Engineering news
"We're using optics to condense the normal setup of cameras, sensors and deep neural networks into a single piece of thin glass," says UW-Madison electrical and computer engineering professor Zongfu Yu.
AI is a resource-hungry process that uses significant computational and battery power every time you, for example, unlock your phone with face ID. Yu’s team envision using pieces of glass with strategically placed bubbles and impurities that would bend light in specific ways to differentiate between different images.
The engineers created a proof of concept that can recognise hand-written numbers. Light reflecting an image of the number enters one end of the glass, and then focuses to one of nine specific spots on the other side, corresponding to the individual digits. The glass could tell, in real-time, when a number 3 was altered to become an 8. "The fact that we were able to get this complex behaviour with such a simple structure was really something," says Erfan Khoram, a graduate student in Yu's lab.
Yu likened the process to ‘training’ a machine-learning algorithm, but with an analogue material instead of code. The engineers place air bubbles of different sizes and shapes, as well as small pieces of light-absorbing materials, at different, specific locations inside the glass. "We're accustomed to digital computing, but this has broadened our view," says Yu. "The wave dynamics of light propagation provide a new way to perform analogue artificial neural computing."
There are many potential benefits of such as a system over current forms of image recognition, as its passive nature means it could be used hundreds of thousands of times. "We could potentially use the glass as a biometric lock, tuned to recognise only one person's face," says Yu. "Once built, it would last forever without needing power or internet, meaning it could keep something safe for you even after thousands of years."
And – although the training process is time consuming, once done, the process literally works at the speed of light. The researchers are planning to scale up the complexity of the tasks that this smart glass can handle. "The true power of this technology lies in its ability to handle much more complex classification tasks instantly without any energy consumption," says Ming Yuan, a collaborator on the research and professor of statistics at Columbia University. "These tasks are the key to create artificial intelligence: to teach driverless cars to recognise a traffic signal, to enable voice control in consumer devices, among numerous other examples."