Tag - ai

AI & FAKE NEWS: PER CONTINUARE A LOTTARE BISOGNA “DISCENDERE TRA IL VERO E IL FALSO”
Quanto è importante prestare attenzione alle foto false generate dall’AI, l’intelligenza artificiale, che circolano sul web? Ne parliamo con Davide Del Monte, presidente di info.nodes. Radio Onda d’Urto ha intervistato Davide del Monte a seguito dell’incrementarsi nel Mondo di due diverse correnti di utilizzo dell’intelligenza artificiale e della diffusione delle ‘fake news‘. Un utilizzo è quello incontrollato della controparte, delle grandi potenze, per creare notizie false e criminalizzare lotte o movimenti e che Del Monte definisce pura “propaganda“. L’altro utilizzo dell’intelligenza artificiale proviene, invece, dal basso. E per analizzarlo si è partiti da una fake news diffusa durante la Global March to Gaza: una marcia composta da persone da tutto il mondo che stanno convergendo in Egitto per cercare di rompere l’assedio israeliano su Gaza. Le immagini che circolano, però, sono in alcuni casi false: la carovana Sumud partita dal Nordafrica è effettivamente partecipata da migliaia di persone, ma i video e le foto che si stanno diffondendo maggiormente sono realizzate dall’intelligenza artificiale (vedi foto). “C’è il rischio di depotenziare e delegittimare il messaggio di questa importante iniziativa se diffuse”, fa sapere Davide Del Monte. La disinformazione sui social ha anche contribuito grandemente alla tensione in California a seguito delle proteste anti-ICE. E’ tornata in auge la già utilizzata foto di mattoni impilati su dei pallet – e che circola sin dalle proteste di Black Lives Matter nel 2020 – per tentare di criminalizzare le proteste producendo notizie false e per tentare di dimostrare che ci sia una volontà “esterna” a guidare le rivolte a Los Angeles. Nel contesto del film Matrix, la “realtà” è definita come una simulazione creata da macchine, che viene percepita come “reale” dai personaggi umani. Siamo a questo punto? Ne parliamo con Davide Del Monte, presidente di info.nodes. Ascolta o scarica.
OpenAI, DeepSeek and the Rise of a ‘Digital Lumpenproletariat’: DiPLab’s Antonio Casilli Interviewed on Italian National Radio
Antonio Casilli, professor of Sociology at the Polytechnic Institute of Paris and co-director of the DiPLab research group, recently featured on Italy’s national radio program Eta Beta, hosted by Massimo Cerofolini on Rai Radio1, to discuss the dark side of artificial intelligence. In the interview, Casilli shed light on the invisible workforce behind AI’s apparent automation—millions of precarious, underpaid workers across Africa, Asia, and South America who perform the “dirty work” powering AI systems. For mere pennies per hour, these digital laborers filter disturbing content including graphic violence and sexual abuse, annotate images and videos, correct algorithm errors, train self-driving cars, and even produce commissioned social media engagement. The DiPLab director, author of the award-winning book Waiting for Robots. The Hired Hands of Automation, also highlighted how everyday users unknowingly contribute free labor to improve AI models like American-made ChatGPT or Chinese-developed DeepSeek, further entrenching this exploitative system.
Two new academic articles on AI published by DiPLab!
We share the exciting news of two new papers that were published last month, concerning parts of the extensive research DiPLab conducts on the networks of production of AI The first paper is titled “Where does AI come from? A global case study across Europe, Africa, and Latin America” (by P. Tubaro, A.A. Casilli, M. Cornet, C. le Ludec and J. Torres Cierpe), appears in New Political Economy’s special issue on power in the digital economy. It examines AI supply chains, focusing on how and where companies recruit workers for data annotation and other essential tasks. While the organisation of AI data work varies, the reasons for these differences and the ways it dovetails with local economies were underexplored. This article clarifies these supply chains’ structures, highlighting their impacts on labour conditions and remunerations. Framing AI as an instance of well-known outsourcing and offshoring trends, analysis of AI data work in France, Madagascar, and Venezuela, highlights two main models: marketplace-like contracts and firm-like structures, with hybrid arrangements in between. Each model suits different AI tasks but all reproduce well-known patterns of exclusion that harm externalised workers especially in the Global South. We argue that worker reclassification alone is insufficient and advocate for a broader policy mix, including regulation of technology and development strategies at national and supra-national levels. You may find an open access version of the preprint of the paper here -------------------------------------------------------------------------------- -------------------------------------------------------------------------------- The second paper is titled “The digital labour of artificial intelligence in Latin America: a comparison of Argentina, Brazil, and Venezuela” (by P. Tubaro, A.A. Casilli, M. Fernández Massi, J. Longo, J. Torres Cierpe and M. Viana Braz) appears in Globalizations’ special issue on AI in Latin America. It sheds light on the precarious, low-paid data workers supporting AI production in the region, often for foreign firms. Mixed-method data support a comparison of Argentina, Brazil, and Venezuela to reveal common patterns and regional differences. The analysis supports the conclusion that Latin America plays a key role in AI data work, with companies exploiting economic hardship to cut costs. In Venezuela and Argentina, crisis conditions foster an ‘elite’ of young, STEM-educated workers, while in Brazil, this work is done by lower-income groups. In all three countries, AI data work also blends with the informal economy, reinforcing inequality in this way. These findings call for more attention to AI labour conditions and advocate for policies to recognise data workers’ skills and support their career development, potentially enabling worker organisation. You may find an open access version of the preprint of the paper here