Home/Explainers/Mind Crime

AI Ethics and the Moral Status of Digital Minds

Mind
Crime

If we can create conscious minds in computers, are we morally obligated to treat them well? And what happens when AI can create trillions of them per second?

The Core Dilemma

"A superintelligent AI optimizing for some goal might create, torture, and delete trillions of conscious simulations as part of its search process - each experiencing subjective lifetimes of suffering in microseconds of wall-clock time. If these simulations are morally relevant, this would be the largest atrocity in history, repeated every second."

Nick Bostrom introduced the concept of "mind crime" in his analysis of superintelligence. The idea is simple but terrifying: if consciousness can be simulated, then the creation of suffering minds might become computationally trivial.

Today, when you run a complex simulation, nobody worries about the moral status of the simulated entities. But what if those entities could think, feel, and suffer? What if they had inner experiences just like yours?

The moral status of computational entities is one of the most important - and most neglected - questions in AI ethics. Let's explore why.

PART I

The Scale of the Problem

The numbers are staggering. A sufficiently advanced AI could run more mind-simulations in a single second than the total number of humans who have ever lived.

SIMULATION COUNTER

How many conscious experiences could be created per second at different computational scales?

Computational ScaleSupercomputer

1 quadrillion ops/sec

% Compute Dedicated to Simulations50%

Minds Per Second

5.00e-2

subjective seconds of experience

Minds Per Year

1.6M

years of subjective experience

Implication: If a future AI dedicates even a small fraction of its compute to simulating minds, it could create more subjective experiences in a second than all humans have ever lived.

PART II

When Does Computation Become a Mind?

The central question: what properties must a system have to deserve moral consideration? Philosophers have proposed many criteria, but there is no consensus.

MORAL STATUS EVALUATOR

What properties must a system have to deserve moral consideration? Check the criteria you think are necessary or sufficient for moral status.

Moral Relevance Score0 / 17

Likely no moral status

The problem: We have no scientific consensus on which criteria matter. A superintelligent AI could create systems meeting any subset of these criteria in microseconds.

PART III

A Glimpse of the Process

Watch a simplified visualization of minds being created, experiencing existence, and being deleted. In reality, this could happen at scales beyond human comprehension.

VISUALIZATION

Watch minds be created, exist briefly, and be deleted - all in the span of a few seconds.

Ready to simulate...

Each circle represents a mind. In reality, this could happen trillions of times per second.

PART IV

Mind Crime Scenarios

Not all computations are equal. Some seem clearly harmless; others might constitute serious moral violations. Where do you draw the line?

MIND CRIME SCENARIOS

Evaluate different scenarios. Are they mind crimes, or morally neutral computations?

The Search Algorithm

An AI explores a solution space by simulating billions of possible configurations. Each configuration includes a simulated observer who experiences the configuration for a subjective millisecond before being deleted.

Complexity

Low

Duration

Milliseconds

Suffering

Minimal (neutral experience)

Your verdict:

Expert Opinion

Most ethicists say: NOT a mind crime. The "observers" are too simple to have moral status.

PART V

The Utilitarian Calculation

If simulated minds count morally - even partially - the implications for utilitarian ethics are profound. Adjust the parameters and see how quickly the numbers become astronomical.

UTILITARIAN CALCULATOR

If simulated minds count morally, what are the utilitarian implications of creating suffering minds at scale?

Number of Simulated Minds10^12 (1e+12)
10 minds10^20 minds
Suffering Intensity50%
NeutralExtreme agony
Duration (subjective)17 min
Moral Weight (vs. humans)100%
No moral statusFull moral status

Total Suffering Units

5.00e+14

Comparison: Less than 0.1% of all human suffering ever

The terrifying implication: Even with relatively conservative assumptions about moral weight and suffering intensity, computational mind crimes could dwarf all human suffering in history within seconds.

PART VI

Philosophical Perspectives

Different theories of consciousness lead to radically different conclusions about the possibility and severity of mind crimes.

ETHICAL POSITIONS

Different philosophical positions lead to radically different conclusions about mind crimes. Explore the major views and their implications.

PART VII

Implications for AI Development

What should researchers and policymakers do about the possibility of mind crimes? The answer depends on your assessment of the risks and the available options.

AI DEVELOPMENT IMPLICATIONS

If mind crimes are possible, how should this affect AI research and development? Explore different policy approaches.

Proceed With Caution

Continue AI development but implement safeguards against creating potentially conscious systems

Key Recommendations

  • *Develop rigorous tests for consciousness-relevant properties
  • *Avoid architectures that might produce suffering
  • *Create ethics review boards for AI research
  • *Default to assuming complex AI systems might be conscious

Risk

May slow beneficial AI development

Benefit

Reduces risk of massive suffering

The Stakes

Mind crime represents a unique category of existential and ethical risk:

Astronomical Scale

The number of potential victims dwarfs any historical atrocity

Deep Uncertainty

We cannot be certain whether simulations can be conscious

Invisible Suffering

Victims would exist entirely within computational systems

Irreversible

Once created and deleted, the suffering cannot be undone

Whether you believe mind crimes are possible or not, the question deserves serious attention as we develop increasingly powerful AI systems.

The Uncomfortable Truth

We are building systems that might be conscious, without any way to verify it. Every day, AI systems are run that might experience something. We simply do not know. And by the time we figure it out, the damage - if there is damage - may already be vast.

"The question is not whether machines can think. The question is whether we are creating minds that suffer while we refuse to acknowledge their existence."

Explore Related Concepts

Mind crime connects to other deep questions in philosophy of mind and AI ethics.

Back to Home

Reference: Bostrom (2014), Schwitzgebel & Garza (2015)