NVIDIA's Drive PX AI platform will be used in future autonomous vehicles from Toyota, according to a press release detailing an announcement made at NVIDIA's GPU Technology Conference. According to NVIDIA CEO Jensen Huang, NVIDIA and Toyota have been working together to build a powerful AI on the platform that will be able to integrate tightly with future Toyota vehicles, and possibly even help speed along the introduction of truly driverless cars to the auto market. The rig that will be running the software is set to be a custom-made variant of NVIDIA's very own Xavier supercomputer chip made for automobiles.
The powerful Xavier SoC kit that will be featured in the new vehicles is set to be small enough to fit in the hand, while laying down a blazing fast 30 trillion complex deep learning calculations every second. The bevy of sensors on each vehicle will give that extremely high calculation figure a run for its money; they will be taking in all environmental variables quickly enough to accurately process things in a matter of milliseconds, which essentially puts the amount of data being moved about well into the GBps area. While some self-driving car systems, such as Google's, rely on the cloud and remote neural networking a bit more, everything the Xavier-based Toyotas see, think, and do will happen onsite.
The powerful chip and advanced software ensure accuracy and smarts, but the system will get smarter over time thanks to automatic updates. This means that changes to local maps, fixes for issues, and improvements to the core program will all come straight from NVIDIA, and will hit consumers' vehicles automatically. NVIDIA nor Toyota announced exactly when this partnership will begin to actually bear fruit, or just how autonomous the very first cars it produces will be. Toyota stated that their goal is to move things along in regards to autonomous vehicles, and eventually see absolutely zero traffic fatalities. They have been working toward making cars more autonomous for 20 years, and it seems that they're looking to NVIDIA for help with the final step; cobbling together a powerful program for a fully autonomous vehicle, and eventually getting that system into consumer vehicles.