Peter Diamandis, founder of XPRIZE and co-founder of Singularity University, believes artificial intelligence will soon deliver not just intelligence but genuine wisdom, fundamentally changing what it means to be human. Yet in a recent debate with astrophysicist Brian Keating on the "Into the Impossible" podcast, a different picture emerged: one where the most advanced AI systems may have already hit a wall that no amount of computing power can overcome.

Diamandis has spent decades building the future. Beyond XPRIZE, he co-founded Singularity University with futurist Ray Kurzweil and launched Human Longevity with geneticist Craig Venter. In his new book with Steven Kotler, "We Are as Gods," he argues that humanity is crossing a threshold into godlike capability. His central thesis: artificial general intelligence (AGI), the theoretical point where AI matches or exceeds human intelligence across all domains, will generate wisdom by simulating billions of possible outcomes.

But Keating, a Chancellor's Distinguished Professor of Physics at UC San Diego, offered a direct counterargument grounded in experimental evidence. He and one of his students ran a test on large language models (LLMs), which are AI systems trained on vast amounts of text to predict and generate human language. They trained these models exclusively on physics papers published before 1911, the same knowledge available to Albert Einstein. The result: the models could not reproduce what Einstein accomplished with that same data.

What Does This Experiment Reveal About AI's Limits?

The pre-1911 Einstein test cuts to the heart of a fundamental question: Is wisdom simply a matter of scale, or does it require something more? If wisdom were purely a function of processing power and data volume, Keating argued, then an LLM trained on all the physics knowledge available in 1911 should theoretically be able to derive Einstein's insights. It did not.

This finding suggests that current AI architectures, built on graphics processing units (GPUs) and LLMs, may already represent a local maximum, a point beyond which incremental improvements yield diminishing returns. Keating noted that while AI is outpacing most mathematics PhDs in certain narrow domains, the ceiling of what these systems can achieve remains unknown and possibly lower than many technologists assume.

Diamandis fired back at what he called doomers and skeptics, but the conversation revealed a deeper tension: the difference between simulation and genuine understanding. Keating emphasized that wisdom may require embodiment, lived experience, and the ability to navigate genuine uncertainty, not merely the simulation of billions of hypothetical scenarios.

How Should Humanity Prepare for an AI-Transformed World?

Despite their disagreement on AI's ultimate potential, both thinkers identified concrete areas where artificial intelligence is already reshaping society and where preparation matters:

• Legal and Regulatory Barriers: Diamandis argued that AI will eventually untangle the legal systems currently blocking technological progress, suggesting that regulatory frameworks designed for an earlier era may become obsolete as AI capabilities accelerate.
• Training Data Quality: Keating raised a critical concern: dystopian training data may be making AI systems more dangerous by encoding harmful patterns and biases into their decision-making processes, a problem that scale alone cannot solve.
• Human Purpose in Post-Scarcity: Both discussed what happens to human meaning and motivation when scarcity disappears, with Diamandis suggesting that humanity will split into creators and consumers, while Keating questioned whether such a division is inevitable or desirable.

The conversation also touched on speculative but serious topics. Diamandis outlined what he calls the "five forks of humanity," divergent paths shaped by technology: radical life extension through longevity science, brain-computer interfaces (BCIs) that merge mind and machine, off-planet colonization and human speciation, the division between creators and passive consumers, and the uploading of human consciousness into digital substrates.

Diamandis

On the question of extraterrestrial life, Diamandis presented his own theory of the Fermi paradox, the puzzle of why we have not detected alien civilizations despite the vastness of the universe. He suggested that humanity may itself be part of someone else's biosphere experiment, a thought experiment that reframes humanity's place in the cosmos.

What Concrete Developments Are Already Underway?

Beyond theoretical debate, both hosts discussed real-world projects advancing the boundaries of human capability. Diamandis mentioned the Future Vision XPRIZE, an initiative designed to shape how AI systems are trained and deployed. He also highlighted David Sinclair's epigenetic age-reversal trials, which are now underway in human eyes, representing a tangible step toward reversing biological aging.

These projects embody the tension at the heart of their debate: technology is advancing rapidly, creating both extraordinary opportunities and genuine risks. The question is not whether AI will change humanity, but whether that change will be guided by wisdom or merely by the simulation of it.

Diamandis closed with a perspective on progress that both thinkers seemed to share: what you accomplish between breakfast and dinner today would be considered godlike to your grandparents. Humanity has simply stopped noticing the miracles of ordinary life.