‘Trompe-l'œil’ automation

Artificial intelligence (AI) is often seen as to signing the end of certain jobs and to the development of highly qualified positions. Professor Antonio Casilli (sociology, Telecom ParisTech) addressing the EUROPA GE90 seminar, presented his investigations, a nuanced vision of the work scene that underpins digital automation.

‘Trompe-l'œil’ automation

“In fact, artificial intelligence, does not eliminate work but rather makes it invisible, with a subdivision into micro-tasks”, sums up Prof. Casilli. The thesis defended by this specialist of digital anthropology contradicts a number of common presuppositions. As it sees it, this idea of AI replacing manpower bears, above all other considerations, the mark of marketing and ideology forces for the purpose of masking free work carried out by consumers themselves and precarious workers thousands of miles away. Comparing corporate communication about the imminent arrival of fully automated services with the reality of technical and human backstage activities, his analysis reveals a major gap.

The real operational functioning of applications still depends, to a large extent, on human interventions, but remains hidden by the prevalent discourse on self-reliant machines blatantly served to the public at large. “Accompanying today‘s scientific and technological challenges, we are witnessing the construction of a specific discourse designed to defend both partisan and economic interests”, he underscores.

 

The people behind AI

The students were able to discover - through the presentation of concrete cases such as the Uber platform, the self-drive cars and research engines – a totally new vision of something they use on a day-to-day basis. The emblematic example is the car booking service of Uber was analysed from various points of view. Normally this servicer is perceived as a person-to-person portal between potential passengers and the Uber drives, the application also collects user data without any service in return. “Uber drivers spend more time on their smartphones forwarding information to Uber than they do behind the driving wheel, inasmuch as the trips help build up maps and trip frequencies that are then used to improve the overall efficiency of the Uber system”, explains our research scientist, illustrating his points on the screen. The project of a self-drive taxi, forecast as the next stage of an AI world in progress, was thoroughly debunked. While the self-drive taxi is supposed to operate without any human intervention, in fact the vehicle requires remote operators and even help from the passenger(s) to warn of presence of obstacles ‘en route’.


The fundamental theme of machine learning was addressed via an explanation of certain operational modes of research engines. This special capacity – to recognize and reference unending new data inputs - presented as a result of self-learning machine modes, in fact calls on human decisions and interventions. Only the case of seeing thousands of data at the input could enable machines to be able to handle certain complex situations. “Millions of invisible workers who live outside our Western economies, in the Philippines, in Indonesia, in India and Africa, work on behalf of the GAFA (Google, Apple, Facebook and Amazon)”, the sociologist explains, who based his findings on enquiries carried out in several of the countries mentioned above. There are, spread round the world, tens of millions of micro-task-workers whose jobs it is to name photographs, to rank videograms and music in categories, to associate words or expressions with web-sites, to translate these words and expressions into their native languages.

They are recruited via on-line platforms such as Amazon’s Mechanical Turk [cf. https://www.mturk.com/ ] at very low pay rates, with these myriad invisible workers omnipresent behind the service offers appearing on the Internet. Amazon is not alone here. In many circumstances, the web surfers are invited to make suggestions to improve a translation or a selection of holiday and leisure sites. “The promises that technological progress implies that machine-based intelligence could become totally independent have been recurrent over the past 70 years, but personally I do not entertain much belief in these prophesies”, concludes Professor Antonio Casilli, with a touch of pessimism notwithstanding.