Enlarge (credit: Aurich Lawson)

The self-driving car reached something of a watershed in 2016. All of a sudden, it seems we're right on the threshold of autonomous vehicles transforming our transportation. I'd ask any engineer in 2015 when they thought we'd have full (i.e., SAE level 5) self-driving vehicles and the answer was always "ooh, that's another 20-30 years away." Fast forward a year, and all of a sudden that target has moved: BMW, Intel, Mobileye, Uber, Volvo, Ford, Delphi, and others have all set 2021 (or earlier in the case of Delphi) as the year by which steering wheels become optional.

The big breakthrough is down to the use of machine learning and deep neural networks, a field that has come along leaps and bounds in a relatively short space of time. For example, Nvidia will sell OEMs and tier one suppliers an open AI platform for automotive uses that leverages the company's GPUs and machine learning—tech that it ably demonstrated with its BB-8 technology demonstrator:

Nvidia's deep neural network allowed BB-8 to figure out how to drive on and off road in a short amount of time.

But along the way, the automakers and tech firms working on the problem have diverged into two groups: the ones who plan to get there in an incremental, stepwise fashion, and the others who plan to skip the intermediate step.

Read 12 remaining paragraphs | Comments