Artificial Intelligence (AI) has not just made a splash but a global tidal wave. Its potential to enhance the lives of people worldwide is staggering. From education and healthcare to agriculture and finance, AI has the power to revolutionise every sector.
However, to fully harness the benefits of this technological leap, countries must first address some fundamental issues.
The starting point is understanding the context in which AI is being embraced in different sectors across the world. What are the prerequisites for making AI interventions meaningful?
What does a nation need to develop and adopt AI systems and innovations?
How can organisations prepare themselves for AI? What is the role of the Global South in all this? What enabling conditions does AI require in terms of infrastructure, governance, financial investment, talent, education, research, value systems, mindsets and entrepreneurial skills? What should constitute a national AI ecosystem?
These challenges must be addressed before any serious and effective effort to develop and adopt AI can be contemplated. Attending to the basics is essential.
Keep Reading
- Time running out for SA-based Zimbos
- More woes for suspended Cottco boss
- Sally Mugabe renal unit disappears
- Epworth eyes town status
Basic infrastructure
AI-enabling infrastructure must exist, starting with basic infrastructure such as energy, communication, transportation, finance, water, and housing. More significantly, affordable, reliable and efficacious digital infrastructure must be in place. Digital infrastructure sits on top of basic infrastructure. It is inconceivable to have it without this foundational infrastructure.
Each basic infrastructure listed above is critical in supporting sustainable socio-economic development and technological innovation.
More significantly, basic infrastructure is essential for economic growth and productivity, enhanced connectivity and mobility, quality of life, public health, environmental sustainability, social inclusion, equity, and job creation. Basic infrastructure is the backbone of modern societies, providing essential services and enabling socio-economic progress.
It lays the foundation for the digital economy and, more specifically, AI-driven innovations and applications. Consequently, strategic investments in basic infrastructure are imperative in pursuit of the development and adoption of AI.
Countries and organisations must work together to build basic infrastructure, thus enabling their participation in the AI revolution.
Digital infrastructure
The development and adoption of AI require reliable and efficient connectivity, digital technology, and devices. Digital infrastructure entails the foundational technology components that support modern information and communication technologies (ICT) by facilitating digital communication, data processing, and online services.
All these foundational digital infrastructure-related matters must be addressed before any meaningful engagement with AI is envisaged.
As already intimated, digital infrastructure encompasses the physical and virtual resources required to collect, store, process, and transmit data and information. It is critical for enabling various digital services, applications, and processes to function efficiently. The essential elements of digital infrastructure can be summarised as networks, data centres, internet backbone, cloud computing platforms, cybersecurity systems, telecommunication, the Internet of Things (IoT), software, and digital applications. Developing and maintaining robust digital infrastructure are crucial preconditions for the AI revolution.
Investments, research and development in digital infrastructure are necessary. They bridge the digital divide, thus ensuring equitable access to digital services in general and AI opportunities in particular.
Digital infrastructure forms the foundation on which AI systems operate efficaciously. It ensures that AI applications can access the necessary data, computational resources, connectivity, scalability, security, and deployment capabilities for successful development and implementation. Additionally, advancements in digital infrastructure in the IT, digital, and communication industries are increasingly intertwined with basic infrastructure.
Thus, digital infrastructure can be used to drive the development of basic infrastructure. There is synergistic and mutual reinforcement.
“Compute Resources” infrastructure
Beyond general digital infrastructure, the AI-specific requirements are more exacting. In the context of AI, the term “compute” refers to the computational resources and processing power required to train and run AI models and algorithms. AI computations involve complex mathematical operations, statistical analysis, and optimisation techniques that necessitate significant computational resources.
The critical aspects of “compute” include training AI models, processing big data, hardware acceleration, distributed computing, inference and real-time processing, optimisation and hyperparameter tuning, and cost and efficiency considerations.
Hence, “compute resources” infrastructure refers to the components and capabilities within a computing system that are used to perform computations, process data, and execute tasks. These resources are essential for running AI applications, performing complex calculations, and managing digital workloads. Compute resources typically include hardware and software components that deliver AI computational power. Critical elements of compute resources include the Central Processing Unit (CPU), Graphics Processing Unit (GPU), Neural Processing Unit (NPU), Random Access Memory (RAM), Storage (hard disk and solid-state drives), Network Interface, Operating System (OS), Virtualisation and Containerisation Technologies, and Cloud Computing Services.
Compute resources infrastructure is fundamental to computing systems’ performance, scalability, and efficiency. Advances in technology continue to improve compute capabilities, enabling complex computations, high-performance computing, and innovative applications in the development and adoption of AI systems.
Of these compute resources, the GPU has been a game changer. A GPU is a specialised electronic circuit designed to dramatically speed up the rendering of images and graphics for display on computer screens. It is a highly efficient processor capable of performing parallel computations, making it an essential component in AI, Machine Learning, Deep Learning, scientific computing, and cryptocurrency mining.
Semi-conductor industry
The semi-conductor industry plays a crucial role in the AI revolution by developing and manufacturing advanced microchips and hardware components that power AI systems.
For a nation to play a meaningful role in developing AI systems, it has to develop its semi-conductor industry. For countries of the Global South to fully benefit from the AI revolution, they must become producers of AI products, not just users or consumers. African countries must participate in the generation of new AI knowledge and innovations. The starting point is being players in the semiconductor industry.
The semi-conductor industry contributes to the AI revolution in several fundamental ways. It designs and manufactures specialised hardware components optimised for AI workloads.
These include GPUs, NPUs, Tensor Processing Units (TPUs), Field-Programmable Gate Arrays (FPGAs), and Application-Specific Integrated Circuits (ASICs) designed specifically for accelerating AI computations. These chips are essential for training and inference tasks in Deep Learning and other AI algorithms.
AI models, especially Deep Learning models (such as Large Language models), require significant computational power to process large datasets and perform complex calculations.
The semi-conductor industry develops high-performance processors (CPUs, GPUs and NPUs) with increased processing cores, higher clock speeds, and efficient memory architectures to meet the computational demands of AI applications.
AI systems require fast and efficient memory and storage solutions to handle large datasets and enable rapid access to data during computations. The semiconductor industry develops advanced memory technologies such as High Bandwidth Memory (HBM), Non-Volatile Memory (e.g., NAND Flash), and Storage Class Memory (e.g., Optane) that are critical for AI workloads.
To accelerate AI computations, semiconductor companies develop AI accelerators and co-processors that offload specific tasks from general-purpose processors to specialised hardware. For example, GPUs excel at parallel computations required for neural network training, while TPUs are optimised for fast matrix operations used in Deep Learning.
Semiconductor companies like NVIDIA invest in AI chip design and architecture innovations to optimise hardware for specific AI algorithms and applications. This includes developing efficient neural network architectures, implementing hardware-accelerated inference engines, and integrating AI-specific instructions into processor designs. It is instructive to note that NVIDIA has a market cap of $2.26 trillion and is the third largest company in the world after Microsoft and Apple.
In summary, the semiconductor industry is the key driver of the AI revolution. Its continuous innovation in developing advanced microchips, specialised hardware accelerators, memory/storage solutions, and manufacturing processes is instrumental in developing AI technologies. These innovations enable faster, more efficient, and more capable AI systems.
Countries in the Global South must develop semiconductor industries. Without these, their accrued benefits from the AI revolution will be minimal. Economic advantages from AI are not only occasioned by adopting AI. No. They also come from developing AI systems and building the physical technologies that enable AI. The fact that one AI semiconductor company (NVIDIA) has a market cap comparable to the collective GDP of the entire African continent speaks volumes.
Energy/Power requirements
AI systems demand a lot of power – they are energy-intensive. In common parlance, energy and power are used interchangeably. However, while the concepts are related, they are distinct in physics and engineering. Energy is the capacity to do work.
It is a scalar quantity measured in joules (J) or kilowatt-hours (kWh). Power is the rate at which work is done or the rate at which energy is transferred or converted. It is a measure of how quickly energy is used or generated. Power is measured in watts (W) or megawatts (MW). Power is essential for understanding the efficiency and capability of systems and devices. Higher power means more energy can be transferred or used per unit of time.
Without enough power, it is inconceivable that a country can maximally participate in the development and adoption of AI. African countries such as Zimbabwe and South Africa are experiencing electricity shortages (characterised by load shedding) before the extensive adoption of AI systems. This means such economies cannot maximally enjoy the benefits of AI. The energy requirements of the technology and its innovations are quite extensive. Countries like SA and Zimbabwe must urgently address their energy deficit and move to an energy surplus before they can have any meaningful engagement with AI systems.
The energy requirements of AI systems can vary significantly based on the complexity of their tasks, the size of the models used, and the hardware infrastructure. However, energy is central. The size of AI models, especially in Deep Learning, has been increasing rapidly.
Larger models often require more computational power and energy to train and operate. For example, the GPT-3 model by OpenAI has 175 billion parameters, making it computationally expensive (requiring much energy) to train and use.
Training Deep Learning models involves running large amounts of data through complex neural networks, and this is typically the most energy-intensive phase. Due to its high computational demands, this process often requires specialised hardware like GPUs, TPUs and NPUs.
Once trained, AI models perform inference tasks, making predictions or processing data based on the learned patterns. The energy requirements for inference are generally lower than training.
However, they can still be significant, especially for real-time applications running on resource-constrained devices like smartphones or IoT devices.
The choice of hardware dramatically influences the energy efficiency of AI systems. Specialised chips like TPUs are designed to optimise certain AI workloads, offering higher performance with lower power consumption compared to traditional CPUs or GPUs.
This article is an excerpt from the upcoming Electrical Engineering book: Design and Analysis of Control Systems: Driving the Fourth Industrial Revolution by Mutambara. He is the Director and Full Professor of the Institute for the Future of Knowledge (IFK) at the University of Johannesburg.
Research into optimising AI algorithms and hardware for energy efficiency is ongoing. Techniques like model pruning (removing unnecessary parts of a model), quantisation (reducing the precision of calculations), and energy-aware scheduling can help reduce the energy footprint of AI systems without significantly sacrificing performance.
Countries in the Global South must radically invest in energy and power infrastructure. Without enough electricity, pontificating about AI strategies and policies constitute an exercise in futility.
Governance and Regulatory Framework
AI development and adoption can only be built based on clean, efficient, transparent and accountable governance. All these African countries riddled with corruption, incompetence, and mismanagement – in the public and private sectors – are not ready for the AI revolution. Good governance is crucial for responsible AI development, deployment, and use. An enabling AI regulatory framework is essential. Specific aspects foundational to AI include establishing ethical guidelines and principles for AI development and deployment, regulatory oversight, transparency and accountability, inclusive public engagement and participation, risk management and resilience.
Good AI governance frameworks must support capacity-building initiatives to enhance AI literacy among policymakers, regulators, and society. There is a need for continuous monitoring, evaluation, and updating of governance mechanisms to address emerging challenges, ensure compliance with evolving regulations, and maintain public trust in AI technologies.
Indeed, good governance and an enabling regulatory framework are essential for fostering ethical, transparent, and accountable AI ecosystems. By integrating principles of fairness, accountability, transparency, and inclusiveness, policymakers can promote innovation while safeguarding against potential risks and ensuring that AI technologies contribute positively to sustainable socio-economic development. Globally, national AI policies and regulations must be developed and harmonised in regional blocs and continents.
AI Ecosystem
An ecosystem approach is required to drive the development and adoption of AI. An AI ecosystem refers to a complex network of interconnected entities, including organisations, institutions, technologies, and individuals, that collaborate, interact, and influence each other within the field of
- This ecosystem encompasses various components that collectively contribute to the development, deployment, and utilisation of AI technologies.
The main constituents of an AI ecosystem include AI researchers and developers, technology companies and startups, financiers and investors (including venture capital firms), academic institutions, government agencies and regulators, industry partners and collaborators, data providers and data infrastructure. Data is the lifeblood of AI. The ecosystem relies on data providers, aggregators, and infrastructure such as data centres and cloud services to collect, store, manage, and analyse vast amounts of data for AI applications. Further participants in the ecosystem include ethics and policy think tanks, end users and consumers, and global collaborators and partners.
All these interconnected and interdependent entities form a dynamic AI ecosystem that evolves.
Technological breakthroughs, regulatory developments, market trends, societal demands, and ethical considerations influence the configuration. Effective collaboration and coordination among ecosystem stakeholders are essential for realising AI’s transformative potential while addressing challenges related to fairness, transparency, privacy, and societal impact. Every country must strive to establish a national AI ecosystem. These can then be expanded and integrated into regional and continental AI ecosystems.
Talent and Expertise
AI development and adoption require diverse talent and expertise across various disciplines. Building and deploying AI systems involves a multidisciplinary approach that integrates expertise in computer science, mathematics, data science, engineering, and domain-specific knowledge. Skills required for AI development and adoption include data scientists, Machine Learning engineers, software engineers, data engineers, AI researchers, domain experts (e.g., agriculture, education, healthcare, and finance), ethics and policy experts, prompt engineers, project managers, AI trainers/teachers and AI product managers. There is a need for dynamic and effective collaboration among these multiple disciplines for successful AI development and adoption.
Countries and institutions (private and public) must invest in developing AI talent, skills, and expertise in a transdisciplinary and interdisciplinary fashion. They must create environments that encourage innovation and knowledge sharing to leverage the full potential of AI technologies.
Organisational Preparedness
As all organisations prepare to embrace the AI revolution, they must seek to create the AI transformation roadmap – a treasure map to a new AI-driven entity. They must build a talent bench with the requisite expertise and skills. More significantly, organisations must realise that their current way of doing things will not cut it. They must develop and adopt new and innovative operating models that are fit for purpose in the AI context. Furthermore, organisations and countries must nurture decentralised and distributed technology environments that enable teams to innovate across institutions and sectors. As explained earlier, data is the key driver of AI systems.
Consequently, there must be efforts to embed data everywhere in institutions. There must be ubiquitous access to and use of this data. The proof of the pudding is in the eating. Countries and organisations must drive and unlock AI user adoption and enterprise scaling.
Financing and Investment
Financing and investment are critical in driving AI development and adoption by providing the necessary resources, funding, and support for research, innovation, and commercialisation. The AI industry requires significant investments to advance the technology, build scalable solutions, and integrate AI into various sectors. Specifically, financing and investments are essential for research and development, fostering and promoting innovations, startups and entrepreneurship, AI infrastructure and technology platforms (cloud computing, AI chips, data centres, and specialised hardware for accelerating AI computations).
The funding and investment from the private sector will be critical. Extremely expensive AI innovations, such as Large Language models, cannot be funded by universities or governments.
There is also a need to fund industry-specific applications, talent acquisition and skills development, ethical AI and governance, market adoption and commercialisation.
Indeed, financing and investment are catalysts for driving AI innovation, accelerating technology development, and fostering widespread adoption across sectors. All countries, industries, and businesses must pursue and promote strategic investments in AI to maximise its transformative impact.
Right Mindset
The right mindset is critical in the development and adoption of AI. First, there must be self-belief and confidence, solution orientation, a can-do mentality, entrepreneurial disposition, possibility thinking, an affinity for technology, multidisciplinarity, transdisciplinarity, and an enduring appetite for continuous learning.
A positive and open mindset encourages innovation and creativity in AI development. An experimental mindset leads to the early adoption of AI systems. Embracing new ideas and
approaches fosters breakthroughs in algorithms, models, and applications, leading to advancements in AI technology. The right mindset for AI must embrace a problem-solving orientation, where developers and adopters of AI focus on addressing real-world challenges and opportunities. A thoughtful, ethical mindset is essential to ensure responsible AI development and adoption. Being mindful of potential biases, fairness issues, privacy concerns, and societal impacts helps guide ethical decision-making throughout the AI lifecycle.
As already asserted, adaptability and continuous learning are critical. AI technologies evolve rapidly, and having a growth mindset encourages individuals and organisations to embrace change, attain new skills, and stay abreast of the latest innovations in AI. Developing and adopting AI often requires collaboration across diverse disciplines and stakeholders. Hence, an interdisciplinary and collaborative attitude within diverse knowledge development and sharing teams will be the most effective way to tackle complex AI challenges.
Adopting a user-centric mindset ensures that AI solutions are designed with end users in mind.
Prioritising user needs, experiences, and feedback leads to more intuitive, accessible, and user-friendly AI applications. The right mindset encourages calculated risk-taking and resilience in AI development. Innovation involves experimentation, and not being afraid of failure allows for iterative improvements and breakthroughs in AI technology.
When all is said and done, cultivating the right mindset among AI developers, researchers, policymakers, and industry leaders is essential for unleashing the transformative power of AI.
However, mindset cannot be legislated or decreed into existence. It has to be built and developed diligently and judiciously over time.
Basic Education and Literacy
Basic education and literacy are crucial in developing and adopting AI systems. Basic education lays the groundwork by equipping individuals with foundational skills in mathematics, statistics, computer science, and critical thinking – all essential for understanding AI concepts and applications.
Literacy and competence in Science, Technology, Engineering and Mathematics (STEM) subjects are fundamental to nurturing the next generation of AI researchers, data scientists, engineers, and developers. Basic education provides the necessary knowledge and skills to pursue careers in AI-related fields. Education fosters AI literacy among the general population, enabling individuals to understand AI systems’ capabilities, limitations, and societal implications. AI literacy enables and empowers individuals to make informed decisions about AI adoption and usage. Basic education instils a culture of lifelong learning, which is essential in the rapidly evolving field of AI.
Furthermore, education promotes ethical awareness and responsible behaviour in AI development and adoption. Basic education emphasises ethical principles such as fairness, transparency, accountability, and privacy, which are critical for ethical AI governance. Literacy and education foster creativity and innovation, enabling individuals to apply AI technologies to solve complex problems in various domains such as agriculture, mining, healthcare, finance, agriculture, and environmental sustainability.
Access to quality education and literacy programmes reduces socio-economic disparities in AI adoption. Providing equitable educational opportunities ensures that individuals from diverse backgrounds can participate in and benefit from the AI-driven economy. AI is transforming industries and reshaping job roles. Basic education equips individuals with adaptable skills and competencies needed to thrive in AI-enabled workplaces, fostering career readiness and employability. Education empowers citizens to engage in informed discussions and policymaking about AI governance, regulation, and societal impact. AI-literate individuals can advocate for ethical AI practices and contribute to shaping responsible AI policies.
Basic education and literacy are essential enablers for AI development and adoption, as they foster foundational skills, promote AI literacy, nurture ethical awareness, and prepare individuals for the future of work.
Concluding Remarks
Countries and organisations must not put the cart before the horse when developing and adopting
- They must address critical pillars such as basic infrastructure, digital infrastructure, governance and regulatory framework, financing and investment, basic education and literacy, talent and expertise, energy/power, the semiconductor industry, “compute” resources infrastructure, organisational preparedness, the AI ecosystem, and the right mindset. These related, interdependent, and mutually reinforcing foundational matters must be attended to in a holistic, systemic, and structured manner. There must be an AI national vision, strategy, implementation plan and policy framework, all premised on resolving the above challenges. Furthermore, AI regional, continental and global partnerships must be developed and leveraged.
Indeed, AI can drive all economic sectors and improve the quality of life for all of Earth’s inhabitants.
However, the journey to the Promised Land must start with the basics.
This article is an excerpt from the upcoming Electrical Engineering book: Design and Analysis of Control Systems: Driving the Fourth Industrial Revolution by Mutambara. He is the Director and Full Professor of the Institute for the Future of Knowledge (IFK) at the University of Johannesburg.