10-01-2024
The American elections will be the crash test for artificial intelligence
- Politique
American elections have traditionally served as a laboratory for emerging digital practices. Every month, leading up to the November election, we delve into the digital campaign with Ronan Le Goff, co-director of La Netscouade.
Context
- A decisive election in the era of generative AI. In 2020, the U.S. presidential elections were already haunted by the specter of deepfakes, but four years later, the landscape has changed: technological means have multiplied and become widely democratized. On the technological front, the latest version of Midjourney achieves nearly photographic quality, further blurring the lines. On the usage side, the recent Argentine presidential election was marked by torrents of AI-generated images.
- The cost of disinformation production has dropped significantly. A staggering story emerged after the 2016 U.S. elections: disinformation was not only Russian; Macedonian teenagers had churned out fake news by the kilometer solely to generate clicks with pro-Trump articles. Handcrafted articles in approximate English. If such a story were to occur in 2024, ChatGPT would undoubtedly be enlisted for a more effective and faster outcome. The production of texts and images on a large scale has literally become child's play.
- A 2016 boosted with AI? Fears are high of experiencing a scenario similar to 2016, where fake news and Cambridge Analytica had muddled the democratic game. Under the influence of AI, "the 2024 elections are going to be a mess," warns Eric Schmidt, former Google CEO, who advocates for "freedom of expression for humans (but) not for computers." According to Democratic Senator Mark Warner, Russian manipulations in 2016 were "child's play compared to what national or foreign AI tools could do to completely ruin our elections." Microsoft has already identified the use of AI-generated images on Chinese propaganda accounts targeting the American public.
- Faulty moderation. While social media platforms have established dedicated units for election integrity, their effectiveness is questionable. Meta laid off over 20,000 people in 2023, including many employees responsible for fake news and election integrity. X (formerly Twitter) remains pivotal in information circulation, and Elon Musk shows pronounced disdain for these issues. Solutions may come from AI companies: Google has announced plans to restrict its Bard tool regarding the 2024 elections, while OpenAI is considering ways to strengthen its moderation for this deadline.
- The United States, but not only. The issue of AI in politics will be even more significant in 2024, as it is a historic election year. No fewer than two billion people are called to vote. In the United States, India, Russia, the United Kingdom, Mexico, Taiwan, South Africa, Indonesia, and also on the old continent for the European Parliament elections.
What to expect for 2024? Slovak and Argentine examples
- Slovakia: The audio deepfake of defeat. This is the most significant case to date of AI-assisted manipulation. 48 hours before the legislative elections in September 2023, an audio clip of progressive party leader Michal Šimečka is posted on Facebook. He is heard discussing election rigging with a journalist. Šimečka immediately denies it; AFP concludes it's a deepfake, but the damage is done: the bomb drops within the 48-hour window during which media and politicians are silenced. In a very close race, Šimečka loses the elections, and Russia gains a new ally within the EU.
- Argentina: Generative AI becomes commonplace. A few deepfakes (often not very credible) circulated during the November 2023 elections, but Argentina stands out for its totally unabashed use of generative AI. Javier Milei, the eccentric far-right candidate who was elected president, didn't hesitate to post an image of his rival Sergio Massa dressed as Mao. His opponent reciprocates, with his campaign team forming an AI-dedicated collective - "Sergio Massa in AI for the nation" - which released a highly successful viral video, depicting Javier Milei in satanic rituals. A sort of Hollywood-like production made at zero cost from Buenos Aires. It's not strictly disinformation (no one believes for a second that these videos are real) but rather a new use of technology for propaganda purposes. Practices that currently float in total ethical ambiguity.
Reasons for reassurance
- Disinformation is rampant, but fact-checkers are vigilant. It has been eight years since the term "fake news" entered common language. Journalism has had time to adapt, and fact-checking services are now seasoned and capable of responding to any false information within a few hours (as seen with AFP in the Slovak deepfake case). If there is an economy of misinformation, there is also an economy of fact-checking: the excellent Checknews site is one of the main subscription supports for the Liberation website.
- Promises of citizen fact-checking. A novelty in 2024: X's "Community Notes" tool could play a decisive role in debunking fake news at the source. This participatory fact-checking tool - one of the few positive contributions from Elon Musk - has already defused a malicious use of generative AI in the American campaign. In June, the team of Republican candidate Ron DeSantis had posted a video showing Donald Trump hugging Dr. Fauci, the American Covid expert disliked by the Republican base. An AI creation, immediately denounced as such by X's astute investigators, through a note published directly under the video. Simple, effective.
Biden releases a first campaign video (without AI!)
Who needs artificial intelligence when news images are enough? Joe Biden's first campaign clip, released on January 4, focuses on the threat to democracy posed by a potential reelection of Donald Trump. The menu includes real images of the Capitol assault on January 6, 2021.
Le condensé de culture numérique qui explore les nouvelles tendances du digital.
- Santé
- Culture digitale
- Politique
- RSE
- Éco-conception
- Média
- Éducation
Picks est désormais
sur WhatsApp
Pour nous rejoindre, rien de plus simple.
1. Enregistrer ce numéro +33615605232 sur WhatsApp sous le nom "Picks".
2. Cliquer sur le bouton "Nous rejoindre" ci-dessous et suivre les instructions.
En procédant à votre inscription, vous acceptez l’envoi ultérieur d’un condensé de l’actualité numérique uniquement sur WhatsApp (aucun sms ou appel). Vous pourrez sortir de la liste de diffusion à tout moment en cliquant sur "Bloquer le contact" dans les paramètres de la conversation. Votre numéro de téléphone ne sera pas récupéré ou utilisé pour toute autre utilisation que celle mentionnée ci-dessus.
Grand bol
d'inspiration
PICKS, made in La Netscouade, le condensé de culture numérique qui explore les nouvelles tendances du digital.
Votre adresse de messagerie est uniquement utilisée pour vous envoyer les newsletters Picks. Vous pouvez à tout moment utiliser le lien de désabonnement intégré dans la newsletter. En savoir plus sur la gestion de vos données et vos droits
Manifesto
Nous croyons, à La Netscouade, qu’il est possible de régénérer la communication digitale tout en restant fidèles à ce que nous sommes.
Plus qu’une marque, La Netscouade est une marque de fabrique : nos clients viennent chercher chez nous une manière de faire, une capacité à imaginer des contenus à valeur pédagogique et créative, une propension à mixer cultures et talents dans un esprit « laboratoire d’idées ». Sans refuser la vitesse du numérique, nous avons toujours privilégié le long terme pour créer du lien entre l’entreprise et ses publics. Ce sont la transparence, l’écoute active, la constance et la consistance des prises de parole qui génèrent de la confiance et in fine de l’engagement durable.
Dans la période actuelle, nous aidons nos clients à faire du sens un vecteur de performance. Les projets que nous menons avec eux, RSE, numérique, nouveaux media, e-santé, éducation, science, font converger intérêt particulier et intérêt général et créent les conditions d’un engagement plus fort et plus durable.