InstaDeep: ‘Africa must join the tech revolution’

In the mid-2010s, Beguir, born in Échirolles in the suburbs of Grenoble (France) in 1976, found himself in Tataouine, the Tunisian hometown where his father – a doctor – grew up.

Upon founding the Academy, an association dedicated to football and computer coding for young people, Beguir wished to create a website to publicise his initiative, so a mutual friend introduced him to Slim, a freelance developer and daughter of a diplomat, who found herself more interested in launching a website design service start-up for local small businesses instead of continuing her studies in English at the Sorbonne in Paris. So, the two had a conversation.

This is where InstaDeep got its start. Founded in 2014 by the entrepreneurial duo, it has since become one of the world references in artificial intelligence and machine learning alongside companies such as DeepMind or OpenAI, the designer of the now-controversial ChatGPT.

After raising $100m in January 2022, InstaDeep has just been acquired by BioNTech, the German biopharmaceutical leader, for a record amount of €408m.

At the head of a 250-person team spread over three continents, the co-founders will be able to pursue the projects undertaken since 2019 with the company created by Uğur Şahin, while continuing to shine in global research – around 30% of the annual budget. InstaDeep is devoted to this activity.

The now ex-start-up will also continue its work in sectors as diverse as transport, logistics, high technology and energy.

Passing through Paris, Beguir and Slim agreed to answer questions from us. The day after this interview, they flew to Tunis, where they were expected to receive the honours of Prime Minister Najla Bouden.

Taking advantage of their pioneering image and their new influence, the entrepreneurs mentioned avenues for regulatory improvement for a revision of the Tunisian Start-up Act. Among them, is a liberalisation of the exchange code.

InstaDeep has just been acquired by BioNTech, during a financially delicate time for companies at such a level. Did this trend influence your decision?

K.B.: We have had no issues with funding since we were profitable three years after our creation and we raised $100m a year ago. The company is growing rapidly. We are at just over 250 employees now, half of whom are based in Africa.

The idea was rather to join the forces of two companies that know each other very well and have been working together on dozens of projects since 2019.

What prevented you from pursuing your development on your own?

K.B.: Together, we can do things on a larger scale. InstaDeep will shine even further in sectors beyond biopharmaceuticals. It’s a bit of an Afro-European DeepMind [Editor’s Note: from the name of the company specialising in AI belonging to Google].

We have a lot in common with BioNTech, including this passion for innovation and a positive contribution to humanity. Uğur Şahin is himself a researcher passionate about science, who, like us, publishes regularly.

The idea was to bring in partners who could open doors for us.

Microsoft recently offered up product exclusivity developed by OpenAI. DeepMind was acquired in 2014 by Google. Logic would have dictated that you approach a large American technology group. Why did you take a different path from your competitors?

K.B.: BioNTech is a leading innovation company in its sector. She is behind the greatest success in the history of biopharma thanks to the first messenger RNA vaccine, developed to fight Covid-19 and which saved millions of lives.

For our part, the field of health fascinates us, and we will be able to work on personalised vaccines against cancers. There are real opportunities to bring something positive to people. BioNTech can also help us build very powerful computing systems.

Would you now consider buying African companies?

K.B.: We are very willing to collaborate in every sense of the word with other African companies and to be a force for the development of the African tech ecosystem. But, for the moment, we have no specific project on this subject.

Not raising funds right away instils patience.

You both own just under 50% of InstaDeep, after controlling over 60%. How do you view the dilution of a business? Should we learn to lose control?

K.B.: The main thing is to find the right partner. Our decision to open capital was never motivated by the simple fact of raising money, since we were already profitable [Editor’s Note: in 2021, according to figures that we were able to obtain, InstaDeep achieved a figure of business of nearly $13.9m and this, with a team of 100 people].

The idea was rather to bring in partners who could open doors for us, such as AfricInvest, which for example helped us set up offices in Lagos and Cape Town. In general, you must first develop the company, strengthen the teams, and experiment with several things before raising money. Because mistakes made without money cost less.

Z.S.: Not raising funds right away instils patience. I see a lot of business founders talking about raising money after only six months. However, I think that focusing on the fundamentals of a company before raising funds gives you more time to choose with whom you want to move forward.

What projects are you currently working on?

K.B.: We like to look at problems where AI has never been applied. For example, we are working on PCB design automation with an industry leader. The team also has strong Tunisian and African leadership which ensures that AI accelerates the design of printed circuits.

Z.S.: The pandemic and the ensuing difficulties in the supply of components such as cobalt have shown the importance of this type of technology.

We hope that in 2023 it will be possible to raise local funding while having local mentors.

Do you have projects in Africa?

K.B.: The projects we take on are not all for commercial purposes. With Google, for example, we have developed, within the framework of research, the first locust cloud detection system. In 2020, this recurring scourge has taken the form of a cloud measuring the size of New York City. In one week, it devoured the equivalent of a stock of food useful to 10 million people.

Our system makes it possible to identify, with a 90-day notice, the areas where these clouds are likely to form. This allows the authorities, in particular the Food and Agriculture Organisation of the United Nations (FAO), to intervene more effectively.

A 100% African team, made up of our researchers in Tunis, Cape Town and Lagos, worked with the Google research centre in Accra, which provided the data.

How do you concretely proceed to predict the advent of such a disaster?

Z.S.: We aggregate all the historical data available (weather, humidity, type of soil) to create a predictive system. We also map the wells and all the wetlands, which allows us to determine where the larval gestation areas are.

You have chosen to open an office in London, in particular in order to be able to attract non-African funds. From now on, you are controlled by a German company. Before you, the Nigeria-based Paystack was bought by the American-based Stripe, when the e-commerce platform Jumia chose a Wall Street listing. Can the financial success of an African company be recognised abroad?

K.B.: It is true that it was very difficult for us to obtain local funding, but that has changed a lot. We hope that in 2023 it will be possible for a start-up to raise local funding while also having local mentors.

But the bottom line is that the young Africans who develop the technologies of tomorrow have a positive impact on their community, and even on a larger scale. The current challenge is to create the conditions that allow them to develop this new economy.

Too often, decision makers see start-ups as side companies tinkering with things no one understands.

Do you invest personally in young African initiatives?

K.B.: Yes, we are going to do it more and more. We have also invested part of our budget in the organisation of the AI summer in Tunis. It took the form of two events made up of the Deep Learning Indaba, which in 2022 brought together 400 AI researchers from across the continent within the premises of the Sup’Com engineering school in Tunis, and the AI Hack, an artificial intelligence hack-a-thon which took place over three days in the multipurpose Olympic Hall in Radès, with 1,000 participants.

We did the same in 2019, spending 20% of our annual budget on the organisation of the first edition of the AI Hack, which has since benefited from the support of the Tunisian government and Google.

Z.S.: Investing in the ecosystem in this way can influence decision-makers who, for some, too often see start-ups as side companies that tinker with things that no one understands. The click has to come and these people discover that the value really exists.

Can your models encounter interpretation biases as we see for certain conversational AIs?

K.B.: Absolutely. We know that AI works, but it can be biased by the data it receives. This is why it is extremely important that Africans or developing countries, in general, take ownership of these technologies.

There will always be biases but, if we want to minimise them, we must ensure these extremely powerful technologies are invested in and developed by people of diverse origins and representative of the world’s population.

In the design of its conversational AI called ChatGPT, OpenAI, one of your competitors, is accused by the American magazine Time and several NGOs of having exploited employees of Kenya-based Sama company to “clean up” the system by processing information deemed “hateful” or “toxic” on a daily basis. Are you just as obligated to resort to such moderation?

K.B.: AI is imperfect because it is created by humans who are not perfect, so obviously we have to go through that. But, at InstaDeep, our goal is to give young African talents, in partnership with their international colleagues, the opportunity to work on real added value.

What OpenAI does in San Francisco, we do in Tunis, Lagos, and Cape Town. The idea is therefore that Africans have access to high-value-added positions rather than being relegated to the role of underpaid subcontractors.

In Africa, the failure of public policies is regularly attributed to the absence of an identification or demographic database. How do we address and fix this problem?

Z.S.: It is true that a few years ago the number of databases available was much smaller. However, something clicked recently during Covid-19. For example, we have seen the creation of public databases on vaccination rates. Governments, regions, and businesses have started to take data seriously and realise the impact it can have on people’s lives.

K.B.: In reality, neither data nor funding is an issue today.

My goal is not to predict the future but to correct injustice.

In terms of your commercial development, which countries or sectors are you targeting as a priority on the continent?

K.B.: The emergence of new world-class start-ups, founded by Africans in Africa, represents an interesting economic breeding ground for us. It is indeed easier to work with companies like Yassir or GoMyCode, which know how to manage data and which are growing exponentially. They will be the first to adopt AI.

Research on artificial intelligence is sometimes perceived as a quest to reduce risk, or even as a means of anticipating the future. Is this your obsession?

Z.S.: I wouldn’t call it my obsession, but I see how the world can be unfair from time to time. I think I am a very honest person and, in this logic, my objective is not to predict the future but rather to correct some injustices.

K.B.: AI is, in any case, the greatest technology of our time. And, like any technology, it is neutral and depends on the use made of it by the people who handle it. With Zohra, we are fighting for a world where AI benefits medicine, development and the economic well-being of all. The idea is to move towards a more harmonious and balanced world, in which, thanks to AI, African states sit at the table of the nations that matter.