Redazione RHC : 12 July 2025 09:40
During an interview on the Manifold podcast, Omar Shams, head of the Google Agents project, emphasized that besides chips, the other major limitation to AI development is energy. It’s often thought that innovation comes only from more powerful hardware, but without adequate power capacity, any advanced algorithm remains on paper.
Shams compared electricity growth between China and the United States, highlighting how the former increases every year as much as the cumulative production of countries like the United Kingdom and France. In the US, however, infrastructure tends to stagnate due to regulatory constraints. A lack of sufficient power means that investments in AI could remain untapped at the operational level.
To overcome this constraint, the manager proposed solutions that seem like science fiction: space-based Solaris power plants positioned in lunar orbit or at Lagrange points, capable of generating and transmitting energy without the limitations of Earth. While acknowledging the difficulty and astronomical costs of such an undertaking, he believes it is a viable long-term path.
Shams recounted his experiences at Mutable, a pioneering startup in the field of AI-assisted programming tools, founded almost simultaneously with GitHub Copilot. Among the key projects was “AutoWiki,” a system that analyzes code to automatically generate wiki-style documentation: a way to simplify understanding and better exploit language models. .
Speaking of the impact on jobs, Shams explained that “smartagents” are revolutionizing the software industry: repetitive and basic tasks are being delegated to AI, and the number of junior programmers is decreasing, leaving room for more senior figures with technical experience and leadership skills to integrate and manage these agents.
Finally, Shams noted that the success of AI projects will increasingly depend on that “tacit knowledge” acquired through years of practice, intuition, and judgment—elements that no algorithm can replace. For young people investing in their future in the sector, the advice is clear: in addition to studying, you need to gain plenty of “on-the-job” experience and learn to evaluate real-world contexts.