
Tech • AI • Robotics • Game
A French tech and AI program outlined an editorial schedule focused on understanding, debating and testing artificial intelligence, while highlighting a new Mistral model and a wide-ranging discussion with philosopher Alexandra Prégent on the ethics and social impact of AI.
The program is now structured around three recurring themes spread across the week: understanding on Mondays, thinking on Wednesdays and acting on Fridays. The goal is to move beyond product demos and cover AI from technical, practical and philosophical angles, with each edition also including a news review focused on the latest developments in the sector.
A major item in the upcoming news review is a newly released Mistral model repeatedly teased as a “big cat.” The launch was described as delayed by roughly three months, but early reactions pointed to encouraging results on some specific benchmarks, including performance said to surpass future-generation competitors on narrow tasks. The broader point made around the launch was that Mistral remains constrained by smaller data-center capacity than giants such as OpenAI, yet is still producing competitive systems.
A separate segment scheduled for Friday is set to showcase a hands-on test of the newest AI model capabilities. The planned demonstration will focus on practical settings, tuning options and real-world uses rather than abstract commentary, with emphasis on how different configurations can change behavior and outcomes.
The central Wednesday discussion featured philosopher Alexandra Prégent, introduced as working on the ethics of AI at the CEA. The exchange focused on the place of humans alongside machines, the social and political implications of AI deployment, and the need to treat these questions as public matters rather than purely technical ones. Her intervention was presented as direct, intellectually demanding and rooted in sustained research.
The discussion stressed that AI policy should not be left only to engineers, executives or product promoters. The case of Alexandra Prégent was used to argue for more visibility for researchers who study AI’s ethical and societal consequences, particularly in mainstream public debate, where highly mediatised tech figures often dominate attention.
One of the most notable ideas attributed to Prégent is the prospect of a future with “two-speed AI.” In that scenario, the most powerful systems would be accessible only to a limited segment of the population, while others would be confined to lower-grade tools. That possibility was framed as a serious line of inquiry because it would deepen existing inequalities in knowledge, productivity and decision-making power.
Another theme concerned low-cost classifier models such as Jeev and comparable tools released by other firms, including OpenAI. These systems are designed less to reason than to classify or route tasks, for example by answering whether a message is spam or by assigning a probability to a category. They were described as fast, cheap and useful in automation workflows, but also as part of a broader shift toward delegating decisions to machines.
A recurring argument was that emotional reactions around AI are often heightened by imagery of robots rather than by clear descriptions of software systems and their actual uses. Framing AI as programs that classify, generate or assist could lower some of the confusion, though not the legitimate concerns. Those concerns, as highlighted in the discussion, relate less to science fiction than to governance, incentives, concentration of power and social effects already visible today.
The broadcast also highlighted a broader community infrastructure including a dedicated website, a subscription-based club, forums, articles and live interaction across platforms such as Twitch, YouTube, TikTok, LinkedIn and Facebook. Viewers were encouraged to use these spaces to continue discussions, especially on interviews dealing with long-term AI governance and human autonomy.
The discussion underscored a widening gap between rapid AI product cycles and the slower ethical reflection needed to govern them. As new models from Mistral and others arrive, the central question is no longer only what AI can do, but who controls it, who benefits from it and how humans remain meaningfully in the loop.
Ask a question