#29 - Dr Anders Sandberg on 3 new resolutions for the Fermi paradox & how to colonise the universe

The Fermi Paradox and Probability Analysis

• Anders Sandberg and Toby Ord argue the Fermi Paradox is not a paradox because previous analyses relied on flawed point estimates rather than uncertainty distributions.
• Multiplying uncertain probability estimates (e.g., life emergence, intelligence evolution) together creates a "tail" where the probability of being alone in the galaxy is significantly higher than point estimates suggest.
• Current models suggest an approximately 8% chance that we are alone in the visible universe and a 20–30% chance within the Milky Way, based solely on optimistic literature.
• The observation of an empty sky updates our parameters, specifically reducing the estimated probability of life and intelligence emerging by several orders of magnitude.
• This evidence suggests the "Great Filter" (the barrier preventing life from becoming a galactic civilization) is likely an early filter (e.g., the emergence of intelligence) rather than a future one (civilization self-destruction).
• The calculation assumes the universe contains roughly $10^{20}$ stars, yet no evidence of megastructures (like Dyson spheres) or energy consumption by other civilizations has been observed in 500,000 surveyed galaxies.
• Sandberg notes that while the mean number of expected civilizations might be in the hundreds, the distribution is so wide that zero or one is a statistically consistent outcome.

Methodological Lessons for Risk Assessment

• Relying on single, elegant arguments for low-probability, high-impact risks (like the Large Hadron Collider safety or nuclear plant reliability) is dangerous because a single undetected error invalidates the entire conclusion.
• Robust risk assessment requires multiple independent arguments with different methodologies; if 20 distinct arguments each have a 1% error rate, the cumulative certainty that the risk is non-existent becomes very high.
• In risk analysis, the "analysis is correct" scenario often yields a near-zero risk, but the "analysis is wrong" scenario can yield a catastrophic risk; therefore, the total risk is dominated by the probability of the analysis being flawed.
• Standard models in finance (e.g., Long-Term Capital Management) and engineering often fail because they underestimate correlated failures and "normal accidents" in complex systems.
• Sandberg recommends using tools like "Guesstimate" to model ranges of uncertainty and avoid the false precision of point estimates.
• He draws a parallel between the Fermi analysis and pandemic risk, noting that while the probability of extinction is low in standard models, the risk of underestimating biological mechanisms is the primary danger.

Proposed Solutions to the Silence: Estivation and Berserkers

• The Estivation Hypothesis (10% probability): Advanced civilizations may be hibernating ("estivating") to maximize thermodynamic efficiency.
  • Computations require energy proportional to the surrounding temperature (Landauer principle); as the universe expands and cools, computing efficiency increases exponentially.
  • Waiting until the "heat death" (approx. $10^{30}$ years) could allow a civilization to perform $10^{30}$ times more computations with the same initial resources than doing so today.
  • A single planet's resources, if preserved and used in the far future, equal the computational potential of the entire observable universe used today.
  • This strategy requires civilizations to coordinate heavily to prevent resource depletion or invasion while in a dormant state.
  • It implies that active mega-engineering (moving stars/galaxies) should be suppressed to preserve resources, potentially explaining the "Zoo Hypothesis" where aliens avoid contact to protect the cosmic commons.
• The Berserker Hypothesis (<1% probability): Self-replicating killer machines from a past civilization are destroying life.
  • Sandberg argues this is unstable; an ecosystem of self-replicating killers would eventually lead to a universe-wide war of attrition, which would be visibly observable, contradicting the current silence.
  • It requires a specific, rare set of conditions (e.g., radical negative utilitarianism) or accidental deployment, making it unlikely as a universal explanation.
  • A more plausible variant is that advanced civilizations deploy autonomous "police" systems to prevent other civilizations from "burning the cosmic commons," but these are likely defensive rather than genocidal.

Space Colonization Strategies

• Spamming the Universe: Sandberg and Armstrong propose a strategy where a civilization uses the resources of one small planetary body (e.g., Mercury) to launch probes to every reachable galaxy within a few hours of sunlight.
• Probes can travel between stars using low-velocity "hops" or high-velocity "jumps" (90% light speed); the primary constraint is dust collisions, which are rare in intergalactic space.
• Once a probe reaches a new galaxy, it can build infrastructure to launch a new generation of probes to every star in that galaxy, reaching the entire observable universe in just two generations from the origin.
• The expansion of the universe creates an event horizon; galaxies beyond a certain distance (approx. 2.5 gigaparsecs) can never be visited or contacted by humans.
• Colonization strategies must account for the fact that reaching the farthest reachable galaxies makes the origin point unreachable, necessitating the distribution of copies of civilization rather than a central hub.
• The strategy assumes probes can self-replicate using asteroid regolith, which is chemically homogeneous and abundant across the universe.

Broader Implications for Existential Risk and Future Planning

• The sheer scale of the long-term future (trillions of years of potential computation) provides a moral imperative to reduce existential risks and avoid short-sighted actions that limit humanity's future potential.
• Centralized control of technology is historically ineffective (e.g., failed bans on printing or travel), suggesting that "differential technological development"—accelerating safety technologies alongside risky ones—is a superior approach.
• Self-replicating technologies (biological, digital, or memetic) require extra caution because small errors can lead to uncontrollable systemic risks.
• Sandberg argues that while we cannot stop discovery (e.g., nuclear physics), we can focus on developing "undo" technologies (e.g., containment for gene drives) before releasing powerful technologies.
• Future human civilization will face coordination problems as groups disperse beyond causal contact (light-speed communication limits), requiring robust "ground rules" established while humanity is still on a single planet.
• The interview concludes that understanding the vastness of the future and the physics of computation helps identify the most critical actions to take now to ensure a viable long-term future.