39624 research outputs found
Sort by
Perceptions of artificial intelligence (AI) professionals in South Africa on the ethical implications of AI
No full textA research report submitted in fulfillment of the requirements for the Master of Management in the field of Digital Business, in the Faculty of Commerce Law and Management, Wits Business School, University of the Witwatersrand, Johannesburg, 2025This research explores the ethical implications of artificial intelligence (AI) adoption in South Africa, focusing on the private sector. Using a qualitative approach rooted in interpretivism and guided by thematic analysis, data was collected through semi-structured interviews with industry professionals. The study aimed to uncover how AI professionals perceive and respond to ethical considerations in the design and deployment of AI technologies across various industries. Key ethical challenges identified include algorithmic bias, lack of transparency, data privacy concerns, and unequal access to AI technologies. Participants highlighted inconsistencies in ethical awareness, an absence of formalised guidelines, and a gap between personal values and institutional priorities. The study also found that regulatory ambiguity and profit-driven goals often overshadow ethical intentions, despite professionals recognising the societal impact. The research proposes practical recommendations, including the establishment of inclusive AI regulatory frameworks, structured ethics training for developers, and the development of transparent governance models tailored to the South African context. These findings contribute to a deeper understanding of responsible AI integration and provide a foundation for future policy, education, and industry-wide ethical standards.MM202
Generalized Task Learning for Robots: Unifying Task Hierarchies through Contrastive Learning
No full textA dissertation submitted in partial fulfilment of the requirements for the degree of Master of Science, to the Faculty of Science, School of Computer Science & Applied Mathematics, University of the Witwatersrand, Johannesburg,This dissertation addresses the challenge of enabling robots to generalize across unseen household tasks by learning abstract task structures from demonstration data. We develop a three-stage pipeline that translates natural language instructions and demonstrations into hierarchical task representations using large language models, clustering, and parameterized generalization. Our approach is tested and evaluated on the ALFRED benchmark [Shridhar et al. 2020]. ALFRED acts as a standardized measure used for training models to comprehend and follow instructions in natural language. It leverages first-person perspective visual input to carry out a series of actions for various household tasks. While this approach doesn’t represent the state-of-the-art, it establishes a foundation for future research to build upon.MMM202
Risk and Security of Digital-only banks: perspective from Digital natives in South Africa
No full textA research report submitted in fulfillment of the requirements for the Master of Management in the field of Digital Business, in the Faculty of Commerce Law and Management, Wits Business School, University of the Witwatersrand, Johannesburg, 2025The increasing adoption of technology has transformed various industries, including banking, where traditional services are now accessible through digital platforms. In South Africa, digital-only banks operate without physical branches, relying solely on digital infrastructure. While this presents opportunities, it also introduces risks such as cyberattacks, data breaches, and financial losses. Understanding these risks is crucial to fostering adoption, particularly among digitally active individuals. This study addresses a gap in existing literature, which largely focuses on general adoption factors rather than the security-specific concerns of digital natives in South Africa. Grounded in the Technology Acceptance Model (TAM) and Perceived Risk Theory (PRT), the research explores security and risk as key determinants of adoption. It also investigates measures that consumers and digital-only banks can implement to enhance security. A qualitative approach was employed, using semi-structured interviews with eleven digital natives selected through purposive sampling. Findings reveal that perceived security risks remain a key barrier to adoption, despite advancements in digital banking security. Threats such as social engineering and phishing persist due to low user awareness, while infrastructure challenges in underserved areas further hinder adoption. Despite perceptions of convenience and usability, digital-only banks are not widely seen as cost-effective. To enhance trust and adoption, digital-only banks must prioritise security awareness, transparency, and collaboration with stakeholders to address infrastructure gaps and cybersecurity threats. Strengthening consumer education and adopting advanced security technologies are essential for ensuring responsible digital banking adoption in South Africa.MM202
Investigation of Wake Redirection Techniques for Wind Farms
No full textA research report submitted in fulfillment of the requirements for the Master of Science in Engineering, in the Faculty of Engineering and the Built Environment, School of Mechanical, Industrial and Aeronautical Engineering, University of the Witwatersrand, Johannesburg, 2025A world-wide increase in wind energy has seen an increase in the need for optimisation of wind farms. In a wind farm, an upstream turbine has several negative effects on the downstream turbine. The upstream turbine extracts a lot of the energy from the freestream, leaving little energy remaining in the wake to be extracted by the downstream turbine. There is also a significantly higher turbulence intensity in the wake leading to uneven loads being exerted. This all results in reduced power outputs of the wind farm as a whole. Wake redirection techniques, and specifically yaw misalignment, could reduce these negative effects and improve the overall power output of a wind farm. The first aim of this study is to investigate the ability of a surrogate model to predict the flow field when the upstream turbine is at different yaw angles. This could assist with real-time wake analysis and optimisation of a wind farm, without the need for running high-fidelity simulations to predict the wake when the yaw angle of the turbine is changed. The second aim is to investigate the physical effect of the upstream turbine yaw on the wake, as well as the interaction between the wake and the downstream turbine. A surrogate model is a combination of a Reduced Order Model (ROM) and a Machine Learning (ML) algorithm. A ROM was created using an existing, open-source dataset that conducted Large Eddy Simulations (LES) on a two-turbine array using the Simulator for Wind Farm Applications (SOWFA) software. The Dynamic Mode Decomposition (DMD) modes generated during this process were grouped according to the physical property of the wake that they represent. These modes were used to train the ML model to predict the development of individual modes at different yaw angles, which were then reconstructed in groups or combinations of groups to predict the entire flow field at the different yaw angles. It was found that the mode group related to the mean flow is essential to reconstruction and it alone results in the most accurate reconstructions with an average error of 7% and a reduction in computational resources by a factor of approximately 300 compared to LES. Smoke visualisation, hot-wire measurements and force tests were conducted to further investigate the physical properties of the wake. A difference of 8.1% was recorded between the experimentally measured maximum angle of deflection and the maximum generated from the surrogate model. It was found that the larger the yaw angle of the upstream turbine, the greater the wake deflection. This relationship is hypothesised to be linear until the tested yaw angle of 45°, however it is expected to plateau after it reaches its stall condition. As yaw increases, both the force in the horizontal direction (Fy) and the side toppling moment (Mx) decrease linearly with the yaw angle, suggesting that a larger yaw angle also helps reduce the structural loads on the downstream turbine.MM202
An exploratory study into sustainability strategies adopted by Spaza Shops in the face of challenges
No full textA research report submitted in fulfillment of the requirements for the School of Economics and Finance, in the Faculty of Commerce Law and Management, School of Economics and Finance, University of the Witwatersrand, Johannesburg, 2025This study was carried out to determine sustainability strategies used by the spaza shops in the townships of Diepsloot, Cosmo City, and Alexandra. The spaza shop market is saturated and most spaza shops fail because of external and internal challenges. A fight for survival is paramount for such businesses and there is a great need for effective sustainable strategies. With very few studies on the success stories of these informal retailers, it is unclear how spaza businesses remain viable in the face of obstacles caused by both internal and external factors. Other studies have examined the sustainability issues affecting spaza retailers, but the goal of this study is to discover the strategies they use to stand out in a sector that is already saturated with competitors. This study was qualitative and exploratory in nature. To get more insightful data, spaza shops that have been operating for more than two years were chosen using a purposive sampling technique. Thirteen spaza shop operators participated in semi-structured interviews to gather information. The findings highlight the following strategies as crucial to the sustainability of the spaza shops: capitalising on business resources: marketing stratagems; selling selected products; having a competitive advantage; and partnerships with other businesses. The neighborhood crime watch associations that the business community established have been instrumental in reducing crime in the townships, which has allowed operators of spaza stores to increase operating hours. The spaza shops reported a reduction in loss of resources due to burglary and theft. The researcher recommends that business management and entrepreneurial support be acquired by spaza shop operators, the government should outsource skills development training for these retailers, and more partnership initiatives should be encouraged to promote sustainability.MM202
The relationship between coping and psychological wellbeing among South African Muslims during Covid19 pandemic
No full textThe Spectrum of Skin Biopsy Proven Cutaneous Invasive Fungal Infections in Patients at Johannesburg Academic Hospitals
No full textA research report submitted in fulfillment of the requirements for the Master of Medicine, in the Faculty of Health Sciences, School of Medicine, University of the Witwatersrand, Johannesburg, 2025Background Fungal infections affect a large percentage of the population, particularly in the setting of HIV (human immunodeficiency virus). It is not well described in low- and middle- income settings. 1 We aimed to describe the aetiology, demographics, and clinical manifestations of invasive fungal infection in Johannesburg, South Africa. Methods We conducted a retrospective, descriptive review of skin biopsies done by dermatologists at Johannesburg Academic Hospitals from January 2019 to December 2021. Biopsies done on patients 18 years, reviewed by the Department of Anatomical Pathology, with the diagnosis of invasive fungal infections, were included. Data was obtained from National Health Laboratory Service (NHLS) reports and patient files. Demographics, HIV status, certain serum tests and biopsy results was collected. Results 60 patients were included with a median age was 40.8 years (median=39; IQR (Interquartile Range) =36-48). The majority of patients were males (66.7%). 46 patients were HIV positive (76.7%). Only 17 (37%) were on ARVs (antiretrovirals). In the HIV positive subgroup, the median CD4 count was 35.5 cells/L (IQR 8-208) with a median VL (viral load) of 102 copies/mL (IQR 0-276 000). The head and neck areas were predominantly involved (40%). Papules were the most common type of lesions (43.3%). Cryptococcosis (25.7%) and histoplasmosis/emergomycosis (27.3%) were - 13 - the most common infections identified. Anaemia was the most predominant abnormality found on laboratory results (67.9%). Conclusion Histoplasmosis/emergomycosis and cryptococcosis in HIV positive patients were the most frequently encountered invasive fungal infections in our study population. Untreated/poorly controlled HIV poses a major risk factor for developing invasive fungal infections.MM202
Photonic circuits with light patterns
No full textA dissertation submitted in fulfillment of the requirements for the degree of Masters of Science, to the Faculty of Science, School of Physics, University of the Witwatersrand, Johannesburg, 2025Optical circuits leveraging structured light—the precise control of light’s degrees of freedom (DOFs), including transverse space (i.e amplitude and phase), frequency, and polarization—offer a promising platform for quantum information processing tasks. Information can be encoded in high dimensional states, and different degrees of freedom can be coupled creating hybrid structures for complex operations. Recent advancements in tunability techniques, particularly through spatial light modulators (SLMs), have enabled the precise encoding of light patterns as holograms. This capability facilitates complex, programmable operations with applications in imaging, cryptography, and communications, making SLMs indispensable tools for manipulating high-dimensional optical fields. Building on these advancements, this dissertation explores how SLM-powered optical circuits can encode unitary operations by dynamically shaping light patterns to perform matrix multiplications—an essential component of quantum computation. The core contribution of this work is the demonstration of optical matrix multiplication as a means to emulate quantum computations, where quantum states are encoded as spatial light patterns and unitary operations are implemented via phase modulation. This approach exploits the inherent parallelism of the transverse spatial degree of freedom to efficiently simulate quantum algorithms within an optical framework. By establishing a framework for optical circuits with structured light patterns, this dissertation provides a pathway for leveraging classical optical systems to simulate quantum computations. Inspired by Jozsa’s insight that classical waves can efficiently mimic quantum algorithms without requiring entanglement, these findings contribute to the growing field of hybrid classical-quantum computing and highlight the potential of optical platforms in advancing scalable quantum technologies.Council for Scientific and Industrial Research (CSIR)MMM202
