Oct. 1, 2018 — Professor Seth Chandler described how neural networks, a technology associated with artificial intelligence, can be used to make accurate predictions in a presentation last week at the University of Houston Law Center.
The talk titled "What the heck is a neural network anyway?" was sponsored by the Intellectual Property Student Organization.
"Artificial intelligence or machine learning is already running our lives to some extent," Chandler said. "It is likely to increase in power over the coming years."
Chandler, the Law Foundation Professor of Law, defined neural networks as "universal function emulators." He said neural network's are accurate and effective because they rely on the same technology found in computer chips originally intended to enhance the market of computer gaming. Chandler cited Siri, a voice command and recognition function on iPhones, as an example of a neural network.
"When you talk to Siri, Siri is processing your language through a neural network to try and figure out its meaning," he said.
Chandler provided a visual explanation of how neural networks operate contrasted their flexibility and resistance to "overfitting" with more traditional statistical tools such as regression. As such they can be used for divergent tasks such as predicting how likely it is for someone to recidivate if they are let out on parole, generating artificial examples of human voices or human faces, and engaging in transformation of data from one representation to another.
"Neural networks are very useful in predicting things that we haven't seen before," Chandler said. "Basically they extrapolate to things we have not seen before and to make good predictions about what's going to happen."