© 2024 KRWG
Play Live Radio
Next Up:
0:00 0:00
Available On Air Stations

Super fast computing with analog

Fast computer vision – from Nature, 2 Nov. 23, pp. 48 ff.

Self-driving cars, robots, and medical diagnosis can all benefit from fast and accurate analysis of images. Think of a self-driving car taking images with, say, 8 cameras, each image with millions of pixels changing many times each second so that a steering or braking response can be taken in a few hundredths of a second. This requires recognizing complex features – roadways, people, other cars, road signs – that are very complicated to compute from the pixels.

Some developers of computer vision such as at Tesla use innovative programs in digital computers. Yitong Chen and 11 colleagues at Tsinghua University in China have now made a very fast computer chip that uses analog computing, where variables take continuous values, not 0 or 1.

Analog was used in the earliest electronic computers but almost universally fell out of favor, being susceptible to cumulative errors such as rounding. However, Chen’s group has very powerful ways to structure the computing, using the diffraction of light right on the chip to extract patterns. Comparing this to digital computing, they claim the equivalent of 4.6 quadrillion operations per second with very little electrical power. Your modern laptop with an 8-core CPU might hit 100 billion operations per second, slower by a factor of about 460,000. Then again, Chen’s computer can’t keep your recipes or even design another chip.

 This has been an outreach activity of the Las Cruces Academy, viewable at GreatSchools.org.



Vince grew up in the Chicago suburb of Berwyn. He has enjoyed a long career in science, starting in chemistry and physics and moving through plant physiology, ecology, remote sensing, and agronomy.
Related Content
  • KRWG explores the world of science every week with Vince Gutschick, Chair of the Board, Las Cruces Academy lascrucesacademy.org and New Mexico State University Professor Emeritus, Biology.