A New Player in Self-Driving Cars

A chilling video spread across the web this week, showing a carload of teenagers getting hit by a freight train in Kentucky.

The young driver attempted to cross a set of railroad tracks before an oncoming train passed but did not make it. The train struck the car, killing two passengers and pushing the broken husk of the white Toyota more than a mile down the tracks.

A quick decision to hit the gas and save a couple seconds resulted in tragedy.

This accident was unusual, but not that unusual.

More than 30,000 people died in car accidents in 2013. A further 2.3 million were injured. In total, there are more than 10 million car accidents every year.

If more people used seat belts, nearly 2,000 of those lives could have been saved, and more than 20,000 injuries would have been avoided.

But safety measures do not prevent the accidents from happening — they just lessen the damage we sustain.

Computers, on the other hand, can prevent the accidents.

If computers took over the task of driving for everyone, the majority of these accidents would not even occur.

This is one of the main reasons why self-driving cars are such an exciting tech development. They will help eliminate human error from one of the most dangerously error-prone activities we take part in.

In the wake of that horrific Kentucky crash, popular chipmaker NVIDIA (NASDAQ: NVDA) has debuted a new computer to help cars drive themselves.

The Platform

Called Drive PX, NVIDIA has built this mobile vision system around its Tegra X1 8-core mobile processor. Because this chip is heavily geared toward image processing, the PX system can handle up to twelve 2-megapixel cameras running simultaneously at 60 frames per second.

Computer vision systems translate video input into actionable data, and they’re going to contribute significantly to the next generation of computers. We’re seeing it not only in self-driving cars but also in industrial applications such as quality assurance and process monitoring.

NVIDIA uses these cameras to watch the road not only ahead but also behind, beside, and even inside into the driver’s seat.

The system can tell the difference between vehicle types, identify obstacles, and automate parking. It’s a heavy-duty machine-learning platform that makes neural connections as new data is introduced.

NVIDIA has already used it to classify more than 68,000 different objects, and that amount will snowball as the platform is put to use in more cars.

With a price tag of $10,000, NVIDIA’s Drive PX platform will be available in May of 2015. This first version is targeting developers, so it’s unclear where it would turn up first.

However, Tesla (NASDAQ: TSLA) CEO Elon Musk appeared on stage at NVIDIA’s announcement of the platform this week, so the electric car company might be one of the first to look for.

Stock in NVIDIA was trading up 1% pre-market based upon this announcement.

For the last two years, the company’s shares have been gently climbing. Their all-time high was $37.39 back in 2007, and they’re currently trading in the mid-$20s.

NVIDIA’s focus on mobile systems-on-a-chip was perfectly timed to match the growth of mass market smartphones over the last five years, but this has taken place somewhat outside of the company’s area of expertise. Not enough to rely on graphics alone, mobile systems-on-a-chip include applications processors and baseband signal processors.

These current endeavors in machine learning might seem a bit on the fringe as well, but they stay in the company’s wheelhouse of graphical processing and are again timed very well for a coming market disruption.

NVIDIA is a solid addition to our self-driving car portfolio.

Good Investing,

Tim Conneally

 @TimConneally on Twitter

For the last seven years, Tim Conneally has covered the world of mobile and wireless technology, enterprise software, network hardware, and next generation consumer technology. Tim has previously written for long-running software news outlet Betanews and for financial media powerhouse Forbes.