Artificial Intelligence

Background

Africa continues to express growing interest in the development of an AI ecosystem. This rapidly developing new technology is portrayed to have the potential to solve some of the most pressing challenges in Africa and drive growth and development. Developments in AI have been predominantly driven by private sector technology actors, but growing interest by African governments has seen the start of conversations around “AI strategies” for growth and governance across the continent. It is in line with these thoughts that policy questions around AI arise with prompts to create clear roadmaps to guide the adaptation of AI technology in Africa. Data forms an integral part in the development of AI, the policy implications on the use of data for AI development engender the need to explore law and regulations that govern the use of data by the investors and developers of AI systems. Regionally there are no AI-specific laws or regulations that govern the use of AI, these laws are still in their inception stages. As the continent continues to explore the implications of AI, we endeavor to explore the extent to which data protection laws, regulations, and guidelines are applicable to AI development and to what extent they present safeguards in how AI is developed and used.

A Brief History of AI

Over the years, various scholars have given different definitions of AI, from basic to highly technical and scientific. In 1956, Stanford professor, John McCarthy, a now well renowned Computer Science and Artificial Intelligence innovator, first defined AI to be, the science and engineering of making intelligent machines. It is from this definition that AI was further broadened to refer to, the creation of intelligent hardware or software, able to replicate “human” behaviors such as learning and problem solving or a constellation of technologies that enable machines to act with higher levels of intelligence and emulate human capabilities to sense, comprehend, and act.

The inception of AI dates to the 1950s when researchers in innovation and computer science spent time trying to better understand and improve the ways in which they could actualize the science of making intelligent machines. Machine learning became a popular focus of attempts at improving algorithms to better offer solutions to real-world problems; the success of which prompted AI advocates to make even greater strides in actualizing the use of AI. In the 1970s, however, AI encountered numerous obstacles, particularly where data storage and fast processing were concerned. Computers at the time could not store large amounts of information or process it fast enough. The AI industry entered a period where commercial and scientific activities in AI declined dramatically -the decline triggered by the American government’s decision to pull back on AI research. Consequently, the work on AI stalled, only to be revived in the 1980s by heavy investment by other governments-Japan taking particular interest in realizing the AI dream through the inception of the Fifth Generation Computer System Project (FGCS). Like the rest of the world, Japan was lagging behind the United States in technology, the project was developed with the intention of giving Japan the lead in technological advancements. This quest to realize the AI dream incentivized the funding of expert systems and other AI related endeavors with the goals of revolutionizing computer processing, implementing logic programming, and improving artificial intelligence throughout the 1980s-1990s.

There were significant developments in the use of AI in the 1990s and 2000s. In 1997, great breakthroughs were seen in AI – the world chess champion and grandmaster Gary Kasparov was defeated by IBM’s Deep Blue, a chess playing computer program. Within the same year, a speech recognition software, developed by Dragon Systems, was implemented on Windows. Since then research and development in AI have only grown, characterized by improvements in the performance of AI due to:

• Abundance in computing power – the development and use of state of the art computers that are faster, making it possible to process very large amounts of data sets within short periods of time as needed for training and fast deployment of AI algorithms.
• Rapidly declining storage costs – it is now easier and cheaper to store large amounts of data which is key for successful training and deploying AI algorithms.
• Surge in data availability – in recent years there has been a rapid increase in the availability of datasets which has primarily been propelled by the increase of internet connected devices as well as arise in the use of social media.
• Increase in AI investment – more money and resources are now being invested in AI both from private sectors,

Current AI Status

AI is expected to affect all areas of socio-economic life with AI innovations already in use in various sectors: banking, marketing, entertainment, education, agriculture, and health. Further, governments around the world have begun to see the potential for AI for their economies, societies, and in delivering public services. In 2017, Oxford Insights created a Government Artificial Intelligence Readiness Index to assess how well governments are placed to take advantage of the benefits of automation in their operations. The index covers 194 countries and territories in respect of their preparedness to use AI in the delivery of public services. Regionally, North America ranks as the best performing region with Africa and the Asia Pacific ranking as the worst. The Government Artificial Intelligence Readiness Index of 2019 estimates that AI technologies will add approximately $15 trillion to the global economy by 2030. The index, through its ranking, also presents clear disparities and inequalities in AI readiness between the global governments; showing higher income countries faring better in the ranking than middle or lower income countries.

While AI presents great opportunities for growth and development, it is important to ensure that AI technologies do not widen already existing social and economic inequalities, especially for vulnerable communities and marginalized groups. AI scholars have argued that caution must also be taken in the implementation of AI technologies. Although the standard narrative created surrounds the positive impacts of AI, when consideration is not given to policy implementation and ethical concerns, AI can have drastic negative impacts especially in exacerbating social inequalities, deepening digital divides by introducing new forms of exclusion and AI labour exclusions.

AI in Africa

The AI ecosystem in Africa is rapidly developing. AI has the potential to address some of the most pressing challenges on the continent and further drive growth and development in the core sectors of agriculture, healthcare, public service, and financial services.
Although The Government Artificial Intelligence Readiness Index of 2019 ranks Africa among the worst performing in respect of AI preparedness, the index ranks at least 12 African countries in the top 100. This index presents the regional readiness of African countries to embrace, promote and use AI – clearly ,a lot still needs to be done within the region to ensure the development of a healthy AI ecosystem. Efforts directed towards changes in education systems, through the creation of frameworks that will develop citizens’ skills needed to fully understand and embrace the use of AI, alongside addressing the ethical implications of fair, secure and inclusive use of AI and the re-assessment of laws and legal frameworks to support data driven technologies, innovation and growth, need to be considered.

A big part of building this ecosystem is ensuring that AI is relevant to African culture and context. Data plays a big role in this; ensuring the availability of a broad and accessible pool of data is important in enabling researchers, developers and users to understand, create and drive the use of AI within the different core sectors.

References:

‘Artificial Intelligence for Africa: An Opportunity for Growth, Development, and Democratisation’.   https://www.accesspartnership.com/artificial-intelligence-for-africa-an-opportunity-for-growth-development-and-democratisation/#:~:text=The%20rapidly%20developing%20set%20of,efficiently%20and%20effectively%2C%20raising%20yields

Clayton Besaw and John Filitz, ‘Artificial Intelligence in Africa is a Double-edged Sword’.   https://ourworld.unu.edu/en/ai-in-africa-is-a-double-edged-sword

Shanhong Liu, ‘Artificial Intelligence (AI) worldwide – Statistics & Facts’, (Statista, 13 March 2020).  https://www.statista.com/topics/3104/artificial-intelligence-ai-worldwide

Chris Smith , Brain McGuire , Ting Huang , Gary Yang, ‘History of Artificial Intelligence’, ( University of Washington , 2006). https://courses.cs.washington.edu/courses/csep590/06au/projects/history-ai.pdf

Rockwell Anyoha, ‘The History of Artificial Intelligence’, (Harvard University , Graduate School of Arts and Science, 28 August 2017).

Rockwell Anyoha, ‘The History of Artificial Intelligence’, (Harvard University , Graduate School of Arts and Science, 28 August 2017). http://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/?unapproved=247225&moderation-hash=9724821d51b2ba31eddbfde04a03949d#comment-247225

Sanjeev Kapoor, ‘A Look at the Current Status of Artificial Intelligence’. (ITexchange, October 2019),  https://www.itexchangeweb.com/blog/a-look-at-the-current-status-of-artificial-intelligence

‘Government Artificial Intelligence Readiness Index 2019’, (Oxford Insights and the International Development Research Centre). https://www.oxfordinsights.com/ai-readiness2019

Alexa Hagerty and Igor Rubinov, ‘Global AI Ethics: A Review of the Social Impacts and
Ethical Implications of Artificial Intelligence’, (Cornell University, 2019). https://arxiv.org/ftp/arxiv/papers/1907/1907.07892.pdf

Mapping of AI Applications in Africa

CIPIT’s AI dashboards provide information on the emerging AI ecosystem on the African continent. African AI applications, in this study, are defined as applications that were:

1. 1. Generated with machine learning and deep learning algorithms.
   2. Created by AI developers located in Africa.
   3. Created by companies composed primarily of African developers creating AI specifically for use in Africa.
   4. Operational at the time of data collection.
   5. Created by companies that were operational at the time of data collection.

We compiled our list by evaluating entries acquired by accessing and analyzing open-access material, including news articles, websites, corporate documents, academic articles, NGO reports, expert submissions, and other public sources against the above criteria. Unfortunately, active AI applications not listed in any of the open access material above would be excluded from the data.

Introduction to our Work

Artificial Intelligence (AI) is a suite of technologies that are experiencing a renaissance moment. No longer merely a dream of computer scientists, AI is now widely deployed in a variety of applications. Our focus is “African AI” – AI that is developed in Africa and/or by African developers for use within the African context or primarily for African users.

By researching AI tools, products, trends, policies, and impacts, we seek to contribute to the conversations that will help to positively steer the development and impact of AI. In particular, we seek to contribute to the definition and pursuit of responsible AI. We are also committed to the incorporation of gender issues in AI research and development as AI becomes more widely adopted.

Our current AI research is focused on the central question “Does African AI differ substantively from AI developed elsewhere?”.

We invite you to explore our work through the tabs at the left.

Below is a sample of our work: Our submission to the World Intellectual Property Organization (WIPO) as they seek consultation on issues affecting Intellectual Property and Artificial Intelligence.

The speaker is Ms. Wanjiru Muchiri, a Research Fellow and IP expert at CIPIT. The full video can be found here.

Our IP team also took part in the World Intellectual Property Organization (WIPO) “How does AI work?” competition in November 2020 where participants were asked to send in short videos explaining AI in a fun, engaging, and easily accessible way. We are pleased to announce that we were selected as one of the winners and our video is displayed on the WIPO AI Exhibition page.

About the project

Climate change is widely acknowledged as one of the greatest threats that humanity will face in the 21st century. Africa is not immune to the challenges, and in some ways will be more greatly affected by them. Climate change models are currently using AI to improve accuracy, but these are largely developed outside of Africa, and may not take into account the specific nuances of culture, geography, and populations in Africa. We seek to know whether African developers are using AI to assist the region with climate change adaptation and impact mitigation.

The African Handbook of Climate Change Adaptation

The African Handbook of Climate Change Adaptation is part of the “Climate Change Management Series,” published by Springer. With over 100 chapters covering the whole African continent, it provides a robust and long-lasting contribution to the literature. The publication is fully peer-reviewed by a panel of editors and reviewers and is coordinated by the International Climate Change Information and Research Programme (ICCIPR). We have contributed a chapter on the use of AI to address climate change in Africa.

Chapter title: Use and impact of artificial intelligence on climate change adaptation in Africa

Abstract: Although Climate Change is a global phenomenon, the impact in Africa is anticipated to be greater than in many other parts of the world. This expectation is supported by many factors, including the relatively low shock tolerance of many African countries and the relatively high percentage of African workers engaged in the agricultural sector. High-income countries are increasingly turning their focus to climate change adaptation, and Artificial Intelligence (AI) is a critical tool in those efforts. Algorithms using AI are making better predictions on the short- and long-term effects of climate change, including predictions related to weather patterns, floods and droughts, and human migration patterns. It is not clear, however, that Africa is (or will be) maximally benefitting from those AI tools, particularly since they are largely developed by highly-developed countries using data sets that are specific to those same countries. It is therefore important to characterize the efforts underway to use AI in a way that specifically benefits Africa in climate change adaptation. These efforts include projects undertaken physically in Africa as well as those that have Africa as their focus. In exploring AI projects in or about Africa, this chapter also looks at the sufficiency of such efforts and the variety of approaches taken by researchers working with AI to address climate change in Africa.

About the project

As with other software specialties, a majority of AI developers are men. It has been proposed that this, as well as biased data sets, leads to cases of gender bias in AI products as currently observed and widely reported. We propose to investigate the issue of gender bias in African AI systems and products, to determine whether similar biases are present, as well as how African developers have sought to minimize such biases.

Results of the study, executive summary

The gap between men and women has widened, if not worsened under the advent of technology. New tools and methods of discrimination for instance in biased algorithms to incorrect labeling leading to algorithms miscategorizing women in certain professions, flawed keyword searches, and sexist terminology in Information and Communication Technology (ICT). The persistence of inherent stereotypes in Artificial Intelligence (AI) including AI assistants with female voices and representation of AI in media on a sexist scale has called into question the assumption of neutrality of algorithms and the potential for harm in the future. While much has been said on the global AI gender gap, the African AI landscape lies largely unexamined in its contribution to the gender gap in AI. This report is an experimental study into the representation of women in AI in Africa based on data collected on the staff composition of African based AI companies and projects.

The gap between men and women in AI in Africa stands at 71-29% compared to the global gender gap of women in AI which stands at 78-22% as of 2018. The proportion of female participation per industry differs from corporate services, healthcare, and agriculture having the smallest gender gaps and entertainment and environmental conservation having the widest.

This study also considered the participation of women in management fields to assess power and decision-making hierarchies in AI companies and projects across gender lines. The study shows women making up 13% of CEOs and 10% of founders with an overall gender gap of 90% to 10% for overall leadership positions between men and women respectively.

This study provides an analysis into the causes of this gender gap and their unique situatedness in Africa noting that, in part, there needs to be greater care into what AI is borrowed from the Global North and whether in borrowing tech we are not also borrowing added layers of gender inequality to an already unequal system. The study also considers that an analysis of AI gender inequality in Africa requires a tripartite discussion nuanced with this study forming one part, the inclusion of women in the design of AI.

Our paper on this study will be available soon.

The Artificial Intelligence Labour Gender Gap in Africa

Mitchel Ondili et al., The Artificial Intelligence Labour Gender Gap in Africa, Report from CIPIT, (2021).

Data Protection in Africa

Data protection and AI can be analyzed from two perspectives: the first, relating to collection, processing, and storage of personal data for AI development and the second, to the protection of data from misuse and exploitation by researchers and developers during data mining, collection, processing, and storage. Both perspectives interlink and can be addressed, to an extent, by use of similar policies and legal frameworks.

The past year has seen African countries set foundational frameworks with respect to data protection necessitated by the changes and growth of the digital space on the continent. To date, 22 out of the 54 African countries have implemented data protection laws. The remaining majority have bills waiting to be passed into law. This is particularly significant to AI systems that rely on the processing of personal data as these laws create the foundational basis for the protection of personal data against misuse and exploitation by researchers, developers, and even the users of AI.

Personal data is any information that relates to an identified or identifiable living individual. Different pieces of information, which collected together can lead to the identification of a particular person, also constitute personal data.  Thus the general data protection principles on the protection of personal data could apply in specific aspects to AI. Specific data protection principles that may apply include:

• Fairness: To function as intended, AI algorithms are trained using datasets. The principle of fairness mandates that the data must be handled in ways that would be reasonably expected and not used in ways that would have unjustified adverse effects. For AI systems, processing personal data in accordance with the fairness principle, means the data utilized must not be biased or flawed resulting in incorrect or discriminatory results.
• Purpose Limitation: Machine learning often takes place by utilizing datasets collected for other purposes. Purpose limitation requires that the reason for processing data must be clearly established and indicated when the data is collected. The purpose of the processing also needs to be fully explained to the data subject, so that they can make an informed decision regarding consent where applicable.
• Data Minimization: The development of artificial intelligence usually requires the processing of huge amounts of data. The Principle of data minimization compels that the gathered data must be used in an adequate, relevant, and limited manner, only to the extent that is necessary for achieving the purpose for which it was collected.  The principle mandates that sufficient efforts should be taken to determine the exact needs of the algorithm and to select only the information that is actually relevant for those purposes.
• Accuracy: This principle of accuracy requires that personal data is accurate and kept up-to-date. This principle is of particular importance for fully automated systems, where the output could have a significant impact on individuals with little human oversight. In the context of AI, feeding an AI system inaccurate data could lower the quality of the output, and this principle requires that AI users take a particularly vigilant approach to ensuring that the data set is not diluted by bad quality data. These principles have been embedded in legal frameworks on data protection both regionally and nationally by individual countries within their specific data protection regulations. The existing regional frameworks are:
• The African Union Convention on Cyber Security and Personal Data Protection: Adopted in 2014, the Convention has so far been signed by 14 states and ratified by 5 countries out of 54 member states. The Convention gives extensive legislative guidelines and member state obligations in promoting cybersecurity governance and the protection of personal data. Chapters II and III give member state obligations and commitments towards the protection of personal data as well as the application of the data protection principles.
  Additional legislative frameworks have been applied through specific African blocks and have led to the implementation of Data protection laws.

The additional legislative frameworks include:

1. Supplementary Act on Personal Data Protection:
   Adopted in 2010 by the Economic Community of West Africa States (ECOWAS), the Act specifies the prerequisites of data protection laws creating an obligation upon the member states to create a data protection authority. The Act is legally binding. Of the 15 ECOWAS member states only Sierra Leone and Liberia are yet to enact data protection laws.
2. SADC Model Law on Data Protection:
   The SADC Model Law on Data Protection was developed by the Southern Africa Development Community (SADC) 2010 and was adopted in 2013.  Out of the 16 member states, 7 member states have data protection legislation, 6 have draft legislation whereas 2 still don’t have any legislation on data protection. The model law provides a template that has been used by member states in developing their own data protection laws.
3. EAC Framework on Cyber Laws:
   The East African Community (EAC) is the regional intergovernmental organization of the Republics of Burundi, Kenya, Rwanda, Uganda, and the United Republic of Tanzania.  The EAC framework on cyber laws was first established through the assistance of UNCTAD through the development of a Task Force on cyber laws in 2007. The Taskforce Prepared and endorsed two cyberlaw frameworks :

• Framework I covers electronic transactions, electronic signatures and authentication, cybercrime, consumer protection, data protection, and privacy.
• Framework II focuses on intellectual property rights, competition, e-taxation, and information security.

Of the EAC states  Kenya and Uganda have enacted data protection legislations. Rwanda and Tanzania are in the process of enacting their draft legislations.

AI Specific Legal Frameworks

The existence of these frameworks and laws form a good foundation for the protection of personal data when collected, processed, and used for AI development. These laws offer a safety net to the data subjects against misuse and possible exploitation of their data. While personal data protection laws offer a great start in regulating the data used for AI technologies, these laws do not cover all aspects of AI. AI specific policies, laws, and regulations are important in creating a clear roadmap towards the use of AI in a manner that is transparent and supports innovation, but also in the protection of fundamental rights and freedoms of the people and protection against exploitation and misuse not only by developers during their data collection and mining process but also by end-users.

The AU through its ministers for communication, and information and communication technologies (CICT) have taken steps towards the development of AI specific legislations for the African region by adopting the 2019 Sharm El Sheikh Declaration that puts special focus on the African Digital Transformation Strategy (DTS). The Digital Transformation Strategy for Africa is based on:

• Foundation pillars: Enabling Environment, Policy and Regulation, Digital Infrastructure, Digital Skills, and Human Capacity, Digital Innovation and Entrepreneurship;
• Critical sectors: Digital Industry, Digital Trade, and Financial Services, Digital Government, Digital Education, Digital Health, Digital Agriculture) to drive the digital transformation;
• Cross cutting themes: Digital Content & Applications, Digital ID, Emerging Technologies, Cybersecurity, Privacy and Personal Data Protection, Research and Development to support the digital ecosystem.

The DTS also includes policy recommendations and actions under each foundational pillar, critical sectors, and cross-cutting themes. The declaration requested member states among others to, establish a working group on AI to study:

• The creation of a common African stance on Artificial Intelligence (AI)
• The development of an Africa wide capacity building framework
• Establishment of an AI think tank to assess and recommend projects to collaborate on in line with Agenda 2063 and the UN Sustainable Development Goals.

Further, African UNESCO member states joined in the consultations with UNESCO for the first draft of a Recommendation on the Ethics of AI, to be submitted to UNESCO Member States for adoption in November 2021. If adopted, the Recommendations will be the first global normative instrument to address the impact of the development and application of AI.

Conclusion

Overall, for Africa to reap the benefits of AI, it will be important to reflect upon the policy implementations and changes that need to be considered to foster innovation and non- stifling oversight; the impact anticipated from the use of AI technologies and the benefits that will be derived from the use of AI both short term and long term. The oversights that relate to the collection, processing storage and use of data for AI must be developed in such a way that is both relevant and applicable to the African context socially, politically and economically.  Policy makers must ensure that the regulations are structured in a manner that protects information from leakages, misuse and exploitation and protects the rights of its people.

References:

Artificial Intelligence (AI) and Education

Summary

AI Laws on Data Processing

Summary

Artificial Intelligence (AI) in Kenya’s Banking Sector

Summary

The New Wave of eHealth: Artificial Intelligence (AI) and Privacy Concerns? A Case Study of Kenya

Summary

This page contains a running list of our past and current publications and engagements relevant to AI.

1. WeRobots Conference

Held virtually in 2020, WeRobots was sponsored by the University of Ottawa and the Center for Law, Technology, and Society. We presented the following paper:

Title: Towards Integration of African Priorities into Artificial Intelligence (AI) Policy Agendas.

Abstract: Artificial intelligence (AI) technologies and applications present both profound opportunities and significant challenges for economic, social, and cultural development in communities, nations and regions on the African continent. There is growing engagement with AI matters by African innovators, researchers and institutions, with support from international technology firms and international development partners. But the African AI policy discourse is still only embryonic. Debates on how policy, law and regulation should address AI’s numerous dimensions, including those that apply to Africa, have to date been dominated by Western perspectives. In support of the need for greater African-centric policy engagement with AI matters, this paper provides:

1. An introduction to AI’s definitions, implications, and current manifestations in Africa.
2. An exploration of AI’s implications in Africa in respect of equity and labour.
3. An overview of some developed-world AI policy responses.
4. Review of some of the initial activities, and principles at play, in African AI policymaking contexts.

2. AI4D Network

The diversity of the African continent demands a diversity of representation. In 2019 and in collaboration with IDRC, CIPIT hosted the first gathering of the AI4D Network. The output of that event can be found here. https://www.k4all.org/event/ai4dworkshop.

3. WIPO call for public input — Conversations on Intellectual Property and AI.

In response to the World Intellectual Property Office’s conversations on IP and AI, CIPIT submitted three letters with perspectives on IP and AI from the Global South.

• CIPIT WIPO IP and AI – 21/10/2020
• CIPIT WIPO IP and AI – 24/07/2020
• WIPO Response to a call on impact of AI on IP policy

Our submission to the World Intellectual Property Organization (WIPO) as they seek consultation on issues affecting Intellectual Property and Artificial Intelligence.

The speaker is Ms. Wanjiru Muchiri, a Research Fellow and IP expert at CIPIT. The full video can be found here.

Funding for this work is provided through the AI4D project of IDRC.

We are part of the Alliance for Inclusive Algorithms.

We are the East Africa hub for the Open African Innovation Research (Open AIR) Network.