[London Data Week Workshop] How Good are AI Language Models in Global Languages?

​​Join us for a one-of-a-kind event bringing together AI researchers and London’s language communities to ask: "How Good are AI Language Models in Global Languages?"

​This interactive day-long workshop organised by grassroots AI researchers from across Africa, the Alan Turing Institute, and more is held as part of London Data Week. It will cover interactive demos of the latest models, lightning talks from leading researchers on the latest research on multilingual and multicultural AI, as well as breakout sessions focused on how language communities can help shape the development of AI systems and address critical ethical considerations. 

​When: 10am - 4pm, July 2nd, 2024

​Where: Great Hall, Strand Campus, King’s College London, Strand, London, WC2R 2LS (room accessibility guide here)

​Getting there: Visitors without a King’s ID card need to be checked in at the Strand reception, marked with a red arrow as the Main Entrance on this map.

​Schedule

​We're lucky to be joined by fantastic speakers and facilitators from Hugging Face, Cohere for AI, ML Commons, the IRC, and more!

• ​9:30 - 10:00 - Arrivals

• ​10:00 - 10.30 - Introduction (hybrid)

• ​10.30 - 11.30 - Lightning talks (hybrid)

  • ​Aidan Peppin and Marzieh Fadaee (Cohere for AI) will present on Aya, a global initiative that brings together 3,000+ researchers across 119 countries to build models and datasets in 101 languages through open science.

  • ​Gina Moape (Mozilla) will present on Mozilla's Common Voice, a crowdsourcing initiative to help teach machines how real people speak in 100+ languages.

  • ​Yong Zheng-Xin (Brown University) will present on his Lab's work on multilingual AI safety (asynchronous).

  • ​Miaoran Zhang (Saarland University) will present on her work on the ability of models to learn new languages.

• ​11:30 - 12:30 - Breakout sessions (in person only)

  • ​Facilitators from the Alan Turing Institute, IRC's Airbel Lab, and more will facilitate discussions around key research and ethical questions related to multilingual AI.

• ​12:30 - 1:30 - Lunch

• ​1.30 - 2.30 - Benchmarking presentations (hybrid)

  • ​Bertie Vidgen (ML Commons)

  • ​Avijit Ghosh (Hugging Face), Jennifer Mickel (UT Austin), and Usman Gohar (Iowa State University) will present on Hugging Face’s groundbreaking social impact evaluations work.

• ​2:30 - 3:30 - Breakout sessions (in person only)

  • ​Attendees will co-create ethical principles to guide the development of multilingual benchmarks for low-resource languages.

• ​3.30 - 4 - Wrap up

• ​4 onwards - informal networking

​Sponsorship

​We are grateful to have sponsorship for the event provided by LOTI, OpenAI, and Kings College London.