RAIL KNUST PRESENTS AT THE RESPONSIBLE AI NETWORK – AFRICA WORKSHOP ON RESPONSIBLE AI APPLICATIONS IN AFRICA
The Responsible AI Lab (RAIL) gave a presentation at the Responsible AI Network (RAIN), Africa workshop on the theme “Responsible AI Applications in Africa” on 7th April 2022 via zoom. Three panellists from Ghana, Uganda, and Burkina Faso presented their current research on applying AI and Machine Learning in the African context.
The workshop was jointly moderated by Prof Jerry John Kponyo, the Scientific Director of RAIL-KNUST, Ghana and Dr. Caitlin Corrigan of Institute of Ethics in Artificial Intelligence (IEAI) at the Technical University of Munich, Germany.
Artificial Intelligence (AI) is a theory and development of computer systems that can perform tasks that typically require human intelligence. Speech recognition, decision-making, visual perception are features of human intelligence that artificial intelligence may possess.
The fields of AI include: Natural Language Processing (NLP) and Computer Vision. These fields can be applied to several areas such as health, agriculture, commerce, industries and transportation.
Dr Rose Nikibuule, a Lecturer in the Department of Computer Science, Makerere University, Uganda, presented on the application of AI in Agriculture which tackles the issue of pests and provides timely information to farmers regarding the weather in Uganda. She stated that the farmers are involved in the research processes and data collection from the onset. She mentioned “bias” as one difficulty she faces with the farmers, where the farmers do not accept the AI technology due to the misconception of AI being the replacement of their work and impact.
Dr. Sadouanouan Malo, a coordinator of the Master in Data Sciences program and the lead of the Data Intelligibility research team at the Computer Science Higher School (ESI), Nazi Boni University (Bobo-Dioulasso, Burkina Faso), showed how the growth of population is affecting the production and distribution of energy and the need to adopt AI. He then touched on how his researchers use AI for electrical energy management and optimization in Burkina Faso. By obtaining historic data on energy consumption and usage trends, they can use AI to predict future energy demands and usage trends and patterns. He added that a perfect symbiosis of AI and fossil and renewable energy industries would increase yield.
Ing. Dr. Henry Nunoo-Mensah, a Professional Engineer and Lecturer in the Department of Computer Engineering, KNUST and the Programmes Coordinator for the Responsible AI Lab (RAIL) at KNUST, presented on AI applications in health. His presentation covered why AI should be responsible, responsible AI considerations, responsible AI pipelines, AI application in health and responsible AI tools.
He mentioned that AI should be responsible for ensuring that solutions are delivered with integrity, equity and social impact. He stated a need to evaluate AI solutions according to established standards (ISO 26000) and frameworks (FACT, FATE, FACETS). Dr. Nunoo-Mensah said these considerations tend to encourage the building of AI systems that are responsible and trustworthy. He mentioned that AI should impact the society and environment positively through fairness, accountability, equity, data privacy and safety. He then touched on a responsible AI pipeline that involves Envisioning, Data gathering, Modeling and Deployment Stages, respectively and the various responsible considerations at each stage of the pipeline.
Dr. Nunoo-Mensah stated that his current AI research in health includes: histopathological breast cancer diagnoses, assistive video conferencing tool for hearing and speech impaired, medical translation of underserved languages in Ghanaian consulting rooms and chronic kidney disease diagnoses.
He highlighted that there are numerous tools available to aid responsible AI development, and they include, but not limited to TensorFlow Responsible AI Toolkit, CodeCarbon and GAMMA-FACET. In addition, participants inquired about the gap in the technologies or applications used by the panellists. The panellists mentioned that data availability is one major challenge coupled with the acceptance of AI technology in indigenous areas. They stressed the need for a centralized data repository, funding natural language processing (NLP) AI projects for indigenous African Dialects and strengthening collaboration among researchers. Finally, the panellists asserted that AI ethics should always be spoken of when AI is discussed and should be introduced in curriculums for teaching and learning.
