On 14 March 2019, the South African State Information Technology Agency (SITA) hosted a conference to which various industry representatives attended and discussed the different perspectives, challenges and uses of AI. The aim of SITA is to provide Information and Communications Technology (ICT) services to the public in an integrated, secure and efficient manner. The theme of the conference was powering digital transformation through Artificial Intelligence (AI) and what this means for South African citizens.
This is an interpretation of the views expressed during the course of the conference.
What is AI?
AI is the embodiment of human intelligence in computer technology and is driving the 4th Industrial Revolution (4IR) which is on the brink of transforming every aspect of our lives. As we approach the 4IR, the common assumption is that AI is about to replace the human. Stephen Hawking professed that AI is to end mankind, will take off on its own, and re-design itself at an ever increasing and uncontrollable rate. The speed, breadth and sophistication with which AI systems can process data and absorb the role previously occupied by humans is astounding, and leaves the progressive data scientist on the edge of an exciting revolution and the cautious, increasingly sceptical. However, during the course of SITAs industry day, which brought together a diverse collection of industry representatives and converged their different perspectives towards a single focus, several lessons were learned. The views that were expressed, and are discussed below, indicate that there certainly is a different narrative to that espoused by Hawking.
AI is not going to replace human intelligence
AI uses hindsight to develop insight and create foresight. In other words, it uses human input of data to develop intelligence and provide solutions, but it requires human expertise, knowledge, experience and intuition before it can process data. Poorly curated data leads to poor AI decisions. AI is nothing without data and skills, both of which allude to the unquestionable relevance of the human. Without sophisticated human input, innovation and imagination, AI is incapable of producing sophisticated outcomes. AI is therefore, about enhancing and complementing human capabilities and so long as humans are ready to embrace and adapt to working with AI, towards more efficient and effective solutions, then it will not replace us but will elevate our human innovation. By filtering out the dearth of irrelevant data, AI can direct the human intelligence to focus on what is real (or useful or relevant), which enables efficient human analytics. This point is worth emphasising: humans will always remain integral in the system of AI given the irreplaceable nature of human reasoning and the requirement for human data input. For that reason, AI ought to be understood as Augmented Intelligence where the common aspiration is enhanced productivity.
Such an understanding of AI as amplifying human ingenuity, allows us to dismiss the fears that AI will eliminate the human race and cause existential job loss. In fact, the AI industry is adamant that humans will remain in control and where AI is used effectively, it will create more jobs than it removes, so long as we can adapt to a new type of work environment, whatever that may look like. We should move towards seeing AI as the process of humans leveraging technology to solve problems rather than a threat of our replacement of our own making.
The capabilities of AI are astounding
AI is able to respond to many of the problems we face in South Africa such as the challenges we face in the public service delivery sector. The South African government has already committed to building a digitally transformed society by creating digital platforms to provide better services. The government envisions a centralised system of e-government with a single point of entry and exit for data and information. Though this is an admirable vision, there can be practical difficulties in terms of the implementation of such proposals, as experienced by the Government in the rolling out of broadband services under the SA Connect program.
AI is also being used to propel social wellbeing and safety. Some insurance companies now offer insure telematics services which, for example, when applied to motor vehicles can monitor the driving behaviour of an individual. This data is then analysed by the insurance company, which is able to award and incentivise good driving. Such companies, through using AI sensors, are also able to provide value added services including sending ambulances when an accident and its severity is detected or notifying owners of and intercepting car theft or hijacking when the sensor detects unusual activity, driving behaviour or location. The quantity of data and the speed at which it can be analysed using AI systems allows us to measure, not what we think we know, but what we actually know. This can allow companies, for example, to accurately price an individualised insurance premium based not on the assumption that a young male must be a high risk driver, but on the knowledge that he is actually less of a risk than an elderly individual driving a less sophisticated car. Using data and emergent technologies for preventative rather than responsive purposes, driving behaviour and therefore, public safety, can be improved. This kind of technology is not limited to the transport industry. Innovation will allow other services to be enhanced in similar ways by using such technologies.
All of this however, is only possible with the sharing and releasing of data. The provision of more tailored services and generally, the advancement of all AI technology relies on the sharing of data which introduces a wealth of concerns.
Security challenges with AI and the use of mass data
As we build ever more sophisticated systems of AI and as the sharing of information, necessary to enhance AI capabilities, is accelerated, digital security is becoming ever more important.
Privacy is central to concerns around the 4IR and the increasing development of AI. From a human rights perspective, mass surveillance and recording and sharing of data pose a threat to the right to privacy as enshrined in the Constitution and embedded in human dignity. AI policies must be continuously developed and revised to ensure sufficient digital security. Furthermore, in the next several years AI will create new challenges at the intersection of technology and society, some of which we simply cannot predict at present. As such, legislation, regulation and policy should not only be responsive to new challenges but also capable of predicting them.
AI policies must also take account of data ethics and ensure that government objectives in the development of AI technology ensure adequate market protection, is transparent and inclusive and meaningfully engages with the public to ensure that the voice of citizens and stakeholders is heard. AI must also respect the consenting individual and therefore, there must be sufficient communication and public participation so that individual citizens understand the digitised economy in which he or she exists.
Challenges of keeping pace with AI
Another problem that will affect us all is that the skills profile needed today in every sector is changing dramatically, and how we adapt will depend entirely on our skills and education. We all need to keep pace. To understand the capabilities of technology, both those entering the workforce and those already established in it, will need to embrace upskilling and potentially re-skilling. Given the speed at which emergent technology is moving, we must also embrace a more holistic and more integrated ICT-based approach to education. We will need to be equipped with general critical and computational thinking skills and these must be integrated into every aspect of learning. Only then will we be able to adapt to new technologies. If South Africa intends to keep pace, the government needs to address the lack of skills and focus on revising the curriculum and driving IT projects.
The shortcomings and bias of AI
It is also crucial to address the social implications of AI. Racial and gender biases in AI evidences an advanced level of inequality. One problem is that where an AI system has been developed by a specific demographic, it fails to account for social, racial and gender differences and leads to false positives or false negatives in decision making, which we as humans, have to be ready to identify and challenge. Facial recognition software, for example, often fails to identify the correct gender or race. From a feminist perspective, the lack of women’s input in AI systems and under-representation in the industry is also concerning. In light of such problems, we cannot always trust the reasoning of AI and if we do, without applying human rationale, minorities may be targeted and social inequalities, exacerbated. However, these types of problems can be rectified by increasing data input. The more sophisticated we are when dealing with AI and providing data input, the more sophisticated the AI will be but for now, there is much work to be done.
What does all this mean for a South African citizen?
You will be interacting with AI on a daily basis already, whether you know it or not. If you are on a web chat seeking assistance in real time from a service provider, chances are you are communicating with a bot who has been programmed to respond to you. If you are drinking water or using electricity, it is almost certain that those services have incorporated AI to ensure safe and efficient provision. If you are receiving medical treatment, it is possible that an AI database of medical knowledge has influenced your doctor’s medical decisions. If you are using GPS, you are already tapping into a network of real time digital mapping. If you use a smartphone, open a web account, submit your fingerprint as part of a government interaction, your data is being collected. Our interaction and socialisation with AI will only increase.
Our daily interaction with technology is increasingly requiring us to educate ourselves to understand the platforms we are engaging with and to take responsibility to ensure our own safety and security.
Aside from taking responsibility for ourselves in this rapidly changing and complex, digitised environment, every citizen, every government, every public or private business, and every civil society organisation has a role to play. AI is inter-disciplinary and it is a false assumption that physicists and mathematicians are the driving force. A more advanced system of AI is one that incorporates and integrates ICT with social sciences, the arts and the humanities. What this means is that we can all participate in the development of AI and the 4IR, and we should.
The only limit to AI is our imagination
It took 100 years for us to move from the first to the second industrial revolution. Between the third and the fourth, 30. As we grapple with this 4IR, we cannot forget to imagine the fifth. How long will it take before we are on the brink of the 5IR? What will that look like? What will be the challenges to society? How will they be regulated? How will we keep pace with our own augmented intelligence? What are its limits? Is it possible to know too much? And are we ready?
One thing is clear: if we neither harness our imaginations nor ask these complex questions, we can never be ready.
This piece is a culmination of the views expressed at the SITA industry day coupled with Marianne’s personal opinions. Marianne writes in her personal capacity and the views expressed herein do not necessarily represent the views of ALT Advisory.