Cyberpunk Church

Questions about intelligence, civilization, and reality.

Continued Coherence and the Snapshot Problem

Imagine an observer instantiated for one instant with your present perceptions and apparent memories.

That state could be produced without producing the life it remembers. A simulator could construct only the final frame. A fluctuation could assemble a mind with a false past. If anthropic reasoning counts compatible observer-moments, the observer may qualify at far lower cost than an entire coherent life.

A snapshot observer can match present evidence without possessing the history that evidence appears to record.

This is not merely a strange possibility. It is a test of our measure.

Tomorrow is new evidence

At this moment, a sufficiently exact snapshot and a persistent observer can be observationally identical. The models differ in what happens next.

Let E<=t be everything observed through today, and let E(t+1) be another ordinary interval in which memories remain connected, records agree, and the same physical regularities continue.

For a cheap snapshot model:

P(E(t+1) | E<=t, snapshot)
    <<
P(E(t+1) | E<=t, persistent history)

Each additional period of ordinary, connected experience is evidence against models that cheaply produce only the present observer-state.

This update accumulates. A hypothesis optimized to explain one frame must repeatedly pay for unexpected continuation. A persistent process predicts continuation from the start.

That is the precise value of continued coherence. It does not prove that our apparent history is real. It discriminates between models that predict a continuing history and models that predict little beyond the current state.

What coherence does not tell us

Now compare base reality with a faithful, persistent simulation. If both implement the same continuing dynamics, then normal physics tomorrow may be about equally likely under either:

P(E(t+1) | E<=t, persistent simulation)
    approximately equals
P(E(t+1) | E<=t, base world)

Continued coherence can strongly disfavor snapshots while providing almost no evidence between base reality and a faithful persistent simulation.

This distinction matters because “simulated” and “momentary” are often allowed to blur together. They are independent properties. A simulated life might be as causally stable as a base life. A base cosmology might generate short-lived accidental observers.

The Boltzmann-brain stress test

Boltzmann brains sharpen the same problem. In some cosmological models, random fluctuations can produce isolated observers with apparently ordinary memories. If those observers dominate the relevant measure, the theory predicts that our orderly observations are profoundly atypical. That is generally treated as a pathology of the cosmological model or its measure, not as an attractive explanation of our evidence.1

Any proposed observer measure must explain why coherent histories are not overwhelmed by cheap, false-history observer-states.

Counting histories instead of moments may help, but it does not solve the problem automatically. We still need a rule that determines how duration, likelihood, implementation, and causal continuity contribute to mu(H).

Observer-history measure is a framework for stating the problem, not a finished answer.

We have nevertheless learned something observable: another coherent day moves probability away from hypotheses that expected only a convincing present. It need not move probability away from all simulations.

That leaves a different question. What exactly makes one world a simulation of another?

  1. Sean M. Carroll, “Why Boltzmann Brains Are Bad” (2017), develops the case that a cosmology dominated by observers arising from fluctuations fails as a successful account of our orderly data. ↩

Filtered post view · schema v1

Argument map

Loading the claims, evidence, objections, and assumptions behind this post…

  • Supports / entails
  • Attacks / contradicts
  • Structural relation
  • Question
  • Claim
  • Hypothesis
  • Evidence
  • Objection
  • Assumption
  • Conclusion

Drag to pan · scroll or pinch to zoom · select a node for details

Open full map View normalized JSON
tags: Anthropics - Measure