Adam Brown — Bubble universes, space elevators, & AdS/CFT

Our understanding of the universe has undergone radical shifts over the last century, evolving from a static picture to one of accelerated expansion driven by mysterious dark energy. Today, we stand at another precipice, not just in cosmology, but in how we derive these truths, as artificial intelligence begins to intersect with theoretical physics. In a wide-ranging discussion, Adam Brown—a theoretical physicist at Stanford and a lead researcher at Google DeepMind—bridges the gap between abstract high-energy physics and the emerging reasoning capabilities of AI. From the terrifying potential of vacuum decay to the engineering limits of black holes and the philosophical implications of holographic duality, Brown offers a glimpse into both the ultimate fate of our cosmos and the tools we will use to understand it.

Key Takeaways

• The universe’s fate is not sealed: While dark energy currently suggests a heat death, the laws of physics may allow advanced civilizations to engineer a "vacuum decay" to transition to a more hospitable state.
• Energy is not globally conserved: In an expanding universe, general relativity allows for the creation of energy, potentially enabling the seeding of new "bubble universes."
• Black holes have strict engineering limits: Mining energy from black holes is theoretically possible but constrained by material science limits, specifically the maximum possible tensile strength of a string.
• Information scales with area, not volume: The Bekenstein-Hawking bound suggests that a 3D gravitational world can be mathematically equivalent to a 2D non-gravitational one, a core concept of the holographic principle.
• AI is currently a tutor, not an Einstein: Large Language Models (LLMs) excel at synthesizing existing physics knowledge and debugging understanding, but they have not yet demonstrated the capacity for novel conceptual leaps like inventing General Relativity.

The Instability of the Cosmos and Vacuum Engineering

Our cosmological worldview has shifted dramatically, from a static universe to one expanding at an accelerating rate due to the cosmological constant (dark energy). This acceleration implies a grim future: a "heat death" where galaxies drift beyond our causal horizon, limiting the free energy available to future civilizations. However, theoretical physics offers a speculative but mathematically consistent escape hatch known as vacuum decay.

• The threat of acceleration: The discovery that the universe is expanding faster and faster was, in many ways, catastrophic news for long-term prospects, as it implies a finite amount of accessible interactions and energy in our future.
• Nature of the vacuum: A "vacuum" in physics is not empty space but a local minimum in the energy landscape; our current laws of physics may simply represent one metastable state among many.
• Vacuum decay mechanics: It is possible to transition from a higher energy vacuum to a lower one, effectively changing the local laws of physics—specifically constants like the cosmological constant—to avoid heat death.
• Engineering the transition: While quantum fluctuations will eventually trigger this decay randomly, an advanced civilization might deliberately engineer a controlled collapse to reach a more favorable vacuum state.
• The danger of "bad" vacuums: Most possible vacuum states are inhospitable to complexity or life; successfully engineering a transition requires precise targeting to ensure the new constants (like electromagnetism) remain compatible with biological existence.
• Technological requirements: Triggering a vacuum decay would likely require a device similar to a particle accelerator but with vastly superior control to create a stable, expanding bubble without collapsing into a black hole.

Proceed with caution in these theories where there's lots and lots of vacuums out there.

Bubble Universes and the End of Energy Conservation

The concept of "bubble universes" challenges our intuitive grasp of conservation laws. In the framework of General Relativity, energy conservation is a local law, not a global one. This opens the door to scenarios where new universes are spawned, creating vast amounts of matter and energy out of the expansion of space itself.

• Global non-conservation: In an expanding universe, total energy is not conserved; as space expands, energy can be added to the system, a feature intrinsic to Einstein’s General Relativity.
• Seeding universes: It is theoretically possible to spawn a bubble universe that expands indefinitely; while it may look small from the outside, it could contain infinite volume on the inside.
• The origin of the Big Bang: Our own universe may be the result of a vacuum decay event or a bubble formed within a larger, metastable ancestor universe.
• Anthropic selection: The reason our universe appears finely tuned for chemistry and life may simply be that we exist in a bubble where these conditions hold, while countless other inhospitable bubbles remain empty of observers.
• The disconnect from the creator: Even if a civilization seeded our bubble, the physical disconnect is absolute; the "creators" would be outside our causal horizon, and we would likely expand at the speed of light away from them.
• Quantum origins of structure: The massive structures we see today (galaxies, clusters) originated as microscopic quantum fluctuations in the early universe, which were stretched across the cosmos by inflation.

Black Hole Mining and Fundamental Limits

For a civilization facing the end of stellar energy, black holes represent the ultimate battery. They are the most efficient energy storage devices in nature, converting matter entirely into radiation via Hawking decay. However, extracting this energy quickly presents profound engineering hurdles that seem to be baked into the fundamental laws of physics.

• The slow leak of Hawking radiation: Left alone, a solar-mass black hole would take exponentially longer than the current age of the universe to evaporate, releasing energy too slowly to be useful.
• The space elevator analogy: Proposals to "mine" black holes involve lowering a box or scoop near the horizon to capture radiation and winching it back out, similar to a space elevator on Earth.
• Material science constraints: On Earth, space elevators are limited by the tensile strength of materials (carbon nanotubes are barely sufficient); for a black hole, the required strength is exponentially higher.
• The speed of light bound: Brown’s research demonstrates that the tensile strength required to mine a black hole efficiently is exactly the maximum strength allowed by the laws of physics (where the speed of sound in the rope equals the speed of light).
• The rope failure: A rope strong enough to mine a black hole is so heavy that it can only support its own weight, leaving zero capacity to lift a payload; this renders rapid mining impossible via this method.
• Baryon number violation: Despite mining difficulties, black holes remain unique because they can violate baryon number conservation, theoretically allowing for 100% mass-to-energy conversion, far surpassing nuclear fusion.

The Holographic Principle and the Nature of Reality

The study of black holes has yielded perhaps the most profound insight in modern physics: the holographic principle. This stems from the realization that the information capacity of a region of space is determined not by its volume, but by its surface area.

• Bekenstein-Hawking Entropy: The entropy (information) of a black hole is proportional to the area of its event horizon divided by the Planck area, a formula connecting quantum mechanics, gravity, and thermodynamics.
• Volume vs. Area: Intuitively, information should scale with volume (like stacking hard drives), but gravity enforces an area limit; packing too much information into a volume inevitably collapses it into a black hole, capping the storage at the surface area.
• Dimensional reduction: This implies that a 3D gravitational world can be fully described by a 2D theory without gravity, suggesting the "bulk" of the universe may be a holographic projection.
• AdS/CFT Correspondence: Juan Maldacena’s famous duality provides a concrete mathematical example where a gravitational theory (AdS) is exactly equivalent to a quantum field theory (CFT) on its boundary.
• The reality of duality: Philosophically, neither the 3D gravitational description nor the 2D boundary description is "more real" than the other; they are precise isomorphisms, different languages for the same physical reality.
• Cosmological challenges: While AdS/CFT is a triumph for quantum gravity, it describes a universe with a negative cosmological constant; applying holography to our expanding, positive-curvature universe remains an active, difficult area of research.

Are you an AdS dreaming you're a CFT, or a CFT dreaming you're an AdS?

Artificial Intelligence as the New Physicist

As a researcher at Google DeepMind, Brown observes the rapid evolution of AI reasoning. While current Large Language Models (LLMs) are not yet replacing theoretical physicists, they are fundamentally changing the workflow of science and rapidly climbing the ladder of abstraction.

• The tutor model: Currently, the most effective use of AI for physicists is as an omniscient, non-judgmental tutor that can debug misconceptions about advanced topics (e.g., squeezed light in gravitational wave detectors) instantly.
• Graduate-level competence: In just a few years, LLMs have gone from failing to acing graduate-level general relativity exams, demonstrating a capability to translate word problems into mathematical formalism.
• The interpolation debate: While LLMs are often dismissed as "stochastic parrots" or interpolators, sufficiently high-level interpolation indistinguishably resembles creativity; deriving General Relativity from Newtonian physics could be viewed as interpolation in a vast enough abstract space.
• The Einstein threshold: The ultimate test for AI is whether it can replicate a conceptual leap equivalent to Einstein inventing General Relativity from thought experiments—a feat Brown speculates may be the "terminal step" before AI encompasses all human intelligence.
• Finding patterns in data: Beyond theory, AI is being deployed to process exabytes of astronomical data, potentially finding patterns and anomalies that human researchers would inevitably miss.
• Representation engineering: Future AI breakthroughs in physics may come from the models inventing new mathematical notations or representations, much like Einstein’s summation convention or Penrose diagrams facilitated human progress.

The Sociology of Science and Human Agency

Beyond the equations, the progress of physics is deeply human. It relies on institutional cultures, the willingness to take risks, and sometimes, the refusal to follow orders. Brown’s investigations into history and his personal experiences highlight the chaotic, non-linear nature of discovery and survival.

• The necessity of poor calibration: Theoretical physicists are often irrationally optimistic about their own theories; this overconfidence is individually epistemically flawed but collectively optimal, ensuring that many disparate paths are explored with vigor.
• Victims of success: The stagnation in particle physics is partly due to the Standard Model's incredible success; it explained experimental data so well that it left few loose threads to pull on for new discoveries, unlike the anomalies that drove early 20th-century physics.
• Nuclear insubordination: Brown’s historical research suggests that roughly 50% of nuclear weapons dropped in combat (specifically Nagasaki) were deployed in violation of direct orders regarding visual targeting, highlighting the precarious nature of chain-of-command in existential scenarios.
• The cost of big science: The extreme cost of modern particle colliders (tens of billions) competes with other scientific endeavors, raising questions about whether smaller, cleverer experiments (like BICEP) offer better returns on investment.
• Counterparty modeling: Brown’s extensive hitchhiking experience serves as a lesson in game theory and psychology; understanding the motivations of others is as crucial in human interaction as it is in modeling physical systems.
• The value of variance: Just as physics benefits from exploring diverse vacuums, human understanding benefits from high-variance interactions—engaging with the "tails" of the distribution offers insights that the average cannot provide.

It does kind of look like 50% of all the nuclear weapons ever dropped in combat were dropped against direct orders.

Conclusion

The intersection of cosmology and artificial intelligence places us at a unique moment in history. We are beginning to understand the boundaries of our physical reality—from the tensile strength of strings mining black holes to the information density of the horizon—while simultaneously building synthetic minds that may one day surpass our ability to reason about them. Whether we are destined to engineer the vacuum to save our descendants or simply serve as a transitionary step toward digital physicists, the pursuit of these fundamental truths remains the most ambitious project of our species.