In the outskirts of the Malagasy capital, a surprising reality unfolds: artificial intelligence (AI) is powered by real small hands. Andy, who works as a click worker, opens the doors to his daily life. Behind his computer, he compares images all day to train online shopping sites to sort products. It is thanks to him and other Malagasy workers that algorithms suggest relevant products to consumers.
This repetitive work allows Andy to earn only 6 cents of a dollar per task. To achieve a decent monthly income, he must complete more than 300 tasks. However, despite the precarious working conditions, Andy claims that AI offers a real employment opportunity for the Malagasy people.
The Outsourcing of AI in Madagascar
This use of Malagasy labor in the field of AI is not limited to independent workers like Andy. Companies dedicated to training algorithms have emerged in Madagascar. Beepeeoo, for instance, employs 100 full-time people to train AI for French clients in various sectors.
French clients are attracted by the wage level and the quality of Malagasy workers. The latter are highly involved in their work and have a perfect command of the French language. Malagasy operators can thus earn around 100 euros per month, which is three times the average salary in Madagascar.
Varied Tasks for AI Training
The tasks performed by Malagasy workers for AI training are diverse. At Infoscribe, for example, they work on training AI to recognize car license plates or to help future self-driving cars detect obstacles in road images.
But the most surprising task is the annotation of pedestrians on street surveillance cameras. This allows for counting the number of people passing a specific point and analyzing individual behavior to improve urban planning or offer targeted advertising in shopping malls.
The French State’s Involvement in AI Outsourcing
It is interesting to note that the French State itself relies on Malagasy labor for AI training. For instance, to identify taxpayers who do not declare their pool or for analyzing satellite images for the Ministry of Armed Forces.
However, the question of the precariousness of the workers at the end of the chain arises. These workers are the invisible cogs of our digital lives. They perform time-consuming and undervalued tasks, often outsourced by tech companies to precarious workers.
The Ethical Issues of AI Training
It is legitimate to wonder whether the rapid learning of AI could render the work of Malagasy operators obsolete. Indeed, some tasks require only two months of human annotation for AI to perform them alone.
However, some industry players, like Rova Rabetoviana from Beepeeoo, assert that there will always be a creative aspect and a human element necessary in the use of AI.
Researchers also emphasize the importance of the legal framework in AI development. Currently, companies that develop these software for supermarkets find themselves in a legal vacuum. The CNIL conducts checks in stores equipped with smart cameras to assess issues related to data protection.
In conclusion, artificial intelligence is indeed powered by real small hands in Madagascar. Malagasy workers play an essential role in training algorithms and thus contribute to the development of the digital economy. However, the precariousness of their situation and the ethical issues raised by this outsourcing deserve careful consideration.
