Artificial intelligence (AI) is increasingly being used successfully in many areas of our lives. AI is a technology that makes machines intelligent, and these machines learn to speak, hear, feel and think. One area where AI is increasingly being used is in the medical field. For example, in South Korea, the company Seegene used AI to develop the coronavirus rapid testing kit. In China, it was used to sequence the Covid-19 genome in a month, a task that could have taken much longer. Taiwan used AI to detect locations where the coronavirus had been identified. In the United Kingdom, the company Benovelent AI is using it to identify the drugs to be used to treat coronavirus in the absence of the vaccine.
Despite the successes of AI, however, it needs to be regulated. As a member of the World Health Organisation (WHO) committee that is developing guidelines for the applications of AI in medicine, I know just how risky not regulating the technology can be.
Last month, Washington became one of the first states in the United States to pass a bill on the government’s use of facial recognition software. The bill requires government agencies to regularly report on their use of facial recognition technology and test the software for fairness and accuracy. The importance of this is apparent when you consider that this could affect jobs, financial services, education, insurance and housing. Although the bill still needs to go to the State House for consideration, it has furthered essential debate on the use of legislation to regulate AI.
Why is this so significant? Facial recognition technology is one of the major inventions in biometric security based on AI algorithms that are trained with many facial images of people, their names and unique identification. The AI algorithm then learns the relationships between the faces of people and their corresponding names. Yet, it turns out that the faces that are used to train these AI machines are predominantly Caucasian. The people least represented in the database are people of African descent. The consequence of this is that the face recognition system then inadvertently discriminates against African people, so there is an inherent bias in this kind of technology. Here, the law must intervene to make sure that such discrimination does not persist.
This is not the only aspect of the fourth industrial revolution (4IR) that needs regulating. Over the past two years, Uber has been trialling self-driving cars in the US. These cars drive themselves but are still subjected to our rules and regulations such as speed limits or stopping at a red light. Yet, if there is no driver, who is responsible for a speeding fine? In a more extreme case, who is responsible for the death as seen in Arizona in 2018 when the car, it is alleged, drove into a pedestrian?
In 2018, President Cyril Ramaphosa announced that through the Presidential Commission on the 4IR, technologies would be used to augment South Africa’s competitiveness. The commission has made eight recommendations that will put South Africa’s fortunes on the upward trajectory. In the past five weeks on this platform, I have addressed five of these recommendations. The first is to build human capacity in the area of the 4IR; the second is to establish the National Artificial Intelligence (AI) Institute; the third is to create the Advanced Manufacturing Institute (AMI); the fourth is the establishment of a National Data Centre; the fifth is to incentivise the adoption of 4IR technologies and the emergence of future industries and platforms. This week I address the sixth recommendation, which is to review, amend or create policy and legislation.
Although the examples of Uber and facial recognition technology are, of course, just two facets, we need to ensure that our legislation is in line with the 4IR across the board. Ministries, or line departments, would be tasked with looking at all our legislation and updating them to be approved by Parliament. This would require the legislature and the state executives to be trained to become 4IR and science-literate in order to implement changes. Because of the fast-paced nature of technological innovation, there should not be a lag between regulation and the effect of technology on the lives of citizens. As distinguished visiting professor of law at University of Johannesburg, Professor Michael Martinek says, legislation should simultaneously enable innovation.
The report Worldwide AI Laws and Regulations, compiled by research firm Cognilytica, found that many governments are adopting a “wait and see” approach to creating laws and regulations in this space. As Kathleen Walch, one of the authors of the report, put it: “Adoption for AI and cognitive technologies shows no signs of slowing and governments are paying attention. AI laws are coming.” This had even been alluded to by the authors of sci-fi novels in the 20th century. In 1942, the American writer Isaac Asimov set out what he called the Three Laws of Robotics in I, Robot. For instance, a robot may not injure a human being; a robot must obey orders given to it by human beings unless it conflicts with the first law and a robot must protect its own existence as this does not conflict with the first two laws. Yet, as this increasingly becomes a reality, what laws do we actually have to put in place?
This is not just for seemingly obscure cases such as facial recognition software or self-driving cars. Our legislation and regulatory frameworks need to encompass technology infrastructure and access to broadband which, as outlined in previous weeks, has been a slow process, as well as laws to regulate cybersecurity, electronic transactions and data, for instance.
It should also extend to relooking at our tax laws. We should develop a framework for tax companies that are domiciled overseas but make their money in South Africa. These companies include Uber, which connects taxi drivers to customers in South Africa from California, and Netflix, which runs a streaming service in South Africa from California. The regulatory frameworks of countries have not sufficiently caught up with new business models. The model of Uber is predicated on keeping prices low by avoiding tax payments. This has not been tested in South Africa as yet, despite the occasional flare-up by the public and to a lesser degree by regulatory bodies regarding tax payments. Tax not generated is income lost to the country that could be used for the public good of the citizens. In the UK in 2019, for example, a conservative estimate of unpaid VAT was pegged at more than £1-billion.
As we begin to see AI and other 4IR technologies as our new normal, there is a fundamental need for our structures – including our regulations and legislation – to adapt to this new wave. This, arguably, is a mammoth task that will require collaboration across ministries and spheres of society before it is brought forward in Parliament.
This is the sixth in a series of eight articles unpacking the recommendations of the Presidential Commission on the Fourth Industrial Revolution.
Tshilidzi Marwala is a professor and the vice-chancellor of the University of Johannesburg. He deputises for Ramaphosa on the Commission