Page 7 - EngineerIT February 2022 UPDATED

ARTIFICIAL INTELLIGENCE

The AI system identified more than “This is a powerful and potentially game-changing technology given the number of
90 percent of variants of concern, on important and highly impactful problems that it could help us solve,” said Arnu Pretorius,
average two months before their the AI Research Scientist who leads InstaDeep’s South Africa team. Pretorius
designation by the World Health himself is a prolific researcher, whose numerous presentations at
Organisation. It detected the world-leading academic AI conferences puts him in the ranks of
highly transmissible Omicron Africa’s top AI researchers.
on the day its sequence South African “Some of our work in this direction looked at managing
became available among scarce resources using MARL to potentially overcome issues
more than 70,000 novel AI expertise to such as the economic problem of balancing community and
variants discovered in self-interest when managing common-pool resources such
October and November address African as water, arable land or the atmosphere. This work was
2021. problems presented at the most prestigious AI conference in the world,
“More than 10,000 the conference on Neural Information Processing Systems, or
novel variant sequences NeurIPS – only the second time that research from Africa was
are discovered every week and presented there,” Pretorius said.
human experts simply cannot cope InstaDeep’s pan-African team also collaborated with Google to
with complex data at this scale,” Beguir predict desert locust breeding grounds in advance to alert local farmers across
said. “For the first time, high-risk variants Africa of potential threats from devastating locust outbreaks, which was also showcased at
could be detected on the spot, potentially NeurIPS in December
saving months of precious time.”
Empowering South Africa’s AI community
South African AI expertise to address “We have research engineers not only focusing on research but also working on applied
African problems projects in collaboration with other offices, clients and the research teams. These projects
Using AI to tackle the world’s toughest range from biotechnology to hardware design using AI,” Pretorius added.
problems is in InstaDeep’s DNA and the “In South Africa’s AI landscape, InstaDeep occupies a very unique space, blending
company’s South African team is busy engineering and research, while having a positive impact on the world,” he said. “We also
working on new ways to advance AI so it see ourselves as playing an important role in the larger AI community when it comes to
can help solve challenges in areas such strengthening connections, upskilling and providing opportunities for learning. Examples
as health, transportation and resource of these efforts include our close collaboration with the Deep Learning Indaba, an
management. organisation dedicated to building Africa’s AI community, where we sit on the steering
A particular interest for the Cape committee as well as participating as sponsors, volunteers and mentors. We have a
Town-based InstaDeep team is “Multi- passion for being of service.”
Agent Reinforcement Learning,” or While the South African operation of InstaDeep is headquartered in Cape Town some of
MARL, an approach for solving large- the researchers are based in Johannesburg and Durban. n
scale problems by getting many AI
devices to work together, for example
eliminating traffic congestion with self- Deep Learning Indaba
driving cars. InstaDeep’s focus on the
subject has already had an impact on he Deep Learning Indaba is an organisation with a mission to strengthen
South Africa’s AI ecosystem with Cape machine learning and artificial intelligence in Africa. Their goal is for Africans
Town fast becoming an academic centre Tto be not only observers and receivers of the ongoing advances in AI, but
for MARL research. active shapers and owners of these technological advances.
Developing and working with Much of the wider discourse at present is permeated with conversations around
local AI communities is also central to the 4th industrial revolution, the need for policies and interventions around changes to
InstaDeep’s ethos. To empower other AI jobs and workplaces, the impacts of increasing automation in societies, of high-levels
researchers, InstaDeep’s South African of global investment in AI and machine learning, and visions of AI-first organisations.
team also created an open-source MARL What underlies this conversation is the ongoing and rapid advances being made in
framework, called Mava – named after artificial intelligence. It is essential for Africans to become not just observers of the
the Xhosa word for ‘experience’ – with ongoing advances in AI, but active shapers and owners of these technological advances.
the aim of accelerating innovative It is for this reason that the Indaba was conceived. And for this reason, it plays a unique
solutions to problems across the and important role within our continent. n
continent and beyond.

EngineerIT | February 2022 | 5