Tech Talk: Considerable Tests Still Needed For A.I. Drivers

Advertisement
Advertisement

Autonomous Vehicles may be well on their way to a street near you – pending much-needed testing – but there are still several problems to be confronted first. There's no guarantee that they will always function as they're supposed to and any accidents will need to be somebody's responsibility, whether the manufacturer's or the individual who was inside when it happened. Beyond ethical dilemmas, there are also problems with what laws or regulations will be enforced, how those will apply across country borders, let alone from city to city or state to state. Pollution, in association with smart cars, could improve or it could get worse. There are issues that will need to be addressed with consideration for state and city budgets, income, and taxation. Finally, infrastructure and components are making steady progress but those can be adversely affected by weather, malicious interference, and other circumstances. All of these problems, setting aside trust issues people still have with self-driven cars, will need to be solved before autonomous cars and trucks become mainstream. The issues also point to the importance of transition between licensed human drivers and computer-driven systems.

First, it needs to be said that self-driving cars are coming, ready or not. Manufacturing, real-world testing, and planned testing have saturated the auto industry, news surrounding technology and automobiles, and nearly all other groups with ties to those things. The technology presents unique problems but could also solve a huge number of problems. Although some have posited that they could result in an increased overall number of vehicles on the road and, by proxy, more pollution, there is a very real possibility that emissions will be drastically cut. So could travel times and the number of accidents caused by faults of human drivers. Self-driven vehicles should also increase the overall mobility of everybody, wherever they operate, especially in the low-income sector – thanks to likely innovations from Uber, Lyft, Google, and others in the ride-sharing industry. At this point, there is almost nothing that will prevent algorithm-driven vehicles from making their debut, for better or worse.

The ethics implications of self-driving automotive can be difficult to examine in any detail, mostly because they are primarily philosophical in nature. They can be mostly summed up with the question, "how should a driverless car react in a circumstance where there is no alternative to an injury to either the driver, pedestrians, or others?" However, the question runs much deeper than that, because cars and trucks are not the only advances approaching a point where A.I. is the sole driver. Many areas of the world are testing A.I. driven contraptions in a wide array of shapes and sizes, intent on finding a way to deliver packages, mail, food, and other goods with ever-improving efficiency. That means there is an increased risk to people, goods, and equipment with every new discovery for how self-driven systems can be used. Tied to the ethical implications is the possibility of injuries themselves.

Advertisement

Most incidents have been avoided so far, with regard to accidents caused by A.I. drivers. In cases where accidents have occurred, they have mostly been blamed on individual drivers' failures to respond to onboard system and take back control. However, some have been prevented by the drivers themselves, too. There will be an estimated 10 million autonomous vehicles on the road by 2025, and many companies plan to have their vehicles operating at near peak performance – as far as current technology and components are concerned – by that time. With that said, the systems need to get a lot better before they become usable everyday. Some have predicted that as many as hundreds of billions of miles of testing need to happen before the vehicles can match human drivers and surpass them. That number is a long way away from the 636,000 miles Google's self-drivers moved over the past year in California. Before autonomous cars can take the place of human drivers, they will need to learn traffic laws, how to discern between objects, animals, and other vehicles on the fly, and how to switch the rules up under extraneous circumstances or when switching locales. Perhaps more importantly, the self-driven vehicles will need to be able to cope with much less predictable human drivers.

Making matters worse, there are other implications beyond those that will still need to be worked out. For example, Uber, Lyft, and others have built a business model centered around taxi-like ride sharing, but have also invested millions into self-driving systems. Those drivers will either have to find new work, or industry leaders will need to find new and innovative ways to improve on human-based services. Other jobs that are heavily dependent on driving skills will also be affected and it isn't only those employees who stand to lose money or livelihood, either. Cities and States operate on the income created by taxation and fees. The nature of non-person drivers will almost certainly eliminate a substantial chunk of revenue created by fees. While taxation could make up some of that, people would likely not be very happy with that as a solution. Another example is the infrastructure. Supporting networks, hardware, and other components that are needed to generate an A.I.'s awareness and comprehension of its surroundings will have to improve to cope with changing conditions. Furthermore, they will need to be able to work, at least at a functional level, when conditions prevent or interfere with car-to-car and car-to-hub communications and interaction.

Self-driving vehicles obviously have a long way to go and, rather obviously, they will probably require a human touch for at least the next few years, if not more. That delay in full autonomy could even last for decades if regulating authorities and automakers can't agree on how to safely implement the technology – or if any of the other issues outlined above can't be addressed. While it can be fun to dream of the future, all of those points suggest that it is going to be a slow march to progress.

Advertisement