Home Cyberpsychology & Technology How to Implement GPT-4 Model Usage in Africa

How to Implement GPT-4 Model Usage in Africa

Reading Time: 6 minutes

During a lecture hosted by the United Nations University Institute for Natural Resources in Africa (UNU-INRA), Nature Speaks: Artificial Intelligence and Growth, at the University of Ghana on 25 May 2023, Professor Tshilidzi Marwala, Rector of the UNU and former Vice-Chancellor of the University of Johannesburg, shared insights on how artificial intelligence can contribute to the achievement of the sustainable development goals in Africa. He asserted that a holistic approach is needed for usage.

Artificial intelligence systems, much like the internet, are swiftly evolving into a foundational infrastructure layer, poised to enact transformative change. Fueled by burgeoning computational capabilities, enhanced connectivity, and vast reservoirs of big data, AI holds promise for propelling progress towards the Sustainable Development Goals (SDGs). It can catalyse the emergence of new ventures, enhance food supply chains, revolutionise education delivery, and play a pivotal role in addressing urgent health and climate-related issues.

By 2030, AI use in general is projected to contribute a staggering $15.7 trillion to global GDP, with $6.6 trillion coming from increased productivity and $9.1 trillion from consumption effects. AI has the potential to fundamentally change the way businesses operate, drive innovation, and improve the lives of millions of people across Africa. Some of the key sectors that could benefit from AI include healthcare, agriculture, education, and finance.

In Africa, generative AI models are already being applied across multiple domains, particularly in climate change prediction, healthcare, clean energy forecasting, water management, economics, finance, and governance. The Generative Pre-trained Transformer (GPT-4) model, for instance, holds significant potential to transform patient care and information accessibility. Through my engagements with clients and patients, they have expressed how ChatGPT-3.5 and GPT-4 can enhance their medication adherence satisfaction and alleviate the workload of healthcare providers. This is achieved through the automation of repetitive tasks and the development of personalised treatment strategies.

Professor Marwala in South Africa revealed that “during Covid, AI was used to predict the peaks of the pandemic”. At the University of Johannesburg, scientists predicted the first wave of Covid in South Africa, before it had spread widely across the country. This AI enables predictions to be made with limited data. Using Bayesian inference with the compartmental SIR models, scientists were able to support public health policymakers in quantify the impact of government interventions, allowing them to plan ahead.

Development of new generative artificial intelligence models

With the development of new generative artificial intelligence models, such as the Generative Pre-trained Transformer (GPT-4), it has garnered considerable attention across various fields particularly healthcare, education, and industry. GPT-4 model for instance, is a natural language processing model that uses machine learning algorithms to generate human-like responses to user input. Its advanced language-modelling capabilities have been shown to enhance access to information, increase productivity, and streamline communication in various fields.

In my recent usage of this large language model (LLM), the GPT-4 in healthcare and academia, it practically has shown immense promise for transforming health-care delivery and rapidly being integrated into clinical practice and knowledge delivery, respectively. Also, it has been observed that several LLM-based pilot programmes are underway in hospitals, and clinicians have begun using GPT-4 to communicate with patients and draft clinical notes in healthcare. While LLM-based tools are being rapidly developed to automate administrative or documentation tasks, many clinicians for instance, also envision using LLMs for clinical decision support and for creating personalised treatment plans.

Though LLM-based tools have shown great potential, there is also cause for concern about using LLMs. Extensive reports by OpenAI have shown the potential for language models to encode and perpetuate societal biases. Because, language models like the GPT-4 are typically trained using a vast corpora of human-generated text to predict subsequent text on the basis of the preceding words,. It can also reinforce structural inequalities and bias, perpetuate gender imbalances, threaten jobs, and facilitate oppressive government surveillance. Some biases, once identified, can be addressed via additional targeted training through a process called reinforcement learning with human feedback. This is a human driven process that can be imperfect and even introduce its own biases.

Encoded biases can lead to poorer performance for historically marginalised or under-represented groups in Africa. For instance, it was reported that LLM trained on clinical notes for clinical and operational tasks, predictions of 30 day readmission were significantly worse for black patients than for other demographic groups. Despite its vast potential, the adoption and implementation of GPT-4 in Africa faces several challenges, including a lack of relevant technical skills, inadequate basic and digital infrastructure, insufficient investment in research and development, and a need for more flexible and dynamic regulatory systems. Research is underway to solve these challenges.

The increased use of the GPT-4 model has further raised ethical and economic concerns regarding job displacement, privacy, and accountability, not just in Africa alone but worldwide. The policies and practices put in place today can shape the benefits and harms of AI in the decades to come. To contribute to these efforts, it was reported that IDRC and the Swedish International Development Cooperation Agency launched the Artificial Intelligence for Development in Africa (AI4D Africa) programme in 2020.

Though, this four-year CA$20 million partnership seeks to address general challenges by supporting the African-led development of responsible and inclusive AI,. The programme will promote excellence in applied research as well as in applying AI technologies to solve development challenges and improve livelihoods for those living in poverty.

Integrating African perspectives in GPT-4 implementation strategies

Well, it can be said that, using AI has been proven to be the first step to addressing Africa’s infrastructure gap, as it allows policymakers to have better information and make more informed decisions about what infrastructure to build, where, and when – and to ensure that current projects can serve Africa’s growing population for the next 50–100 years.

It was earlier reported by African Centre for Economic Transformation (ACET) after ACET and Omdena worked with nearly 40 data scientists and machine learning (ML) experts from around the globe to identify machine learning tools and approaches to inform policy decisions. In the first AI challenge, which was on infrastructure, the data scientists created models and designed methodologies to determine the most needed infrastructure, the best locations, and the factors that will affect long-term economic impacts. The AI Challenge lasted ten weeks and used machine learning techniques with computer vision, natural language processing, and exploratory data analysis (EDA).

These tools allow new ways to source and utilise data on scales previously unavailable. For example, EDA identifies the best model to fit the data, and can prepare the right data for machine learning and AI algorithms, using a myriad of data sources, such as satellite images, socioeconomic data, climate and topological data, population and demographic data, Google Trends, Google business data, and social media data, to understand the aspirations, needs, and sentiments of the people living in the region. 

But in comparison, the potential use of GPT-4 in Africa can augment these efforts by providing advanced natural language understanding capabilities. GPT-4 could enhance the analysis of textual data from various sources, including social media, news articles, and government reports, to gain deeper insights into public sentiments, policy preferences, and socio-economic trends.

Apart from the considerable costs associated with integrating GPT-4 model technologies into economic development, there’s a crucial need to finance workforce retraining, enabling individuals to effectively utilise these technologies. Enhancing skills training in data science and integrating GPT-4 AI models into educational curricula, along with providing accessible platforms and tools, can facilitate the provision of AI-driven services in economic development.

Furthermore, to maximise the benefits of AI technologies, particularly the GPT-4 model, in Africa, it’s crucial to address various social, cultural, ethical, and gender-related considerations. These encompass challenges in education and the workforce, ethical implications of AI, legal hurdles, and socio-cultural factors. But it’s imperative to recognise that these challenges give rise to numerous additional sub-issues, further emphasising the complexity of the task at hand.

The ongoing debate surrounding AI ethics often overlooks the perspectives and contributions of Africa, despite the fact that AI applications like the GPT-4 model are likely to be utilised within African contexts. The ethical principles guiding this debate often stem from non-African cultural and moral frameworks, creating a disconnect between global AI ethics discussions and the realities of AI implementation in Africa. This omission can be seen as a form of epistemic injustice, unfairly marginalising African voices and perspectives as knowers.

Africa possesses rich philosophical and cultural traditions that can offer valuable insights into ethical principles for the design, development, and application of AI, including the latest GPT-4 model. To ensure that AI technologies, particularly the GPT-4 model, are sensitive to African socio-cultural contexts, it’s imperative to incorporate African perspectives.

This entails considering the values, interests, and moral traditions of African societies in the design and deployment of AI technologies. Therefore, there is a pressing need for global frameworks on responsible AI and GPT-4 model ethics that encompass diverse societal needs, concerns, and interests. By incorporating African perspectives into these frameworks, AI can better align with the realities and nuances of African societies, fostering more inclusive and ethically sound AI development and deployment.

African perspectives are and should be critical components of these frameworks in their development, deployment, and use. There are two major implications for this. First, it will mean the development of AI applications that respond to African needs, expectations, interests, values, and beliefs. Second, it will contribute to epistemic justice in the global AI ethics discourse.

Onah Caleb is a research assistant at Benue State University (Nigeria). He runs the blog KaylebsThought.


© Copyright 2014–2034 Psychreg Ltd