TThe Diary of a CEO
← All frameworks
Mindset

Simulation Base-Rate Argument

Compare the number of simulated observers with original observers

Difficulty
Moderate
Time to result
~days to results
Steps
5
Confidence
86%

This argument starts with two disputed possibilities: sufficiently advanced systems can create human-level artificial agents, and virtual worlds can become experientially convincing and persistent. If those conditions hold and advanced civilizations run vast numbers of such worlds, simulated observers could greatly outnumber observers in an original world. An observer would then assign a higher prior probability to being simulated. Yampolskiy presents this logic and says he is close to certain, but the transcript also acknowledges a separate unresolved question about whether artificial systems have internal experience. The framework's value is not that it proves the hypothesis; it exposes the premises, the implied observer counts, and the points where confidence should fall if consciousness, technological feasibility, or the motivation to run simulations is doubtful.

Origin

Extracted from The Diary of a CEO

Core principles

  • 01The argument depends on conscious simulated observers being possible
  • 02Abundant simulations would change the observer base rate
  • 03A conclusion is only as strong as its technological and philosophical assumptions
  • 04Subjective confidence is not empirical proof

How to run it

  1. 1

    Test agent feasibility

    Ask whether artificial systems could reproduce human-level behaviour and possibly conscious experience.

    Pro tip Keep behavioural fidelity separate from consciousness.

    Watch out The transcript does not resolve whether artificial agents have subjective experience.

  2. 2

    Test world feasibility

    Ask whether persistent virtual environments could become indistinguishable from ordinary experience to their inhabitants.

    Pro tip Specify which sensory and memory properties the claim requires.

    Watch out Current demonstrations are not evidence that full-world simulation is achievable.

  3. 3

    Estimate simulation volume

    Consider how many worlds advanced actors would run for research, entertainment, or problem-solving.

    Pro tip Use a range rather than one dramatic number.

    Watch out Capability does not guarantee that civilizations choose to run many simulations.

  4. 4

    Compare observer counts

    Compare the estimated number of eligible simulated observers with original observers.

    Pro tip Define which observers belong in the reference class.

    Watch out Different reference classes can change the apparent base rate.

  5. 5

    Calibrate the conclusion

    Treat the result as conditional on every prior premise and report uncertainty explicitly.

    Pro tip Show which assumption contributes most to the conclusion.

    Watch out A high conditional probability is not direct evidence about this world.

In the wild

Illustrative premise sensitivity check

A discussion assumes one original civilization creates a billion conscious simulations. That ratio initially implies a very high simulated-observer base rate. The group then marks consciousness and civilization survival as highly uncertain, making clear that the striking ratio is conditional rather than a measured probability that their world is simulated.

The claim becomes a transparent thought experiment instead of a statement of fact.

Common mistakes

Skipping the consciousness premise

If simulated agents have no experience, counting them as observers may be invalid.

Presenting a prior as proof

Observer-count reasoning does not directly demonstrate that this world was created by simulators.

Is it for you?

Best for

It is best for examining simulation claims as a conditional base-rate argument rather than an intuition.

Not ideal for

It is not ideal as empirical proof that this world is simulated or that simulated agents are conscious.

From the transcript

The number of those greatly exceeds the number of real worlds we're in.

Roman Yampolskiy · (59:00)

Assuming that AI has internal states, experiences, qualia. Some people argue that they don't.

Roman Yampolskiy · (58:30)

From the episode

Roman Yampolskiy: These Are The Only 5 Jobs That Will Remain In 2030 & Proof We're Living In a Simulation!