Eric Schmidt on the Robotics Race, Singularity Timeline, and Energy Shortage | 241

In this talk

1. 2:21

   AI agents and the pace of change

   Schmidt estimates that AI’s effects are still early, while reasoning agents already help people and could soon take on more work. He describes a possible path to recursive self-improvement, but says current systems do not yet meet that standard. He points to AI agents writing and testing software overnight as an example of how work is changing.

2. 10:45

   Skills, jobs, and responsibility

   Schmidt expects programming to shift from writing code to directing systems, and says strong technical judgment will remain valuable. He urges universities to teach students how to use AI tools, while Diamandis argues that students should encounter them earlier. Schmidt also flags job disruption, risks to young people, and unpredictable behavior when agents from different providers work together.

3. 17:49

   Power and infrastructure constraints

   Schmidt identifies electricity as a major limit on AI growth and cites an estimated U.S. power shortfall of 92 gigawatts by 2030. He says data centers may consume 10% of U.S. electricity and describes facilities around 400 megawatts in size. He argues that efficiency gains may lead to more computing demand, and discusses space-based data centers as a possible business option with engineering challenges.

4. 29:05

   Robotics and competition with China

   Schmidt calls China a competitor and says its low-cost robotic hardware appears to be ahead. He links that advantage to China’s electric vehicle supply chain, manufacturing scale, and expertise in motors. He warns against ceding the low-cost robotics market and notes that complex, high-skill assembly work, such as building rockets, remains difficult to automate.

5. 35:31

   Safety, governance, and American values

   Schmidt says researchers have not settled how recursive self-improvement might work, though limited demonstrations provide some evidence. He discusses biological and societal risks and says a serious incident could prompt governments to coordinate, while clarifying that he is describing a possibility rather than advocating it. He wants U.S. AI development to preserve freedoms and protect children, with political will to address risks without abandoning innovation.