How close are we to an autonomous future?


If Elon Musk and the people at Uber are to be believed, we’ve seen the future of transit. Advanced computer systems will, they tell us, free everyone from the constraints and dangers of operating a vehicle. The transportation we’ve imagined usually involves an electronic brain attached to sensors, giving the rider a hands-free experience while making the road safer for everyone. Since 2014, Tesla, Google, Apple, and Uber (among many other others) have all been actively trying to make that future happen. But a recent fatality involving an autonomous Uber vehicle reminds us that, though this technology has progressed quickly, there is still a long way to go before we fulfill our dreams of an automated future.

A leap in imagination

According to our cultural imagination, automation frees human beings from otherwise impossibly complex or time-consuming tasks. In the future, science-fiction posits, super-intelligent computers will give us seamless transit, analytics, and basically anything we tell them to do. Though scientific research and theory have helped spur imaginations, it’s important to remember that some of our ideas of automation and the future arise from the needs of writers to advance a plot. Solutions in the real world generally do not work so smoothly or efficiently.

So how close are we to fulfilling our dreams for an automated future? Are companies like Uber getting us closer or are they looking for the wrong solution? A good place to start might be in the examination of common traits of autonomous vehicles in fiction, and how closely they match today’s reality.

1. Keyless or Biometric Entry

A simple fingerprint, retina, or other biometric scan frees drivers from the inconvenience of lost keys in the future. Today, fingerprint scanning is probably the closest match, though it is not currently widespread on car doors. A closer contender might be our smartphones, which are used by rental companies, like ZipCar, to unlock shared vehicles.

2. High-efficiency and speed

Tesla’s battery-powered vehicles seem to be leading the way towards the fuel economy we expect for futuristic vehicles. The newest long-range battery packs are said to travel around 300 miles on a single charge. Though 300 miles is nothing to shrug at, the 9.5 hour charge time doesn’t make for the fastest long-distance travel...especially if you’re cruising along over the speed limit.

3. Detailed, accessible, systems analysis

Cars and computers have become increasingly intertwined over the last 20 years. OBD ports are generally standard and give those with complementary technology a view into the system—helping us decipher what warning lights are trying to tell us. As computers regulate more of the mechanical aspects of cars, the amount and clarity of information available through those ports and onboard displays has increased, giving us a real time view of fuel economy, tire pressure, and other variables. However, we’re ultimately a long way off from replacing a mechanic's diagnosis with AI.

4. Multi-modal control interface (vocal, touch, eye tracking)

Onboard computers that are hands-free are not truly hands-free unless they can be controlled entirely without hands. Can Alexa or Siri be trusted to take the wheel in our stead? We all know the answer is a hard “no.” Perhaps, one day, we’ll have our neurons hooked up to our vehicles. Then again, our human potential for daydreaming might mean our hands are best kept on the wheel.

5. Automated acceleration, braking, and navigation

Automated driver systems are much more than simply following a GPS map. Autonomous cars need detailed virtual models of the world around them to make decisions. Speed limit, traffic directives, lane closures, and (most importantly) people need to be “visible” to the computer. Replicating the advanced spacial awareness of human beings is still a long way away, but the use of LIDAR mapping has gotten us close. Though these systems are able to perform essential driving functions well on highways, the complex environments and snap decisions of city driving are still far beyond their capacity as evidenced by the recent tragedy in Arizona.

Teaching the Car to Drive

With our five points of comparison in mind, the cars of our imagined future do not seem entirely out of reach, but the steps we must take to fully realize them are massive, complicated, and limited by current capabilities. In science-fiction, artificial intelligence is used recursively to create programs designed for humans. Those tasks, today, take time and a huge scope of understanding. Rather than using a computer generate code and layout, developing software requires a team of developers and designers (like us!) and a large chunk of time. Even if a computer was made to generate sections of code which deliver particular results, we’re a ways away from teaching computers to truly understand the purpose and functional requirements for what they are producing. Computers simply do not have the context yet. Until we can truly produce computers that think for themselves, we probably won’t have the tools to make “real” autonomous vehicles.

Uber’s self-driving car fatality, then, feels like the growing pains of an emerging technology. Like the early days of consumer automobiles, the technology is still in development and prone to unpredictability. Even with a human behind the wheel of the car responsible for the fatality, putting a computer that only partially knows what it’s doing in charge is dangerous. Of course, there are also plenty of self-driving vehicles which have not killed anyone. There are also plenty of tasks which automated systems may perform with near perfection—think assisted parallel parking.

Is any of this reason for ceasing development into autonomous vehicles? No. However, the recent fatality should remind us to be wary of new technologies and remember that revolutionary developments rarely work out of the box. For Uber, that means it’s back to the controlled test-drive environment. For everyone else, it means we can expect humans to remain behind the wheel for at least another few years.