From March 11 to 14, Khipu 2025 will take place in Santiago, Chile. Vía Libre will participate in various sessions and panels on AI and machine learning.
Khipu is one of the most important events in the region, bringing together researchers, students, and experts to discuss the present and future of AI in Latin America.
To learn more about the event, visit the following link.
📌 Some panels will be streamed live and can be watched online from the link below.
For those in Santiago, Chile, it is still possible to join in-person events on Thursday, March 13, and Friday, March 14.
Luciana Benotti is also part of the organizing committee of Khipu 2025.
Panels and workshops we will participate in:
- Empowering teachers with AI for participatory bias assessment of language technologies: Scaling teachers’ engagement to Latin America
Monday, March 10 – 14:00 PM to 16:00 PMPresentado por: Pietro PalombiniAssessing AI technology for biased and toxic behaviors is crucial for developing human-centered applications. However, current approaches often overlook culturally situated evaluations and lack representation from marginalized regions like Latin America. This project aims to empower teachers and students in bias assessment of language technologies, fostering more responsible and culturally aware AI. Currently, a course in Argentina involves 30 university professors, 400 high school teachers, and 6,000 students, who generate and analyze data to identify stereotypes and biases. The initiative will expand to Brazil, Mexico, Chile, and Uruguay, covering 60% of the Latin American population. Methodologies will be adapted to each context, and datasets will be released for future research on society-centered AI.
- Sociocultural evaluation of AI
Tuesday, March 11 – 5:45 PM to 7:45 PM
Participants: Luciana Benotti, Marcos Gómez, Pietro Palomini, and Vinod (India).
This workshop is a joint effort with Guido Ivetta, Pietro Palombini, Sofía Martinelli, Marcos Javier Gómez, Vinodkumar Prabhakaran, and Sunipa Dev. It is called “Socio-Cultural Evaluation of Language Models” and builds on collaborative work with Laura Alonso Alemany, Beatriz Busaniche, Nair Carolina Mazzeo, Emilia Echeveste, and other colleagues from Fundación Vía Libre and the National University of Córdoba. - Challenges and New Questions for Security Engineering in ML
Tuesday, March 11 – 12:00 PM to 2:00 PM
Presenter: Enrique ChaparroSecure ML systems are a common goal, but security requires engineering. Beyond general computing risks, ML introduces new challenges. Current approaches resemble a “patch and penetrate” cycle, akin to late-20th-century software security.
Early ML security and privacy research has focused on “operations security,” aiming to protect and maintain existing ML systems. This ongoing work seeks to establish security engineering best practices specific to ML.
We argue that Architectural Risk Analysis (ARA) is essential at this stage. ARA provides a design-level perspective, exposing systemic risks to ensure proper mitigation and management throughout system development. Unlike traditional security testing, ARA prioritizes design trade-offs and robust engineering over fixing isolated code errors.
Additionally, we explore security guidelines such as the Saltzer-Schroeder principles and the IEEE’s top software security design flaws. - Research in academia and industry
Thursday, March 13 – 6:00 PM to 6:40 PM (Chile
Participants: Laura Alonso Alemany (Vía Libre – UNC), Nando De Freitas (Microsoft), Jeff Dean (Google Research, Google DeepMind), Guillermo Sapiro (Apple, Princeton University), Nayat Sánchez-Pi (INRIA, Chile). - AI governance panel
Friday, March 14 – 3:40 PM to 4:20 PM (Chile)
Beatriz Busaniche will be the moderator of the panel, in which the following will participate: Ignacio Silva Santa Cruz (Ministry of Science, Technology, Knowledge, and Innovation), Marina Bericua (Microsoft), Luca Belli (Getulio Vargas Foundation, Brazil), and Juan Carlos Lara (Digital Rights).