Stellenbosch University
Welcome to Stellenbosch University
Africa needs fundamental research in machine learning
Author: Wiida Fourie-Basson
Published: 11/09/2018

Fundamental research in Artificial Intelligence and machine learning is essential if Africa wants to have control over some of the really big ideas in the world.

This is the message from Dr Nando de Freitas, one of the world's leading thinkers in machine learning and principal scientist at Deepmind. This South African presented the keynote address at the 2018 Deep Learning Indaba that is taking place at Stellenbosch University from 9 to 14 September 2018 this week.

The Deep Learning Indaba is a volunteer-driven grassroots organization whose aim it is to build pan-African capacity in Artificial Intelligence by creating communities, building leadership and recognizing excellence across the continent. During the week, more than 500 students and researches from 30 African countries and 19 international countries are exposed to several of the world's leading thinkers in this field.

Dr De Freitas said it is still very difficult for machines to recognize shapes, such as that of Table Mountain, or to emulate even the most basic of human actions, such as pouring milk into a cup, which is something a three-year-old also has to master: “Hardcoding has its limits, and it is very hard to get machines to do what human minds can do. We are trying to do this the machine learning way, but for that we need massive datasets of images and comparative training data sets. This takes a lot of effort and ingenuity. It cannot be taught. You have to figure it out as you go along," he explained.

However, most of the new frontiers in machine learning, such as reinforcement learning, meta learning and supervised learning, requires fundamental research: “Fundamental research is important, even for Africa. Because it is from fundamental research that the big ideas emerge. It is not only about applications. Only when you have control over the big idea, will you be able to build things. The rest is trial and error."

He also emphasized the ethical aspects of machine learning: “We will never have ethical machines. There will always be some kind of bias. But it is essential that we all have access to the benefits that technology will bring," he concluded.

The next plenary speaker was Dr Mustapha Cissé, head of the Google AI Lab in Ghana, and an expert on fairness, transparency and reliability in machine learning.

He said there are important problems in Africa and the developing world that we want to solve. But there is an inherent bias due to the fact that the community solving the problems are doing it from the available but limited datasets: “It is important that we are aware of what we call the 'white guy problem', and that we pay more attention to the impact we could have on the a global level." He presented two sessions on the Fundamentals of Machine Learning at the Indaba.

Prof Louise Warnich, Dean of the Faculty of Science at Stellenbosch University, said in her welcoming address that the Deep Learning Indaba could not have happened at a more ideal time, as South Africa's Department of Science and Technology has just released a new White Paper that aims to shake-up South Africa's science and research landscape:

“The fourth industrial revolution is predicted to disrupt industries and transform production, management and governance systems. Will we here in Africa be able to make the technological leap, as we did a decade ago from landlines to mobile technology? Will Africa and Africans benefit from the cutting-edge advances in artificial intelligence, data science and machine learning, while at the same time addressing Africa's unique challenges? Or should I rather say, can we afford not to?"

Dr Shakir Mohamed, one of the lead organisers of the Indaba, said the Indaba aims to work with local universities and to strengthen local initiatives, so that students do not need to leave Africa to learn from the best.

Other key events this week include a Women in Machine Learning event sponsored by Microsoft, the inaugural Kambule and Maathai awards that recognise excellence in the research and application of machine learning in Africa, and the closing keynote address to be delivered by Dr Jeff Dean, global head of Google AI, and Google Senior Fellow, on Friday, 14 September.

For media enquiries, contact