France and India open joint call on math foundations of AI

This call is small compared with hyperscaler capex, but it targets a crucial bottleneck in the AGI race: theory and evaluation. By funding joint projects on mathematical foundations, optimization, safety and PDE‑based modeling, DST and ANR are explicitly trying to strengthen the scientific underpinnings of AI rather than just chasing bigger models. That kind of work—on convergence guarantees, robustness, and trustworthy behavior—is exactly what’s needed if society is going to keep up with rapidly scaling capabilities.

Strategically, it also tightens the research ties between a fast‑growing AI player (India) and a leading European science system (France). That blurs the old core–periphery picture in AI research and creates a pipeline of students, postdocs and PIs who are comfortable working across borders on the hardest theoretical questions. Over time, those networks can seed new labs and startups that are less dependent on US‑centric ecosystems.

For the AGI timeline, this kind of program tends to be double‑edged: better theory and evaluation can make powerful systems safer, but they can also make training and deployment more efficient. Either way, it increases the odds that when truly general systems emerge, more than a handful of US‑ and China‑based labs will understand how they work and how to test them.