The Dawn of DriveNet: Peering into the Future with Google, Uber and Facebook.
The future of transportation was coming at us fast last week in cryptic headlines. First, news broke that Google was “building their own Uber, ” a taxi hailing service to augment their famous self-driving car program. Almost in response, Uber announced the opening of a research center at Carnegie Melon University to develop self-driving cars, hiring 50 senior scientists and cleaning out the talent pool at the Robotics Institute that designed the Mars Rover.
Later, the story changed. Google was said to only be testing a carpool service for employees, an app that ”isn’t associated with the company’s driverless car program.”
Are these moves by the titans of transportation technology a coincidence? Unlikely.
Google, like Uber, is building toward an ambitious future: efficient cities served by universal broadband with on-demand driverless cars and delivery networks. To make the vision work at scale, a successful driverless car is only part of the solution. It will need to be combined with an algorithmic ride dispatch platform and, ultimately, with a ubiquitous social network like Facebook.
A service like this — DriveNet — will lead to fundamental changes in our society: the end of individual car ownership, enabled by a hyper-intelligent algorithm that predicts and provides for our daily transportation needs.
The Driverless Car
The autonomous driverless car is wondrous. And inspiring. Which is why Google has taken it on. When perfected — and it has a long way to go — it will make millions more productive, letting them work safely during their commute. It will also greatly reduce highway fatalities by letting tired, drunk or inexperienced drivers sit back and give up the wheel.
But how will these driverless cars fit into our lives? Will they be purchased by individuals, to be parked in their driveways to serve their specific needs, just as cars are today? Probably not. Early movies looked a lot like the stage plays that preceded them. And early TV programs sounded a lot like radio. Its often hard for us to make the leap of imagination, so we keep thinking of the new thing as just the old thing, improved.
The endgame: the end of individual car ownership, enabled by a hyper-intelligent algorithm that predicts and provides for our daily transportation needs.
But if we do make that leap, its likely that our driverless cars will not be parked in driveways but be available in a whole new way — shared via an on-demand service.
Google is already thinking along those lines. Chris Urmson, the head of Google’s Driverless Car program, shared:
“We’re thinking a lot about how .. this might become useful in people’s lives. One is in the direction of the shared vehicle. The technology would be such that you can call up the vehicle and tell it where to go and then have it take you there.”
They made a great video of the first test with real people: http://youtu.be/CqSDWoAhvLU
The idea is that pools of autonomous cars roam the streets and move between neighborhoods to anticipate and respond to demand. They are hailed via phones, arrive within a few minutes, and after dropping you off move on seamlessly to the next passenger. Congestion and the need for parking in crowded cites is eliminated. Costs drop dramatically (.50 cent taxi rides?) as no drivers are needed to maintain the service.
Moving to such a system will deliver a windfall of savings to our society. In a recent study called “Transforming Personal Mobility,” a Columbia University think tank looked at the costs of driving in Ann Arbor, MI where 120,000 privately owned cars are used by daily commuters. It concluded that all these private cars can be replaced with a shared fleet of just 18,000 autonomous cars (an 85% reduction!) with the net effect of better mobility experiences at radically lower cost to consumers
and society.
Going further, a PWC report predicted that, in the transition to shared autonomous cars, an astonishing 99% of all cars on the road can disappear without much impact. Targeted savings: the $2 trillion spent per year in aggregate in the U.S. on car-related activities, with tsunami-like impact on the auto industry.
The Self-Driving Carpool
Achieving these levels of efficiency will be impossible if, as we have today, each car on the road is carrying only one passenger. If cities are to reduce congestion, keep costs down and drastically reduce the number of cars on the road, the car dispatch service will need to accommodate multiple passengers per trip, matching people up for carpooling together.
And while this is happening on a limited basis today, with services like UberPool and LyftLine, people-matching in the driverless car era will need to get even more intelligent. Specifically, it will need to address the safety concerns and social compatibility of the people sharing the ride.
Today, the people who get in your car through UberPool are random strangers. At Hovee, we’ve learned that sharing a peer-to-peer ride with someone is an intimate experience. And trust is important. In an autonomous car, with just the two of you in the car without a driver, this applies even more. Before you share the ride, you want to be confident about who you are getting in a car with.
This is where Facebook, or other social networks come in. The dispatch algorithms will not only calculate the most efficient carpool route connecting passengers, but also look at the social profiles of the people involved to see if they are really compatible. So friends (Facebook and RL) could be given preference, as well as those who work for the same company or like the same music or movies. If the match is done well, the shared ride can be an opportunity to meet new people, or catch up with neighbors traveling in the same direction. Someone you meet on a shared ride might turn into a friend, a work connections or even your soulmate life partner.
DriveNet Dispatch
Today Uber and other taxi-hailing services deploy sophisticated algorithms and teams of analysts to allocate cars and maximize driver utilization. These algorithms use the flood of data from the phones of drivers and passengers to constantly test assumptions, learn and improve. For example, Uber sees every time you open your App, even if you don’t request a car, and registers an indicator of demand. The system analyzes these signals and manages the inventory of available cars. It uses techniques like demand prediction, supply positioning (sending cars proactively to expected areas of high demand) and surge pricing to balance supply and demand across localities and time periods.
Dispatch is imperfect because today humans are still involved. Uber has teams of former Goldman Sachs i-bankers sitting in front of spreadsheets, balancing supply & demand, and social and personal preferences are not part of the equation.
In a fully scalable Google autonomous-car world, dispatch will need be fully automated, and be intelligent not only about traffic patterns, but also about social compatibility and individual usage patterns.
Managing transportation on this scale is how DriveNet will be born as an autonomous, predictive system. And because of the scale, and the deep domain knowledge required, the ideal group to build DriveNet is a partnership of the big players: Uber, Google and Facebook.
With deep predictive capability for both supply and demand, autonomous cars will be guided by increasingly sophisticated algorithms that will know your social connections and personal preferences, and can plan based on knowledge of when you, your friends and co-workers go to school, to work or to social activities.
So as you walk out your front door one morning in the future, on the way to a work meeting, you will count on DriveNet to send a shared driverless car, where, as you settle into the plush seats, you find the other people in the cabin share your interests, have friends in common and keep your ride fun and blissfully stress free.
That is a future I would trust to an algorithmic intelligence.
Paul Kogan is the CEO of Hovee, a social ride-matching service in San Francisco
