can be viewed here: www.sciencemag.org/news/2017/12/these-science-videos-topped-charts-and-stole-our-hearts-2017.

The Science video explains how neural networks are becoming more widely used in many industries, including voice-recognition technology, autonomous cars and genetic sequencing.

The concept can be explained by a network of neurons connected with one another loosely inspired by our brains. These networks receive data (images) as inputs and must recognize patterns in the data to make decisions, such as drive cars autonomously.

The “Deep Visualization Toolbox” Can Decipher how Neurons Interact with Data

While this technology is powerful, there are limitations, including the fact that it can be difficult to decipher how they work.

That’s where the “Deep Visualization Toolbox” comes in. It can isolate individual neurons and discover how they interact with data to shed light on how these neural networks do the things they do, Clune says.

“I am delighted that the journal Science is covering this important research,” Clune says. “Artificial intelligence and, in particular, deep neural networks will lead to dramatic changes in every economic sector, scientific field and in many cultural areas.

It is increasingly being deployed throughout society, despite the fact that we do not know exactly how it works, when it is biased and how to prevent it from being easily manipulated.

It is Critical to Understand the Benefits and Harms of this Technology

“Our work is part of an effort to better understand this technology and improve it, and we are delighted that the scientific community is so interested in such research.

“The possible benefits from artificial intelligence are enormous, but so are the potential societal harms,” he adds. “It is, thus, critical to improve our understanding of the technology to harness it as best we can while mitigating its downsides.”