The bold promise of autonomous vehicles has collided with a sobering reality: turning driverless cars loose on every road will take more time and more money than originally expected. However, driver-assisted artificial intelligence (AI) technologies will see rapid adoption. Rather than replacing drivers completely, AI will be used to make driving easier, safer and more enjoyable.
Person using a car in autopilot mode hands free
The future of autonomous cars will be human plus machine, not human versus machine, with AI complementing human judgment instead of acting as a complete substitute for humans.
The pattern that underlies the commercialization of driverless cars is something I call the “Myth of Substitution.” There is no shortage of audacious predictions about driverless cars, such as those made by Tesla CEO Elon Musk. But reality has deflated those promises. In one of the latest developments, General Motors’ Cruise unit announced that it was delaying commercial deployment of its autonomous cars because more testing is required. Cruise’s postponement comes as other autonomous vehicle companies and automakers have reached similar conclusions about how long it will take for autonomous cars to be used on public roads without restrictions. This has triggered disappointment, with some comparing the vision of driverless cars to a mirage.
In the Myth of Substitution, when a new technology emerges, the widespread hope is that it will perfectly substitute the technologies that are currently in use. However, this rarely turns out to be true. For example, television was initially thought to be a substitute for radio. While families who gathered around the radio in their living rooms switched their focal point to TV, radio did not die. It became a complement to TV as entertainment on the go with drive-time radio. The microwave was originally positioned as a substitute for stovetop cooking, yet it is hard to imagine getting by with only one or the other. Digital marketing was predicted to be the victor over other channels, including traditional television ads. Yet, Super Bowl advertising is as healthy as ever.
Twenty years ago, the buzzword was “pureplay” for digital companies, with the expectation that all things digital would overtake bricks-and-mortar. But as I observed then, and continue to see now, customers do not think of themselves as “online” or “offline” customers. Retailers are realizing that customers are best served with an omnichannel experience, as we’ve seen with Amazon opening bricks-and-mortar stores, and cosmetics retailer Glossier launching storefronts in major cities.
Essentially, technology substitution is not an either-or-world; the truth lies in between. Hybrid solutions should not be viewed as disappointing, but a true advancement. In fact, a defining aspect of technology is the role that it plays in extending human capabilities. AI will be most useful in helping humans to drive safer, but true driverless autonomy remains a distant and difficult goal.
The Real “Autonomous” Expectation
To understand how pragmatic hybrid driver-assistance models may evolve, it is useful to point out that the “driverless car” is a poor label for advancement in autonomous vehicle technology. There are five levels of autonomy in vehicles. The first three fall into the category of advanced driver assistance systems (ADAS). At Level 1, the driver maintains control over the car, but some driver-assistance features are incorporated into the vehicle design. At autonomy Level 2, technology can control functions such as acceleration and steering, but the driver remains in control. This is the level that Tesla has been at for the past three years; the vehicle steers, accelerates and brakes autonomously, but the driver needs to constant monitoring the vehicle. Level 3 is conditional autonomy, in which technology monitors the environment, but the driver must be ready to take control of the vehicle.
At Levels 4 and 5, there is a fundamental shift from driver assistance to driver substitution through automated driving systems (ADS). Level 4 is “high automation,” in which the vehicle is capable of handling all driving functions under certain conditions. An example is what Walmart is piloting in Bentonville, Arkansas, allowing an autonomous vehicle to move customer orders along a two-mile route between two stores. Level 5 is fully autonomous; the vehicle is capable of handling driving under all conditions. (Even at Levels 4 and 5, the driver may be able to take control of the vehicle, including a “kill switch” that could make it acceptable to regulators and consumers who fear letting go of the wheel.)
While people are seduced by the sexiness of Level 4 and Level 5 autonomy, ADS that substitutes technology for a human driver isn’t coming anytime soon. This level of full autonomy is neither pragmatic nor technically possible now in terms of infrastructure and safety. Rather than get caught up in the delusion of full autonomy, transportation providers would be better served by focusing on creating smarter driver-assistance systems with progressively advanced capabilities.
The trade-off between cost and the value of autonomous technology will probably be optimized at Level 3 or 4 for several years. In the short-term, that creates many opportunities to make driving safer with technology that warns drivers or directly takes action. Longer term, we’ll likely see more Level 4 autonomy in certain settings, such as cars that don’t just drive autonomously on the highway, they also enter and exit highways autonomously, with the driver taking control on the side streets. We might see Level 4 autonomy in specific use cases like “platooning” of trucks to reduce road space requirements and wind resistance.
When looking at advancements, we cannot be distracted by the hype of black-and-white thinking that it’s either driver-controlled or driverless. The truth is that autonomous vehicles will evolve, featuring humans and machines collaborating to create safer, cheaper, and more enjoyable driving solutions. When AI collaborates with humans, it can create more cost-effective outcomes than when it tries to become human.
In the next article, we’ll look at another myth in the development of autonomous vehicles – the Myth of the Mass Market.
