Multimedia Gallery
The Discovery FilesA faster way to estimate uncertainty im AI-assisted decision-making could lead to safer outcomes
Increasingly, artificial intelligence systems known as deep learning neural networks are used to inform decisions vital to human health and safety. With support from NSF, researchers at MIT and Harvard have developed a quick way for a neural network to crunch data and output not just a prediction but also the network's confidence level in that prediction, based on the quality of the available data.
Credit: National Science Foundation/Karson Productions
Certain uncertainty.
I'm Bob Karson with the Discovery Files, new advances in science and engineering from NSF -- the U.S. National Science Foundation.
An artificial intelligence system -- AI -- can quickly analyze and sniff out patterns and assist in complex decision-making. As it gets better, we're relying on AI for more critical life-and-death applications like self-driving vehicles (Sound effect: honk) (Sound effect: heartbeat) even medical diagnoses. Which begs the question, when an AI system makes a decision, how confident is it that it's correct?
Many current methods for an AI system to estimate its own uncertainty can take time and lots of computing power. Now a new method can do that in a split-second. (Sound effect: race cars) Pretty important in, say, high-speed traffic.
With funding from NSF, researchers at MIT and Harvard came up with a way to have neural networks quickly crunch data and output not just a prediction, but the AI's level of confidence in that prediction -- based on the quality of the available data. The AI asks itself, "just how sure am I? How much should I trust my decision?"
The answer can be the difference between a self-driving car determining "it's clear to go through the intersection" or, "yeah, it's probably clear, but I'll stop just in case."
Estimating uncertainty. Building trust in AI by AI knowing when it shouldn't trust itself.
"The discovery files" covers projects funded by the government's National Science Foundation. Federally sponsored research -- brought to you, by you! Learn more at nsf.gov or on our podcast.
Images and other media in the National Science Foundation Multimedia Gallery are available for use in print and electronic material by NSF employees, members of the media, university staff, teachers and the general public. All media in the gallery are intended for personal, educational and nonprofit/non-commercial use only.
Images credited to the National Science Foundation, a federal agency, are in the public domain. The images were created by employees of the United States Government as part of their official duties or prepared by contractors as "works for hire" for NSF. You may freely use NSF-credited images and, at your discretion, credit NSF with a "Courtesy: National Science Foundation" notation.
Additional information about general usage can be found in Conditions.