Kyutai, the first European private-initiative laboratory dedicated to open research in artificial intelligence, was launched today at Station F (Paris) in the presence of its scientific team and its three co-founders, Xavier Niel, Rodolphe Saadé and Eric Schmidt.
Kyutai, an AI open-science lab
Kyutai is a non-profit laboratory entirely dedicated to open research in artificial intelligence (AI). Its objective is to tackle the main challenges of modern AI, particularly by developing large multimodal models – using text but also sound, images, etc. – and by inventing new algorithms to enhance their capacities, reliability and efficiency. To do this, the laboratory will use the computing power made available to it by Scaleway, an iliad Group subsidiary. Scaleway’s supercomputer has the highest-performance computing power for AI applications deployed to date in Europe.
Resolutely committed to the democratization of AI, Kyutai is positioning itself as a leading player in AI open science. Its ambition is to share its advances with the entire AI ecosystem – the scientific community, developers, companies, society at large and decision-makers in democracies.
Kyutai will also contribute to the training of future AI experts, by offering internships to students on Master’s programs and supervising PhD students and postdocs.
A private, non-profit initiative
Kyutai was founded jointly by the iliad Group, the CMA CGM Group and Schmidt Futures.
The iliad Group and the CMA CGM Group have each contributed €100 million to funding the research laboratory.
Schmidt Futures has also joined as a co-founder, forging a strong partnership based on a shared commitment to this project.
With nearly €300 million already invested in it, Kyutai is seeking to bring other private investors on board. The co-founders are inviting other entities to join them to provide long-term financing for the work of this non-profit organization.
An exceptional scientific team based in Paris
Kyutai has already brought on board a team of researchers with outstanding academic and business backgrounds. Kyutai’s offices are based in Paris.
Alexandre Défossez
Alexandre studied math and physics at École Normale Supérieure (Paris) before specializing in applied mathematics for machine learning, graduating with an MSc from École Normale Supérieure (Saclay). He completed his PhD in collaboration with Inria and Facebook AI Research (FAIR) in Paris, covering both the theoretical properties of the training of deep learning models and their applications for audio source separation. In 2020, he joined FAIR as a research scientist, continuing his work on audio modeling. He led the research and development for the open source AudioCraft framework, featuring many state-of-the-art audio models, such as EnCodec (compression/tokenization), AudioGen, and MusicGen (audio and music generation). With a strong interest in science at large, he has also made key contributions to the field of computational neuroscience alongside Dr. Jean Remi King.
Edouard Grave
Edouard is a research scientist with eight years’ experience in the corporate research labs, Apple MLR and Facebook AI Research (FAIR). Before that, he did two postdocs, at UC Berkeley and Columbia University. He obtained his PhD in computer science from Université Paris VI and graduated from Ecole polytechnique with an MSc in machine learning and computer vision. Edouard’s research interests revolve around large language models, natural language processing and efficient deep learning. While at FAIR, he co-led the development of the LLaMA large language model and the fastText library.
Hervé Jégou
After obtaining a PhD in compression and error-correcting codes, Hervé was hired as a research scientist at Inria, where he invented algorithms for searching in very large databases of images and videos. He joined Facebook AI Research when the Paris laboratory was founded in 2015, serving as manager and then as director. His research interests include computer vision, efficient training of neural networks, compression, data protection and indexing. He invented, in particular, the “product quantization” compressed-domain search algorithm utilized by many tech and vector-search companies. The Faiss library that he founded at FAIR is one of the most, if not the most, popular open-source vector search library.
Laurent Mazaré
After studying applied mathematics and computer science at Ecole Polytechnique, Laurent received a PhD at the intersection of cryptography and formal methods. He then moved to the finance industry, working as a quantitative strategist at Goldman Sachs. In 2013 he joined Jane Street as a software developer, working on distributed systems, trading analysis and strategies. In 2017 he moved to DeepMind, working as a research engineer in their computer vision group. He came back to Jane Street as a researcher in 2018 and spent his time on the equities desk creating strategies based on new modeling techniques, as well as building infrastructure for research. Laurent has also made significant open-source contributions over the years, with the two most notable being among the most used machine learning libraries for Rust.
Patrick Pérez
A graduate in applied mathematics from École Centrale Paris, Patrick has more than thirty years’ experience in computer vision and machine learning, shared between public and private labs: Brown University as a postdoc and Inria as a PhD student, researcher and senior scientist in the public sector; and in the private sector, Microsoft Research as a research scientist, Technicolor as Distinguished Scientist and Valeo as Scientific Director and VP of AI. During his career, Patrick has explored various fields of applications (medical, cinema and media, automotive), processing and analyzing signals from all types of sensors (cameras, MRI, microphones, radar, laser scanners), and has built up and supported long-term research teams. Some of his most recent projects with the Valeo.ai team led to embarking transformer-based 360-degree visual perception in cars and generating visual counterfactual explanations for camera-only autonomous driving systems.
Neil Zeghidour
Neil joins Kyutai from Google DeepMind, where he started and led a team working on generative audio, with contributions including Google’s text-to-music API, a speech-to-speech translation system that preserves your voice, and the first audio codec entirely based on neural networks. He previously spent three years at Facebook AI Research, working on automatic speech recognition and audio understanding. He graduated with a PhD in machine learning from Ecole Normale Supérieure (Paris), and holds an MSc in machine learning from Ecole Normale Supérieure (Saclay) and an MSc in quantitative finance from Université Paris Dauphine. In parallel with his research activities, Neil teaches speech processing technologies at the École Normale Supérieure (Saclay).
The Scientific Committee: three internationally renowned researchers
Kyutai’s Scientific Committee is made up of three leading AI specialists:
Yejin Choi: Yejin is a South Korean scientist specialized in natural language processing and computer vision. In 2018, she joined the Allen Institute for AI (AI2), and in 2020 became a Professor at the University of Washington. Her research focuses on the many ways in which human intelligence differs from that of large AI models. In 2022, Yejin became a MacArthur Fellow and was awarded the Foundation’s prestigious “genius” grant, and in 2023 Time Magazine named her one of the 100 most influential people in AI.
Yann LeCun: Yann is a French researcher specialized in artificial intelligence, machine learning, computer vision and robotics. He is one of the pioneers of deep learning, and the inventor of convolutional neural networks particularly used in image analysis. His work earned him the 2018 Turing Award alongside two other researchers. Yann is a professor at New York University and Chief AI Scientist at Meta where he founded Facebook AI Research (FAIR).
Bernhard Schölkopf: Bernhard is a German researcher known worldwide for his work in machine learning. After spending time working for corporations, including at AT&T Bell and Microsoft Research laboratories, he joined the Max Planck Institute for Intelligent Systems in Tübingen as director, where he headed the empirical inference department. He is also an Affiliated Professor at ETH Zürich and President of ELLIS (European Laboratory for Learning and Intelligent Systems). He has received a number of awards for his research work, including the Leibniz Prize in 2018 and the Royal Society Milner Prize in 2014.
Xavier Niel, Chairman and Founder of the iliad Group: “Europe’s got everything it takes to win the AI race. French researchers are known as being the best in the world. We have all the computing power that’s needed. And we have a hyper-dynamic AI ecosystem. By creating an AI open research lab here in Paris, we’re stepping up the pace even further. Kyutai will give us ultra high-performing, reliable AI models, which the entire European AI ecosystem will be able to benefit from. I’m delighted to see that many private partners have joined us in this venture. Because together we go further!”
Rodolphe Saadé, President and CEO of the CMA CGM Group: “Through this new investment in artificial intelligence, I would like the younger generation to benefit from all the opportunities that this technology has to offer. Thanks to Kyutai, researchers will have all the resources they need to create and invent the world of tomorrow. It’s a venture that will bring together the world’s best experts to address the challenges of AI. The CMA CGM Group’s decision to get involved was also driven by a desire to place France and the rest of Europe at the forefront of artificial intelligence research. We need to create a dynamic and innovative ecosystem to keep our businesses competitive. Joining forces is the best way to prepare for the future.”
Patrick Pérez, CEO of Kyutai: “We’re embarking on an extraordinary adventure with the launch of Kyutai! It’s an adventure of cutting-edge artificial intelligence with open research available to everyone. To achieve our ambitions, we’re building an exceptional research team and are immersing ourselves in the major challenges of modern AI. We’re looking forward to sharing with the community the first results of our open research.”
