One of the most promising applications for Artificial Intelligence is driverless transportation. Self-driving cars - once the exclusive domain of big companies like Google - are now facing competition from startups. One of the hottest is Silicon Valley's Drive AI, recently named as CB Insights most promising AI startups. Mark Niu goes for a ride.

Mark "Okay I hope in with anticipation, for my driver today, has a hands-off approach.

Mark "What's gonna happen here?"

Tao Wang: "I press this button and it will go on its own."

Mark: "That's pretty much it."

Tao Wang: "Yep."

In the parking lot, the van starts off conservatively. But once it hits the open road you might say it has a free-wheeling spirit, kind of like co-founder Tao Wang.

Mark :"Do you ever get nervous taking out this car?"

Tao: "Not really."

Mark: "First time you must have been nervous."

Tao: "A little bit. But It's very fun after you program the vehicle yourself and see it working."

Wang - from Wuxi, China - founded Drive.ai with fellow students from a Deep Learning research group at Stanford University.

TAO WANG, CO-FOUNDER DRIVE.AI "In a lot of cases machines are more reliable than humans because humans get distracted, get tired when they are driving and that's when most of the accidents happen. But there are very hard challenges building an AI system that can drive very reliably."

Drive.ai is building a complete solution - both software and hardware - that can turn almost any vehicle into a self-driving car. They're focusing on business fleets that can have pre-programmed route-though the cars still have to be able to adapt to whatever comes its way. In 2017, Drive.ai's vehicles traveled more than 10,500 kilometers in autonomous mode and had to disengage 151 times to allow a human driver to take over.

Tao Wang: "Oncoming traffic still has a little problem dealing with that. Still working on that."

TAO WANG, CO-FOUNDER DRIVE.AI "Actually every single disengage is like a golden learning experience for us."

Mark: "What's the leading cause for disengaging from self-driving mode?"

Tao Wang, Co-founder, Drive.ai: "I think one of the major contributors are behaviors of other agents on the road, like pedestrians behaving in a weird way or cars cutting us off really aggressively."

All the data collected - and the camera views from every angle, of each journey - can be dissected later. Visualization projects leader Pat Marion shows us how he can program adjustments and replay simulations to see if it works going forward.

Pat: "The blue path is our navigation route where we wanna go. But in case there's a parked car so what we're gonna wanna do is deviate a little from the normal path here and put a little more room for the parked car."

Wang says one of the ongoing challenges for any self-driving vehicle is knowing, like humans, when it's okay to break a rule, for example, in emergency situations. And while acknowledging that the technology will cut the number of professional driving jobs, Wang believes increased safety and less traffic will help create a more productive economy. Mark Niu, CGTN, Mountain View, California.