A New Way of Thinking About the "Trolley Problem" of Artificial Intelligence

Over the next 20 years, as robotics advances, self-driving cars, industrial robots, and medical robots will have greater capabilities, greater autonomy, and more widespread adoption.

Responsible application of robotics and machine learning can save more lives, and the benefits outweigh the drawbacks. 

Image credit: Unsplash/Gabriella Clare Marino

1. Autonomous robots and autonomous decision-making can indeed cause fatal errors.

2. Deaths caused by robotic errors will create a new age "trolley dilemma" of moral dilemma.

3. If society accepts machine learning and makes every effort to introduce robotics in a responsible manner, more lives will be saved.

Over the next 20 years, as robotics advances, self-driving cars, industrial robots, and medical robots will have greater capabilities, greater autonomy, and more widespread adoption. Inevitably, these autonomous robots may make mistakes in decision-making, resulting in hundreds or thousands of deaths. But such disasters can be avoided if humans are involved.

Such a future is certainly scary, but once human society is able to use robotics responsibly, more lives will be saved than deaths will be caused.

The Process of Machine Learning

Robots are not "programmed" by humans to mimic human decision-making processes. They learn from big data, using complex mathematical formulas derived from the data to perform tasks such as "recognizing traffic lights". The machine learning process requires far more data than humans need. However, once trained, robots will outperform humans in any given task. Thanks to machine learning, AI and robot performance have improved dramatically in the last five years.

The points I make in this article apply to healthcare, manufacturing, and other industries that are rapidly becoming automated. Let's take autonomous driving as an example. While an experienced human driver may have hundreds of thousands of miles of driving experience in a lifetime, Waymo, Google's self-driving car company, will have completed 2.3 million miles of road testing in 2021 alone, with the AI technology behind it learning from data about each vehicle's driving experience, and these self-driving vehicles never get tired or, unlike forgetful These self-driving vehicles never get tired and, unlike forgetful human drivers, may forget the mistakes they've made.

When Tesla's "Smart Call" feature was first introduced, the cars could leave parking spaces and navigate around obstacles without being operated by their owners. At first, many users complained that the new feature did not perform as well as it should have, but within a few weeks, Tesla collected data from early adopters and retrained the machine learning model behind the new feature. Since then, "Smart Call" has improved significantly and become a key competitive advantage for Tesla's new cars.

Autonomous Robots Can Save Lives

As the amount of data available for learning increases, the capabilities of artificial intelligence are rapidly improving, and AI is becoming more accurate, adaptive, and safe. As more and more robots enter mainstream everyday applications, their applications are becoming more widespread, signaling that functional robotics is gaining traction. Autonomous driving will evolve from "holding the steering wheel" to "letting go of the steering wheel", to "no need to monitor" and even "no need to pay attention to "and eventually evolve to a fully automatic state of "no steering wheel".

A good example is the Chinese autonomous driving company Wenyuan Zhixing. The company has deployed uncrewed minibusses and uncrewed sanitation vehicles in several Chinese cities. They operate in a more restricted environment than driverless cabs, but with a significant increase in safety compared to human drivers. These vehicles will eventually be freed from these initial road-going constraints after operating in specific environments and collecting large amounts of data.

As robotics evolves from simple applications to complex scenarios, we will have access to more data that will improve its performance and safety. For example, by reducing human error, which is the most common cause of road accidents, self-driving cars could prevent 47,000 serious crashes and save 3,900 lives in the UK alone over the next decade. The RAND study found that even if autonomous driving were only 10 percent safer than human driving, that translates into many precious lives saved.

Ethical Dilemma

Most people still show great concern about the mass adoption of robots, including the moral controversy over human lives lost due to machine error. The classical "trolley dilemma" refers to a moral dilemma in which a bystander can switch the tracks of a runaway trolley, killing one person but saving five others. This dilemma illustrates that the choice between "who lives and who dies" is essentially a moral judgment and that such decisions should not be left to uncaring machines.

However, the "moral dilemma" is further exacerbated by the fact that robots and humans have different perceptions and naturally make different types of errors. For example, robots are quick and always focused, but they can misjudge obstacles, such as the driverless car from Uber that mistook a pedestrian pushing a bicycle across the street for a car and thus predicted that it was moving faster than it was.

The difference between human error and machine error makes it even harder for the public to accept deaths caused by robots. It must be even harder for them to let go if they have ever heard media spin like that following the 2018 Phoenix crash in the United States. Once the various media outlets continue to lambast every self-driving death with big, condemning headlines and biased amplification, it's likely that people will lose faith in self-driving systems altogether, even if the technology has the potential to ultimately save millions of lives.

If a human driver causes the death of another person, they will face legal judgments and consequences. But the "black box of artificial intelligence" cannot explain to judges and the public the reasons for the decisions made in human-understandable terms, or in legally and morally justifiable human language.

Another point that needs to be debated is accountability. In that Phoenix accident, the human driver in the car was charged with manslaughter. But should the car manufacturer, AI algorithm provider, or engineer be held responsible and liable in this case? Only when accountability is clearly attributed can we have a basis for consensus and guidelines to build a sound autonomous driving ecosystem.

The "Trolley Conundrum" in the Age of Artificial Intelligence and Machine Learning

Since robotics can save countless lives, there is a strong case to be made for a major push for automation, as long as it proves to be marginally better than humans. Let us take every opportunity to introduce robotic automation tools that benefit humans, initially in restricted and specific environments, and then gradually scale up adoption, eventually gaining more autonomy and a broader push. Through this step-by-step approach, we can collect more data and improve the performance of robots on the one hand, and minimize accidents that endanger human lives on the other.

Given the potential for dissent, we need to work together to make the public more aware of the short-term pains and long-term benefits of robotics. Only then can we gradually develop a responsible and rigorous attitude as we embrace automation technology so that robots can better serve human society.

Link to the original article.

https://www.weforum.org/agenda/2022/05/ai-s-trolley-problem-debate-can-lead-us-to-surprising-conclusions/

Related News

1、Chip Packaging Lead Time Has Grown to 50 Weeks

2、Eight Internet of Things (IoT) Trends for 2022

3、Demand for Automotive Chips Will Surge 300%

4、Volkswagen CFO: Chip Supply Shortage Will Continue Until 2024

5、BMW CEO: The Car Chip Problem Will Not Be Solved Until 2023

6、Shenzhen: This Year Will Focus on Promoting SMIC and CR Micro 12-inch Project