DARPA has funded the creation of Singular’s chip because the fact it gets it wrong is jolly useful when comes to making sense of video or other messy real-world data.
A chip that can’t guarantee perfection can still get good results on many problems but needs fewer circuits and burns less energy, Bates says.
It solves problems when data has built-in noise from the real world, or where some approximation is needed, are the best fits. He has already seen good results from high-resolution radar imaging, extracting 3-D information from stereo photos, and deep learning.
It is also easier on the electricity. Singular’s approach was capable of processing frames almost 100 times faster than a conventional processor restricted to doing correct math—while using less than two per cent as much power.
The chip works alongisde a single conventional processor.
DARPA funded Singular’s chip as part of a program called Upside, which is aimed at inventing new, more efficient ways to process video footage. Military drones can collect vast quantities of video, but it can’t always be downloaded during flight, and the computer power needed to process it in the air would be too bulky.
Deb Roy, a professor at the MIT Media Lab and Twitter’s chief media scientist, says that recent trends in computing suggest approximate computing may be useful if you are processing any kind of data that is noisy by nature.