Autonomous cars are fast becoming a reality — and they are getting smarter each and every time they take to the road. This is thanks to deep structured learning algorithms that allow self-steering vehicles to assess, react and adapt to changing road conditions and file the memory away for later. As demonstrated in this Nvidia Drive PX video clip, they are also proficient at identifying individual vehicle types — including cop cars.
Drive PX is one of the industry's leading self-driving car platforms. It's powered by two Tegra X1 processors and comes with 10GB of DRAM; enough to handle data from up to 12 cameras simultaneously. Drive PX forms the brains behind numerous self-steering vehicles, including the Audi A7 and Tesla Model S electric car.
At GTC 2015, we were given a demonstration of how Drive PX "sees" nearby vehicles and assesses the different shapes and sizes in real time. Interestingly, this included police cars which the platform was able to correctly separate from other SUVs and sedans and identify.
This got us wondering about how a self-driving car should react on the inevitable day that a cop attempts to pull one over. Presumably, teaching it to stop safely in response to a police siren wouldn't be too difficult — but this just creates its own set of problems.
For instance, nefarious types could trick your car into pulling over, even though they aren't police. These are the kinds of hacks and exploits that car manufacturers will need to overcome before the technology becomes mainstream. In any event, interesting times are definitely ahead!
Gizmodo travelled to GTC 2015 in San Jose, California as a guest of Nvidia.