Human touch boosts Africa-Europe AI research cluster

AFRICA-EUROPE

The Africa-Europe Cluster of Research Excellence (CoRE) on Artificial Intelligence (AI) may be focused on ‘computational thinking’, but it has the ‘human touch’ to thank for its progress so far.

Less than a year after it was established in June 2023, the cluster has defined an impressive list of research topics, has been hosting regular masterclasses and webinars, and is planning to produce PhD graduates at a considerable rate.

And it all started with a deliberate decision that those involved should bond – both as people and as academics.

“Science is a human activity. People working together have to develop personal relationships and learn to trust each other. And for that, you need human connections,” Professor Francesco Petruccione, co-leader of the cluster, told University World News. He is based at (Stellenbosch University) in South Africa, where he specialises in quantum computing.

Professor Francesco Petruccione, co-leader of the Africa-Europe Cluster of Research Excellence in Artificial Intelligence. Photo: Stefan Els, Stellenbosch University.

“So, from the beginning, I insisted we meet every week for an hour, even if only virtually. Initially, we just chatted over a cup of coffee, sharing ideas, but that’s how we got to know each other better.”

His fellow co-leader, Professor David Sumpter, an applied mathematician at Uppsala University) in Sweden, takes up the theme: “Our summers up here are short and therefore sacred. So, when Francesco said we must meet every week, I had my doubts. But getting together regularly has been very useful.”

Professor David Sumpter, co-leader of the Africa-Europe Cluster of Research Excellence in Artificial Intelligence. Photo: Supplied.

The cluster continued this momentum with an in-person workshop in Uppsala in November last year, attended by representatives of most of the nine African and six European institutions making up the structure.

“A bunch of visiting PhD students came along as well – all women, making for a nice change in IT circles. And while we planned all day, they got together with our students, laying a foundation for the future,” Sumpter said.

Research topics

It was at this event that the cluster agreed on research topics, including speech recognition and translation models for African languages, computational approaches to drug discovery, using AI in conservation and agriculture, data-driven disease mapping, and revolutionising sport data with machine learning.

In its research brochure, the cluster explains: “AI methods have become powerful tools for addressing societal challenges in almost all areas of life. This is fuelled by improved connectivity, increased computational power and data access.

“Applications of AI will play an essential role in tackling problems in a multitude of areas, like climate change, food security, education, public health, access to energy and clean water, and social justice, to name just a few.

“The cluster aims to strengthen the capabilities for research and postgraduate training in these areas at key universities in Africa and Europe, as well as initiate research projects in areas of global relevance.”

Participants

In Africa, the universities of Lagos, Nairobi and Rwanda, as well as Addis Ababa, Makerere, Rhodes and Stellenbosch are main partners.

The African Institute for Mathematical Sciences in South Africa and the country’s National Institute for Theoretical and Computational Sciences are also involved.

In Europe, the universities of Ljubljana, Tübingen, Uppsala and Warwick are main partners, and two bodies in Italy are also involved – the Abdus Salam International Centre for Theoretical Physics and the World Academy of Sciences.

A map showing the participants in the Africa-Europe Cluster of Research Excellence in Artificial Intelligence. Image: Supplied.

‘New form of collaboration’

The cluster’s full name is Artificial Intelligence, Data Science and Theoretical and Computational Thinking, but its participants soon shortened it to CoRE-AI.

It is one of 20 clusters launched by the African Research Universities Alliance (ARUA) and The Guild of European Research-Intensive Universities in June and September 2023, framed by the AU-EU Innovation Agenda that was agreed to in February 2022.

Each of the clusters connects several African and European universities multilaterally to collaborate on an area of common interest, addressing major scientific and societal challenges through interdisciplinary research and higher education that leverage the skills and excellence of the participating institutions.

Every cluster falls under one of four thematic areas: public health, the green transition, innovation and technology, and capacities for science. CoRE-AI is part of innovation and technology, but also collaborates with clusters in other themes, particularly public health, and the green transition.

The CoRE initiative brings together partners from 120 institutions in 42 countries on the two continents. It is described as a “new form of collaboration between some of the best researchers from both continents, striving for equitable partnerships in an unequal world”.

One of the participating researchers, Dr Joyce Nakatumba-Nabende, a senior lecturer in computer science at Makerere University, says the notion of an equitable partnership has been put into practice in the CoRE-AI.

Dr Joyce Nakatumba-Nabende, head of the Makerere AI Lab in Uganda. Photo: Supplied.

“We have a co-leader from each of our two continents, we make joint decisions in our weekly meetings, and we take turns to present masterclasses. So, it’s not one grouping being prioritised over the other,” she told University World News.

African languages

She heads up the Makerere AI Lab in Uganda, where the human connection lies at the heart of her own research into natural language processing and using machine learning to build translation and automatic speech recognition models for African languages.

“If you want a computer to understand the way human beings speak, you need a lot of data. We have 52 languages in Uganda, but there’s a shortage of data in those languages on which to train computer models,” she explained.

“So, we started analysing local radio broadcasts, which are plentiful, with speech recognition systems. And we collaborated with Mozilla’s Common Voice platform to collect more speech data.

“Now we are applying the models we developed for particular uses, like agriculture. For example, we write programs so that computers can listen out for keywords in talk shows to create alerts about possible outbreaks of crop diseases or pests in Uganda.

“And in agricultural extension work, there’s question and answering systems that can provide agricultural advice to farmers, but mostly not in local Ugandan languages. Translation models can help, but there are biases that have to be considered.

“For instance, Luganda does not have gendered pronouns, unlike English. And because of stereotypes in society, standard machine translation systems will assume a ‘nurse’ is female and a ‘doctor’ male. That must be corrected,” she said.

Sumpter says this shows that expertise being developed in Africa can be put to good use elsewhere. “Joyce’s group has been very successful in training language models on small data sets, which is something that can be applied in various kinds of situations worldwide.”

Petruccione says the work of the CoRE-AI cluster can prevent Africa from missing another technological wave.

“Countries like Japan and South Korea made the most of opportunities like the semiconductor or silicon chip revolution when that came along. Africa missed out completely. Now we’re in the AI revolution, and we can’t afford to miss out again.”

PhD pipeline

A major focus of the CoRE-AI cluster is postgraduate training, especially at doctoral level. The cluster wants to build capacity around core competencies in AI, data science and theoretical and computational thinking with a view to addressing key developmental challenges.

This forms part of ARUA’s plans to “revolutionise doctoral training in Africa”.

ARUA’s intention is to increase the production of doctoral graduates at its 15 member universities five-fold – from the current annual average of 200 each to 1,000 each per year.

Last month, the body’s secretary general, Professor Ernest Aryeetey, said the Mastercard Foundation had expressed interest in funding a 10-year pilot programme in this regard.

Aryeetey said ARUA’s 15 member universities expect to admit the first cohort of about 1,000 students each in January 2026.

CoRE-AI has its eye firmly set on this initiative. Both Petruccione and Sumpter said the cluster is planning to admit 100 to 200 PhD candidates a year. It is putting the necessary systems in place and preparing to submit the required funding applications.

For its doctoral training initiative, CoRE-AI aims to use a ‘sandwich programme’ model to enhance student mobility across participating institutions.

‘Slice 1’ would be coursework at a student’s home university but ‘fortified’ with online interaction, the ‘filling’ would be a research project at a partner university, while the student would return to their home institution for ‘slice 2’, their thesis.

This would be in line with the plans outlined by Aryeetey for enhancing not only the quantity but also the quality of PhDs. Co-supervision by strong interdisciplinary teams across several participating universities, both African and European, would be a key feature of the programme, and the possibility of joint or dual degrees is being explored.

Entrepreneurship

CoRE-AI wants to produce PhD graduates who can innovate and not only go into academia but also start their own companies because technology and start-ups are often intertwined.

“We will encourage our students to start their own companies on the basis of their research, which will be addressing societal challenges, so they will be contributing to society,” Sumpter said.

“Our cluster is also addressing job creation, skills training and creating new career paths aligned with the fourth and fifth industrial revolutions,” Petruccione said.

Funding

The question of money inevitably comes up in research programmes, and the Africa-Europe CoRE initiative is no different.

Professor Peter Maassen of the University of Oslo, who is also an Envoy of The Guild to ARUA, told University World News that “most CoREs have been successful in applying for funding in one or more of the regular European Union programmes, especially Erasmus+ and Horizon Europe”.

The clusters were not provided with funding from the start, apart from some seed funding by the home institutions of participants. Uppsala University, for instance, funded the CoRE-AI’s kick-off workshop in Sweden last year, making it possible for representatives of all participating institutions to attend.

Each CoRE has to go out and find the necessary funding, depending on their research programmes.

Asked whether this is not a heavy burden, considering that all cluster participants have their own responsibilities at their home institutions, Petruccione said: “We find ways to stretch the day because we must. If we don’t pull together, we won’t lift Africa, or Europe, for that matter.”