French entrepreneur and CEO of Iliad, Xavier Niel, shared further details about his vision for an AI research center in Paris earlier today at Scaleway’s ai-PULSE event.
The newly established facility, Kyutai, will operate as a non-profit organization, backed personally by Niel, concentrating on artificial general intelligence. It will engage in collaborative research projects and open-source initiatives with PhD candidates, professionals, and experts. When Iliad first introduced this research hub, Niel pledged €100 million (equivalent to 109 million at the current exchange rate) towards this endeavor.
During the event, Xavier Niel announced, “Thanks to some incredible supporters present here today, we have now secured close to €300 million for the development of this project.” Among these supporters is another prominent French figure, Rodolphe Saadé, the CEO of the renowned shipping and logistics giant CMA CGM, who has also committed €100 million. Additional contributions have been made by Eric Schmidt’s foundation and several undisclosed donors.
As Kyutai continues to welcome further contributions, this marks just the initial phase of its journey. Following the announcement, Saadé mentioned at a press briefing, “The presence of numerous journalists here today indicates that this project is likely to attract additional investors.”
Given Kyutai’s focus on fundamental models, substantial computing power will be essential. Fortunately, Iliad’s cloud division, Scaleway, recently acquired 1,000 Nvidia H100 GPUs. These state-of-the-art GPUs are crucial for training and inference tasks and will be tailored to Kyutai’s needs.
The recruitment process for Kyutai’s core academic team has already commenced. Esteemed professionals including Patrick Perez, Edouard Grave, Hervé Jegou, Laurent Mazaré, Neil Zeghidour, and Alexandre Defossez presented their backgrounds and outlined their research plans earlier today. These individuals bring experience from reputable entities such as Google’s DeepMind, Inria, and Meta’s AI research division FAIR.
Heading the research laboratory will be Patrick Perez, a former Valeo employee. Additionally, Kyutai has enlisted a panel of medical experts, including Yein Choi, Yann LeCun, and Bernhard Schölkopf, distinguished figures in the field of AI. They will convene regularly to review progress and provide insights to the team.
Kyutai’s commitment to enabling researchers to publish their findings is a key aspect that sets it apart. Niel emphasized during the press conference the importance of fostering research and advancing knowledge beyond individual gains.
While Kyutai represents a new entrant in the realm of open AI research labs, it joins the ranks of organizations like OpenAI, which initially operated as a non-profit entity. However, the landscape shifted significantly after Sam Altman’s full-time involvement in OpenAI in 2019, leading to collaborations with tech giants like Microsoft and a shift towards a more conventional structure.
Kyutai aims to release its models as open source, distinguishing its approach as “accessible technology.” The team intends to share not only the model designs but also the source code and methodologies behind them. Alexandre Defossez mentioned during the press briefing that while speed is not their primary focus compared to other initiatives, they aim to provide a scientific foundation and transparent results. The team anticipates sharing insights within a month.
The positive reception and engagement with the Mistral 7B model by community members underscore the potential impact of Kyutai’s open-source approach. Laurent Mazaré, another member of Kyutai’s team, praised the model’s accessibility as a significant achievement.
The utilization of Kyutai’s research outcomes by private enterprises for commercial applications, and the comparative effectiveness of research labs versus private entities in releasing foundational models, present intriguing prospects for the industry.
Photo: TechCrunch/Romain Dillet
In a prerecorded video message played at the event, French President Emmanuel Macron expressed his support for open-source initiatives, emphasizing their value for France.
France’s Approach: Regulating Use Cases, Not Models.
Macron also utilized the platform to clarify France’s stance on the AI Act within the European context, advocating for the regulation of AI use cases rather than the creators of the technology. France has engaged in trilogue negotiations with the European Parliament, Commission, and Council to influence the AI Act.
Macron stressed that rules are essential for safeguarding citizens, economic players, and democracy, emphasizing the need to govern usage rather than the technology itself. He highlighted the ongoing collaborative efforts within the European framework to shape regulations for artificial intelligence.
Xavier Niel echoed France’s position on AI regulation, asserting that Europe must prioritize innovation over stringent regulations to remain competitive globally. He cautioned against regulatory hurdles hindering the progress of emerging AI players in Europe.
The conversation shifted towards the potential for European AI firms to achieve significant success, prompting Niel to envision a future where Europe plays a leading role in AI innovation.
