One Topic, One Loop: John Jerry Kponyo
In the third part of our global discourse series "One Topic, One Loop", Prof. John Jerry Kponyo from Kwame Nkrumah’ University of Science and Technology (KNUST) in Ghana explores the concept of responsible AI. He emphasizes the importance of ethical, transparent, and inclusive approaches, describes how Afrocentric datasets are generated at KNUST’s Responsible AI Lab, and ends with a question for Prof. Sune Lehmann of the Technical University of Denmark.Artificial Intelligence touches every facet of the human condition. It is having enormous impact on our workforce, making us more productive, efficient, and impactful. However, AI comes with a host of other unintended consequences, including the perpetuation of existing stereotypes against minorities and the ease of misinformation.
The need for responsible AI
In order to design AI solutions for the public good it is important to consider a set of principles that ensure the ethical, transparent, and accountable use of AI technologies consistent with user expectations, organisational values, and societal laws and norms. For us at the Responsible AI Lab at KNUST, responsible AI refers to the practice of designing, developing, and deploying artificial intelligence in an ethical manner. This means ensuring that AI solutions are delivered with integrity, equity, and respect for individuals, and that developers of AI solutions are always mindful of the social impact of what they are building. This is because AI systems are fundamentally socio-technical, including the social context in which they are developed, used, and acted upon, with its diversity of stakeholders, institutions, cultures, norms, and spaces. In sum, responsible AI is human-centered AI.Responsible AI understands how to maintain public safety, how to prevent harm against minorities, and how to ensure equal opportunity. This requires AI systems that operate reliably and with low error rates , minimising risks, and preventing potential harm to individuals and society. Additionally, responsible implementation of AI is crucial to avoid reinforcing biases or discriminating against minority groups, and to promote fairness and inclusivity. By adhering to ethical principles and rigorous testing, AI can mitigate biases and promote equitable outcomes.
Generating Afrocentric datasets
While STEM professionals build AI systems, responsible AI is a multidisciplinary field and requires a variety of essential contributors to ensure a holistic and comprehensive approach to AI development. Computer scientists and engineers could focus on creating transparent, fair, and unbiased AI systems, while data scientists and statisticians ensure data integrity and develop methods to counter bias.Legal experts are also needed, as are social scientists, economists, and humanists, as well as communications and cybersecurity experts. Only a broad, multidisciplinary approach can succeed in developing AI that benefits society while minimizing harm.
In the Responsible AI Lab at KNUST, our core priorities revolve around the creation of Afrocentric datasets and the development of solutions specifically catered to the African continent. This commitment arises from recognizing the fact that Africa has been left behind in the AI dialogues. By curating datasets that accurately represent the diverse African context and crafting AI solutions that directly address the continent’s challenges, we aim to bridge the gap and contribute to meaningful technological progress. Our emphasis on Afrocentric datasets and solutions reflects the need to ensure that AI discussions encompass and benefit the unique realities of the African continent.
Making Generative AI responsible
To design Generative AI more responsibly, it is important to foster further research aimed at comprehensively assessing the efficacy of Generative AI across a range of contexts, including critical settings like classrooms and hospitals. Currently, there is a lack of quantitative research that delves into the impact of Generative AI on education, learning outcomes, and its potential to assist individuals with learning disabilities.The establishment of ethical guidelines stands as a critical pillar in the responsible advancement of AI technologies. Collaboration between local governments, academia, industry stakeholders, and international organizations plays a key role in formulating ethical standards that span domains like data privacy, bias mitigation, transparency, and accountability. Education, too, occupies a central position in the responsible use of AI systems. The ethical application of AI should be inculcated into the educational curriculum, running in parallel with the teaching of technical skills. Addressing key issues such as fairness, accountability, confidentiality, ethics, transparency, and safety is vital across all disciplines.
At this point, I would like to hand over to Prof. Sune Lehmann with the question of what data sets we need to ensure Responsible AI.