For that matter, none of his counterparts, back then, could. Deep learning as a potent tool of detection of nuclear testing and activity – no matter how silent – was, after all, not around some years back.
But today when researchers are pleading for more bad data (so that there are enough signals, even if bad, in the data stream for the machine to learn), this unique set of ears can tell a lot, and much early.
Deep neural networks are the new set of toys and challenges for scientists at the Department of Energy’s Pacific Northwest National Laboratory. Powered with deep learning, where machines can learn and make decisions without being explicitly programmed for all conditions, researchers can sniff radioactive decays with a new precision (yes, it has been shown it can sort 99.9 percent of the pulses correctly) and speed. An underground nuclear testing, hence, can not escape its radar – thanks to these new intelligent machines.
This becomes very interesting in light of the current nuclear talks and tensions pulsating around the globe. This week itself North Korea claimed demolition of a nuclear test site. The concerned secret location, Punggye-ri since 2006, was reportedly only seen from satellite photos so far.
Satellite eyes have got new ears now, though, with the advent of AI in this space.