Albert Tort

⚫The first paper that uses Quantum Spain quantum computer is published! ⚛️Get to know more about Qdislib, a hybrid distributed quantum-HPC library to cut and execute quantum circuits. ➡ https://lnkd.in/d47NbrkQ Alba Cervera Lierta, PhD Berta Casas Font Mar Tejedor Ninou Javier Conejero Bañón Rosa M. Badia

451

2 comentarios

Today, Íñigo Martínez de Rituerto de Troya presented a core piece of his PhD work at ACM FAccT in Athens. In “Misabstraction in Sociotechnical Systems”, we provide a richer perspective on how abstractions are used in the design, development, analysis, use and governance of sociotechnical systems. While, in recent years, the societal implications of AI and algorithmic systems have led to an enriching of scholarly, advocacy and policy efforts to better understand and address these, we still lack shared understandings of how we are or should be abstracting AI systems and their effects, not just at the technical level, but also in how these systems are embedded in operational contexts where they impact lives and professions, how organizational actors and processes go about designing/maintaining/governing/overseeing/etc, and the institutional arrangements that apply to or otherwise affect these practices. As Íñigo concluded, we hope this paper helps you think about abstraction differently. Furthermore, there is a lot of follow-up work possible to further crystallize conceptual and methodological approaches to guide abstraction in the growing efforts to operationalize AI governance that allow for effective AI oversight but also better system design, in ways that represent and empower the needs and concerns of those bearing the brunt of consequential AI systems. Read the paper here: https://lnkd.in/eQHU4dDn With Jacqueline Kernahan, Neelke Doorn and Virginia Dignum.

89

3 comentarios

🗣️ Will AI shape the way we speak?  This is the question Jūra Miniota, Míša Hejná and I asked in our paper that was presented last week at the 15th International Workshop on Spoken Dialogue Systems Technology (IWSDS-2025) in Bilbao 🇪🇸 We are thankful to the IWSDS community for creating an inspiring environment where interdisciplinary ideas like this can be explored and discussed openly. In this position paper, we argue that the increasing realism and prevalence of synthetic voices and conversational AI could lead to significant sociolinguistic shifts. Passive media already impacts everyday language, but as conversational AI is interactive, this effect can be magnified through acoustic-prosodic entrainment. The effect of speakers adapting their speech style during conversations can persist beyond the interaction and become part of their habitual voice, and with it an expression of social identity. 🔄 We further speculate that conversational AI’s socioindexical influence might become a powerful channel for organizations, brands, and movements to shape public perception and identity. We call for interdisciplinary collaboration to further understand and ethically manage this emerging dynamic. We are eager to know what the broader scientific community thinks about this. Is engineered language change already happening? Or is this even something to be concerned about in the future? We made a poll in the poster session where people could vote with stickers ⭐ and we had many interesting conversations which we would like to continue. What do you think? 📄 Full paper is available in the ACL Anthology: https://lnkd.in/dpYfaDxm #ConversationalAI #Sociolinguistics  #HumanComputerInteraction  #EthicsInAI  #VoiceAI #AIandSociety

118

5 comentarios

*** 𝗧𝗲𝗰𝗵𝗗𝗶𝘀𝗽𝗮𝘁𝗰𝗵: 𝗙𝗲𝗱𝗲𝗿𝗮𝘁𝗲𝗱 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 (𝗙𝗟) *** Today, in cooperation with the Agencia Española de Protección de Datos - AEPD, #EDPS publishes a new edition of TechDispatch focusing on Federated Learning. Federated Learning is developing machine learning by allowing multiple data sources to collaboratively train a shared model while keeping data decentralised. This approach mitigates privacy risks as raw data remains local, which is particularly beneficial for scenarios with sensitive data or regulatory constraints. FL's applications are numerous, from personalised recommendations and healthcare data analysis to autonomous transport systems, where privacy and data protection are crucial. 📈 𝗞𝗲𝘆 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: ☑️ Minimises personal data sharing, aligning with data protection principles like data minimisation and purpose limitation. ☑️ Enhances accountability and auditability by keeping raw data on local devices and sharing only model updates. ☑️ Reduces the impact of large-scale data breaches. 📉 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀: ☑️ Potential data leakage through model updates leading to membership inference attacks. ☑️ Security must be robust across the entire ecosystem to prevent systemic compromise. ☑️ Ensuring data quality and eliminating bias in distributed training. To fully leverage the benefits of FL while addressing its challenges, a holistic approach to personal data processing is essential, focusing on privacy by design and by default, and ensuring balanced data access among federated parties. FL can develop AI systems that are both powerful and respectful of individuals' rights and freedoms. ▶️ Read TechDispatch here https://europa.eu/!qWvRYk

28