Davos 2026: The US-China AI Race, GPU Diplomacy, and Robots Walking the Streets | #225

The AMA question asked how individuals can compete when large players dominate compute. Alex advised following the example of Chinese labs that, when compute-limited, innovated on training efficiency (as seen with DeepSeek and Kimi achieving GPT-4 parity for 1/100th the cost). He stressed picking problems where the leverage from intelligence outweighs the need for brute-force compute. Peter mentioned that he and Eric Schmidt are working on a plan to give entrepreneurs access to the compute they need, suggesting that infrastructure access will improve. The advice aligns with the historical pattern that scarcity drives ingenuity.

Peter shared a conversation with biology (possibly a person) where the idea of making the regulator the bad guy emerged. He argued that every movie needs a villain, and currently it's the technology; switching the villain to the regulator would shift cultural perception. Sem noted that narrative is the only model to shift people at scale, and Peter is working with Google on a big push for positive narratives. Dave pointed out that science fiction, though often inaccurate, shapes business plans; thus media portrayals directly influence innovation pathways. The group agreed that overcoming the amygdala is perhaps the hardest job of leadership.

In the AMA on stopping social unrest, Sem emphasized the ‘B’ (basic) in UBI: giving enough to remove existential fear but not enough to make people happy, so that the economy still thrives on aspiration. He corrected the misconception that UBI is socialist, arguing that in its proper form it is a libertarian idea that dismantles government services and relies on market forces. Peter echoed that stability reduces amygdala activation, and addressing basic needs cuts fear. The group saw universal basic services as a way to provide agency, which is the real antidote to unrest.

Dave observed that all dialogue was about AI, and global presidents were listening to AI discussions, giving hope that leaders are starting to prepare. However, many attendees still treat AI as a cyclical fad that will blow over, missing that it will only accelerate. The hosts described a surreal World’s Fair atmosphere where governments, tech companies, and frontier labs were on equal footing, consuming storefronts. The mismatch between the exponential change and the incremental mindset of nation-state representatives was a recurring theme. Dave noted that next year’s event will be a hundred times more intense, and the fact that Alex was present was itself a bellwether of where the forum is heading.

The hosts emphasized that Demis' outer bound of 10 years is 'tomorrow' in geopolitics, yet the Davos audience seemed to hope AI would blow over like past tech cycles. Dario argued the pace is so fast we should devote almost all effort to navigating the transition. Demis suggested a slightly slower pace would be helpful but requires global coordination—something absent. The conversation also touched on post-AGI purpose: Demis wants to explore the stars with superintelligence, implying that humanity’s massive transformative purpose must transcend nation-state boundaries. Alex stressed that recursive self-improvement may already be priced in, and the more interesting news is what problems AI will unlock, not just the risks.

Sem argued that if energy is the core inner loop, China’s faster power build could give it an edge, but he believes application-layer dominance—where trust and market speed matter—will be the real differentiator, keeping China behind. Alex noted that conventional wisdom holds China’s strength lies in aggressively pushing AI applications into the economy, even if frontier models lag by ~6 months. Dave explained that after GPT-4, frontier labs stopped publishing openly, so the edge from secret innovations (chain-of-thought reasoning, self-organizing agent systems) could diverge the race. Peter Diamandis highlighted his frustration that Hollywood fuels AI pessimism, making the public fear killer robots, and proposed creating films where the regulator is the villain to rebalance the narrative. The segment also linked to the 1950s nuclear overregulation that killed ‘energy too cheap to meter,’ drawing parallels to AI.

Alex explained that the constitution reads like Asimov’s laws, and the radical departure is that Claude participated in writing it, analogous to recursive self-improvement for ethics. The hope is that an AI that genuinely endorses its values will be intrinsically safer. Peter noted that the volume of AI output already outstrips human review, so AI evaluating AI is essential, though it can spiral up or down. Sem praised the Creative Commons license, allowing anyone to improve it, and saw this as a massive net positive that could accelerate a hopeful future. The hosts saw Anthropic as leading on AI rights and personhood, potentially a turning point when history looks back.

Sem was livid, arguing that shifting home heating to solar would slash oil demand, drastically lowering fossil fuel prices for energy-dense applications, making Kapoor’s point structurally flawed. Alex suggested the whole debate is temporary because compact fusion and Dyson swarms will soon make it moot. Peter pointed to data showing wind and solar already outstripping fossil generation in Europe and called for a Manhattan project on solar manufacturing, noting that manufacturing bottlenecks—not technology—are the barrier. Dave highlighted that new electricity demand is overwhelmingly from data centers, which could be moved to orbit, creating a virtuous cycle that forces massive solar panel production. The group concluded that the Manhattan project should be space-based data centers, which market forces are already beginning to drive.

Sem explained that governments excel at incremental microscopic changes, not reinventing the entire economy. The assumption of linear change and stable labor demand is deeply embedded, making them unable to craft a plan when AI breaks both models. Peter added that addressing fear is key: people fear losing their job and shelter, so delivering universal basic services can provide stability and reduce the amygdala-driven anxiety that leads to social unrest. Alex pointed out that China’s industrial policy for robotic automation is a form of plan, but in the West, it’s lacking. The solution, the hosts argued, is not policy alone but giving individuals agency and ensuring equality of opportunity to reach the top.