The US Food and Drug Administration (FDA) has already approved at least one product that uses artificial intelligence (AI) to analyze mammograms for signs of breast cancer, while European regulators have approved a different system that uses AI to spot lung cancer.
Clinical trials indicate that the mammogram software — acting entirely on its own — spots breast cancer about as well as a typical radiologist, but humans and software working together perform better than either of them working alone.
“Cancers can grow for some time before even the best eyes can see them. AI, as it exists right now, can assist radiologists in detecting some cancers and that has the potential to save lives,” said Paul Weber, associate dean for continuing education at both Robert Wood Johnson Medical School and Rutgers New Jersey Medical School.
Weber will outline the amazing progress of medical AI on Tuesday, February 25, and his presentation will be just one part of a larger AI event that runs from 10 a.m. to 4 p.m. at the Busch Student Center at Rutgers University–New Brunswick.
Different sessions will explain what AI can do today and what it will likely do in the next few years, not only in medicine, but also in other vital fields such as education and security.
A session on campus security, for example, will explore the potential of using AI to detect crime as it happens by analyzing the data generated by the thousands of connected devices on any campus.
Many intriguing applications are under development, but several are already available. Existing software can monitor hundreds of video feeds simultaneously, detect the sorts of quick, aggressive movements that are typical of physical assault, and immediately ask humans to review suspicious clips. AI can also be taught to use video feeds to detect people who are carrying guns.
AI has made less of an impact in education than it has in security, but, as at least two of the sessions at Tuesday’s event will demonstrate, many leading AI researchers from both academia and corporate America are working to change that. The first of those sessions will feature Google’s thoughts on how AI will be used in both education and industry over the next five years. The second will provide a more personal view from a leading AI researcher at Rutgers University–Newark: Patrick Shafto.
Shafto is working to build AI that can improve new-language instruction by teaching all learners the way parents intuitively teach their children, by starting with simplified sounds that convey general meaning and adding nuance as learners improve. Shafto and several colleagues have already published research that used mathematical models to show how “baby talk” creates understandable speech more efficiently than standard English. He’s now working on software that can use simplified speech as a teaching tool.
“The goal of language is to communicate, so the goal of language instruction should be to teach the sounds that are most vital to communication first,” said Shafto, an associate professor of mathematics and computer science.
Shafto’s presentation, which will feature several demonstrations, will also touch upon another important area of his research: creating AI that can explain to users why it reaches any particular conclusion.
“Most of today’s state-of-the-art AI models are entirely opaque. No one, not even the people who build them, can really understand why the act the way they act in any given situation,” Shafto said. “We still use these models because they perform well, but they do make mistakes, and we want to be able to figure out what caused them to go wrong and how to fix them.”
The AI event, which is free to all, will take place in the multipurpose room and the center hall at Busch Student Center. It will feature speakers from both academia and large companies such as Google, Dell, NVIDIA, Intel and event-sponsor CDW-G. Some sessions will be reasonably technical but most will be aimed at a general audience. For more information visit the official event page or the event’s Facebook page.
Tags: IT community