June 6, 2023


Complete Canadian News World

Robots will gain a vision similar to that of humans.  How can we use this?

Robots will gain a vision similar to that of humans. How can we use this?

Artificial intelligence has huge potential, but in many respects it is not yet able to match humans. One such aspect so far is sight.

That’s why scientists at the University of Central Florida designed a device that mimics the retina. In the long term, this should lead to the emergence of systems that use artificial intelligence, for example, for facial recognition or control of self-driving cars.

Read also: How will optical illusions help create the next generation of artificial intelligence?

Research team members describe their project in a format ACS nano. According to them, such a system works better than a biological eye, due to the range of wavelengths it can perceive. As if this were not enough, the proposed solution combines three phases into one: data collection, processing and remembering. Until now, each had to be implemented separately.

However, we will have to wait a little longer for the practical use of this technology. According to the authors, this should happen in the next five to ten years. Interestingly, the same team has previously created brain-like devices that could enable AI to operate in space, for example. Tanya Roy, one of the authors, explains that thanks to the ability to perceive images, scientists have clamp-like devices that act like “smart pixels” in a camera.

If you’re in your self-driving car at night and the car’s imaging system only works at a certain wavelength, such as the visible wavelength, it won’t see what’s in front of it. But in this case, with our devices, he can actually see his surroundings in all their glory.Mullah Mangur Islam, University of Central Florida

Read also: An hour on the way. This is enough for the robot to learn to walk

See also  a test. These tasks are terrifying in high school math. Can you handle the possibility?

Scientists tested the device for accuracy. You were supposed to “see” mixed images: the number 3 in the ultraviolet and a three-fold inverted and interconnected image of the number 8 in the infrared. It turned out that the system coped with this task by distinguishing two numbers, considering them 3 in the ultraviolet and 8 in the infrared, respectively. Accuracy in this range was about 70-80%.