Most French AI start-ups have so far invested little in the political world, but already offer services to local authorities.
590 French start-ups specializing in artificial intelligence (AI), including 76 that deal with generative AI – which creates new content – and not a single one directly linked to politics. In its latest census, the France Digital association, a lobby for digital companies, has nevertheless identified around twenty different sectors of activity. From there to deduce that the market for AI in politics does not exist? “No, there is a market, but companies do not advertise politicians as first customers. That’s not how they define themselves.” observe Marianne Tordeux and Thomas Barreau, at France Numérique.
“Generative AI will shake up the world of communication and political marketing by shortening the time to produce content and targeting finely”, considers Othmane Zrikem, chief data officer and founder of an artificial intelligence company. The change could be rapid in the United States, where spending on political advertising reached in 2022, the year of the mid-term elections, 9.7 billion dollars (8.86 billion euros), according to estimates by the AdImpact company. The first official campaign video produced by an AI was also posted online on April 25 by the Republican Party to respond to the announcement of Joe Biden’s candidacy. It describes a scenario of anticipation considered as catastrophe, which begins with the (false) images of the re-election of the Democratic president, and continues with an equally fabricated sequence supposed to illustrate what the consequences of his re-election, of a crisis financial contribution to China’s invasion of Taiwan.
The road is still long
In France, the great upheaval that AI could bring to political practice has not yet taken place. “Everyone is reeling from generative AI, which is a game-changing technology update. Data analysis software tries to say that they use AI, it’s true, but in reality few successful things are marketed. We will undoubtedly see actors appear by the end of the year,” believes Edouard Fillias, head of the Jin consulting agency.
Stéphane Boisson also considers that the road is still long. In Montpellier, he runs Poligma, a company of eight employees that aggregates, analyzes and visualizes data, mainly public, for campaign candidates or communities. Cross-checking of data which can, for example, allow an elected official to anticipate the needs for places in a nursery in a district or for parking spaces to be reserved for the disabled. “AI will be able to do very interesting things, its algorithms will be able to imagine scenarios, it will be able to become a decision-making aid, without replacing elected officials. But for the moment we are still far from the mark, he judges. He is convinced that “the more relevantly the databases are aggregated, the better the results will be obtained by running the learning models”.
He is not the only one to have a company that uses public data, the quantities of which are increasing with the development of open data. After having worked on approximately 1,500 electoral campaigns, including that of François Hollande in 2012 and that of Emmanuel Macron in 2017, Arthur Muller, Guillaume Liegey and Vincent Pons stopped electoral targeting with their company LMP to devote themselves to Explain, a company which markets an “AI assistant” for companies that respond to public orders.
“A propagandist’s dream”
Among the thirty customers they claim: Engie, Veolia or Bouygues Telecom. “This allows them to understand the needs, by analyzing thousands of pages of administrative documents, deliberations, municipal reports, which contain a lot of informationexplains Guillaume Liegey. It is at the intersection of two trends: the explosion of public data that a human cannot process by hand, and the arrival at maturity of LMMs, the large language models, which allow reading and automatic writing.They completed in April a funding round of 6 million euros (to which Minister Jean-Noël Barrot, shareholder of LMP since 2020, as detailed by Releasedid not participate this time, indicate his firm and the company).
In the realm of politics, it may not be for its commercial prospects that generative AI will attract the greatest interest. “It will enable much more sophisticated and difficult to detect manipulation systems, on a much larger scale than todaypredicts Othmane Zrikem. More than private companies, it will undoubtedly be state authorities, such as China, who will be the first to use them for these purposes. It’s every propagandist’s dream.”