Pair of Dima Droubi
The mirror is chipped, but the toilets are clean and the LED bulbs, which comply with ecological standards, give off a sad white light. Houcine, 29, is struck by the look he sees in his reflection: frustration and disenchantment. Graduated in computer scienceHoucine dreamed of designing innovative technological solutions for companies in his country. Five years later, he crosses the busy streets of Casablanca to join an office building where he works as a data labeler. All day long, in front of his screen, he classifies images and texts in order to train artificial intelligence (AI) algorithms. This repetitive and poorly paid work, which perpetuates the trend of subcontracting low value-added tasks in developing countries with a qualified but underpaid workforce, is far from the dreams that nourished his youth.
Houcine is just a drop in an ocean of broken dreams. His case reflects a systemic problem. Although AI has immense potential to transformglobal economyit risks further widening inequalities between developed and developing countries. AI, which certainly boosts productivity and theinnovationleaves behind millions of people, and as many dashed dreams, like that of Houcine.
According to a report from PwC in 2017, artificial intelligence could add $15.7 trillion to the global economy by 2030. However, 84% of this wealth would be concentrated in Chinein North America and in Europe. Regions likeAfrical’Oceania and parts of theAsia will receive only a tiny portion of these profits. The industries most affected by this inequality are those where machines, which have become cheaper, replace human workers. Low-skilled jobs, often outsourced to developing countries, are now automated in rich countries. Workers from the countries of the South, once drivers of the world economy thanks to workforce cheap, see their jobs disappearing at an alarming rate. In short, the jobs that were “offshored» are going to be «onshored back» in rich countries.
Several reasons explain this growing gap. One of the main barriers to AI adoption in developing countries is the lack of infrastructure. In Africa, only 25% of the population has access to Internetand the tools needed to fully exploit AI technologies are rare and expensive. AI models, often designed in Western contexts, are not adapted to local realities. For example, they mainly use theEnglish. Data in English is better suited to rich country contexts.
Local companies struggle to compete with technology giants who monopolize computing and data resources. These monopolies limit the opportunities for local innovators like Hussain to develop solutions adapted to their environment. Just to correct the biases generated by data predominantly in English, these local innovators must devote considerable time and effort to produce applications adequate for their country’s context. Despite these challenges, initiatives are starting to emerge. For example, AI training programs are multiplying in countries like Moroccoand some entrepreneurs premises adapt technologies existing to local needs. For the moment, these efforts remain marginal given the scale of inequalities.
A report of McKinsey evaluated the industries who would benefit most from advances in AI. Unsurprisingly, the sectors of high-techbanks and retail are winning. L’agriculturean economic pillar in Africa, remains lagging behind. However, we should not lose hope: advances in AI are growing at an exponential rate. All it takes is for a young man or woman, armed with a solid and adequate education, to identify a need for him or her to begin looking for solutions with the help of AI. In this way, humans and machines move forward together.
For AI to become a lever of opportunities rather than an amplifier of inequalities, it is crucial that governments invest in digital infrastructure andeducation. International partnerships could also help democratize access to technologies and reduce entry costs, particularly with regard to computing resources. The Global South represents only 1% of the world’s supercomputers. It is equally imperative that AI is trained taking into account the lifestyles, needs and realities of developing countries in order to avoid harmful biases. It is crucial to act so that Houcine’s frustrated look gives way to a look of hope. Houcine, his wife and their children must be able to continue to dream of a future where artificial intelligence will not be a tool of exclusion, but an opportunity for all.
