SA needs laws to regulate selfdriving vehicles
Artificial intelligence should be embedded with decisionmaking values that are in line with our values
LAST WEEK, I presented a talk in Parliament about artificial intelligence (AI), ethics and the law. For starters, AI is increasingly being used to perform tasks previously done by human beings. Doctors, for example, look at the electroencephalography (EEG) signal to detect epilepsy. For a variety of medical, interpretive and social reasons, epilepsy is misdiagnosed. The process of looking at the EEG signal to detect epilepsy can now be performed by an AI doctor.
The AI doctor cannot become tired or biased and has been found to be more consistent than a human doctor. Despite these impressive results, some serious ethical issues need to be tackled.
The company Uber makes more than R80 billion a year. Companies like Uber are managed services providers (MSPs) that connect customers to suppliers.
Uber runs a taxi business, but does not own any taxis. Because these MSPs are based on an IT platform and are domiciled in the cloud, they can easily avoid local regulations and control. When one uses these taxis, for instance, a customer in Joburg often pays a taxi driver through a platform based in San Francisco.
Uber has been making huge investments into self-driving cars. They drive on our roads and, therefore, are subjected to our rules and regulations, such as speed limits. So who is responsible for the fine if the self-driving car runs a red light or drives over the speed limit? According to our laws, if a driver is caught driving over the speed limit, he or she – and not the owner of the car – is liable for a fine. Given the fact that a self-driving car drives autonomously of its human owner, do we still charge the owner?
A few weeks ago, a self-driving Uber killed a pedestrian in Arizona. According to preliminary investigation reports, this selfdriving car actually noticed her, before running her over. If this car had had a driver, he or she would have been charged with involuntary manslaughter, but as this was a selfdriving car, no one was arrested for the crime.
For Uber to have released the car on to the roads, the chief technical officer (CTO) would have had to give permission. Is Uber’s CTO liable for this alleged crime?
It is time that the South African Parliament created laws to govern autonomous robots. Suppose a self-driving car is carrying four passengers. If it reaches a point where it has to either hit a pedestrian or go over a cliff to avoid them, killing all of the passengers, what should this car do?
The philosopher Jeremy Bentham came up with the theory of utilitarianism. If the self-driving car applies utilitarianism, it will do that which will bring the “greatest amount of happiness to the greatest number of people”. So if it saves the four passengers and kills the pedestrian, then the passengers will be happy to be alive. If it kills the passengers and saves the pedestrian, the pedestrian (one person) will be happy to be alive. So, to get the most people happy, it should kill the pedestrian and save the passengers.
Now, if we are to move away from utilitarianism to ubuntu philosophy and the pedestrian is a 3-year-old child and the four passengers are all over 60, then the car should kill the four passengers and save the child. Our legislature should enact laws that will ensure that these self-driving cars and any intelligent machine in our factories operate according to our values, which are based on the principles of ubuntu.
To do this, our legislator will need to understand the principles of AI and its implications. Our engineers will have to develop the capability to be able to remodel these robots, so that they are embedded with decision-making capabilities in line with our values.
Another example of ethics in AI is the area of social networking. The business model of Facebook, Twitter and Instagram is based on the principle that they will give you an account in exchange for your data (where you go, what information you searched for, etc). This is sold to advertisers. When Mark Zuckerberg was asked about this in the US Congress, he replied that he believed people should have the right to do what they want with their data.
The problem with his response is that many people have no idea where their data ends up when they sign up for these applications. With South Africa, there is an added security issue in that much of the data collected in the country is warehoused in California.
Parliament should consider introducing a law that guarantees the basic right to privacy that cannot be taken away because of some legal contract. One example of this is the right to life that cannot be waived away because of some legal contract. As of May 25, 2018, the EU introduced the General Data Protection Regulation (GDPR) and this will have an impact on South African businesses that collaborate with European entities.
One of the major inventions in biometric security is face recognition technology, based on AI. Face recognition databases and algorithms are trained with many facial images of people and their names. The AI algorithm then learns the relationships between the faces and their names. These systems are now in our phones. It turns out that the faces used to train these AI machines were predominantly of Caucasian people and those least represented were of people of African descent.
The consequence of this is that the face recognition system then discriminates against African people. This, of course, is not ethical and our laws must intervene to make sure that such discrimination does not persist. Our legislature should develop laws to ensure that products imported into South Africa comply with our constitution and do not discriminate.
The other example that has a significant ethical dimension is the issue of clinical trials. A few years ago, we registered a patent in the US on a “robot voice”. This voice is used by a person who has had her voice box surgically removed because of cancer. A big international medical appliances company took interest in our patent.
As we were discussing this patent, an issue on where these devices would be tested arose. Our international counterpart indicated that their laws of clinical trials were stricter than ours and therefore the clinical trials would have to be done in South Africa, despite the fact that the device was to be sold on the international market. Studies have shown that Africa is becoming a home for clinical trials.
However, the South African regulatory framework is strong and perhaps, through the Pan African Parliament, it should help other African countries to develop a robust policy to protect human lives.
To be able to regulate technology so that we can protect human lives and human dignity, we need to understand it.
IN ORDER TO PROTECT LIVES, WE MUST BE ABLE TO UNDERSTAND TECHNOLOGY.
■ Marwala is the Vice-Chancellor and Principal of the University of Johannesburg and the co-author of the book Smart Computing Applications in Crowdfunding. He writes in his personal capacity.
ETHICAL MINEFIELD: If a self-driving car breaks the law or causes a person’s death, who is liable? asks the writer.