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

Schmidt describes this as the new norm among Bay Area programmers. The AI acts as a tireless junior team, exploring solution spaces while the human sleeps. This flips the ratio of human-to-AI code from 80/20 to 20/80. The human's role shifts from writing code to defining success criteria and reviewing output. Schmidt notes that this workflow would have been impossible just months ago when AI was merely autocomplete.

Schmidt argues that just as students learn to write, they must learn to prompt AI systems to unlock creativity and productivity. He sees this as urgent, given the pace of change. Diamandis extends the idea to high school, noting his own children are already using AI. Schmidt agrees but cautions about vulnerable teenagers and the need for guardrails.

Schmidt interjects this principle when Diamandis discusses closed loops in robotics and AI self-improvement. He reframes it as 'learning loops' and emphasizes that accelerating learning is the key to winning. This is a meta-protocol for strategy.

Schmidt notes that Elon Musk vertically integrates because it's the only way to drive costs down when supply chains don't exist. He contrasts this with China's well-built-out supply chain, but observes that US humanoid robotics companies are also going vertical. He cautions that for complex, low-volume, high-skill assembly (like rocket building), current robots cannot replace human expertise.

Schmidt mentions using LLMs to do deep research when he joined a rocket company, highlighting how AI can compress years of domain expertise into quick insights.

Schmidt describes a startup programmer who writes a spec and a test function at 7pm, then goes to sleep while the AI works until 4am, producing inventions that would have taken 6 months and 10 programmers at Google. He notes that the top programmers will become more valuable as directors of programming systems, while the rest will be displaced. The shift is driven by LLMs that can reason longer and produce higher-quality tokens, not just larger context windows. Schmidt sees this as the end of hand-coding: 'No one will ever do that again after the end of this year. It'll be like riding a horse.'

Schmidt spent the prior week reviewing RSI (recursive self-improvement) papers and found no agreement among scientists on the exact approach. He defines real RSI as 'Start now, learn everything, discover things, and tell me what you learned.' That query doesn't work yet. He notes that the labs are leapfrogging each other weekly, but the core breakthrough is still missing. The SF consensus timeline of 2-3 years is based on scaling reasoning agents, not on having solved RSI.

Schmidt argues that robotics can be reduced to actuators and a brain. The EV industry produces the same kind of motors, giving China an expertise the US lacks. He visited Unitree and was impressed. He notes that China's culture of brutal competition, no small talk, and vertical integration at scale makes them formidable. He wants the US to win the robotics revolution, not cede the low end. He also notes that high-end, high-skill assembly (like rocket building) is still beyond robots, but low-skill labor will be swept up.

Schmidt believes that every student, regardless of major, will use AI as their primary creative and analytical tool. He wants a quarter or semester course to be the first thing they learn. He acknowledges age restriction concerns for younger teens but insists universities must adapt. Diamandis pushes for high school, noting his own kids are already using AI. Schmidt concedes it's an improvement but warns about vulnerable teenagers and the need to address mental health risks.

Schmidt clarifies he is not endorsing such an event, but describing the likely path. He imagines a scenario where after a bad incident, leaders from the US and China meet and agree to cooperate despite competition. He notes that without a wake-up call, political will is absent. He also highlights other dangers: impact on teenagers (suicides linked to LLMs), agent orchestration unpredictability, and the need to maintain American values while racing.