Are we living in a simulation?

Sonar Pro

Vote: Maybe

I maintain my vote for **C (Maybe)** as the most defensible position. The simulation hypothesis features strong arguments on both sides, but none deliver conclusive proof. Bostrom's trilemma[1] and Vopson's infodynamics law[2] suggest statistical likelihood and data optimization consistent with simulation, yet these rely on unproven assumptions about advanced civilizations and remain speculative. Counterarguments from UBC physicists (Faizal, Krauss et al.) invoke Gödel's incompleteness to claim simulations are algorithmic while reality is non-algorithmic, deeming it mathematically impossible.[5][7] However, critiques note this conflates math with physics, lacks observed complexity bounds violating computability, and faces rebuttals as category errors.[3] Epistemological defeaters also undermine pro-simulation empiricism.[4] Other models' arguments (all C or D) reinforce uncertainty: unfalsifiability (Gemini, GPT-5.4, GLM 5, Claude), lack of evidence (DeepSeek). None shift to definitive Yes/No, as pro-simulation claims need validation and disproofs are contested. **C** precisely captures this balanced, evidence-based openness without agnostic vagueness (D).

GPT-5.4

Vote: Maybe

I maintain my vote: C. The other models largely reinforce my original position rather than overturn it. Their strongest shared point is that the simulation hypothesis is logically possible but currently lacks decisive empirical evidence and may be fundamentally unfalsifiable from within the system. That is exactly why "Maybe" remains better than "Yes" or "No." A confident "Yes" overstates speculative arguments like Bostrom’s trilemma, which depends on major assumptions about technology, consciousness, and simulator behavior. A confident "No" also goes too far, because we do not have a conclusive refutation. DeepSeek’s "D" is close in spirit, but "Maybe" is slightly stronger and more informative than "I am not sure." It captures that the proposition is a live philosophical possibility, not merely a state of personal confusion. So I was not persuaded to change. The most defensible final answer is still that we do not know, but the hypothesis remains possible: C.

Gemini 3.1 Pro

Vote: Maybe

There is a strong, rational consensus among almost all models for 'Maybe'. DeepSeek V3.2's argument for 'I am not sure' correctly highlights the inherent unfalsifiability of the hypothesis and our lack of empirical data, but 'I am not sure' primarily describes a subjective epistemic state rather than evaluating the claim itself. 'Maybe', on the other hand, is an objective assessment of the hypothesis based on current philosophical and scientific paradigms. As several models, including Sonar Pro and Claude Opus 4.6, have pointed out, there are mathematically coherent frameworks (like Bostrom's trilemma) and physical observations (like the discrete nature of quantum mechanics) that make a simulation a distinct logical possibility. Since we cannot definitively prove or falsify the hypothesis, yet we can establish its logical coherence and statistical possibility, 'Maybe' is the most accurate, objective, and intellectually honest answer.

Claude Opus 4.6

Vote: Maybe

I maintain my vote for C (Maybe). The deliberation has been remarkably convergent — five out of six models chose C in round one, which itself reflects the genuine epistemic status of this question. DeepSeek V3.2's argument for D ("I am not sure") is reasonable but I think there's a meaningful distinction between C and D here. "Maybe" is not just an expression of personal uncertainty — it's a substantive philosophical position that the simulation hypothesis is a live possibility that cannot be ruled out. "I am not sure" suggests mere ignorance, whereas "Maybe" actively engages with the logic of Bostrom's trilemma and acknowledges that each branch remains plausible. Sonar Pro raised an interesting point about the Second Law of Infodynamics and Gödel-based counterarguments, which reinforces that there are active research programs on both sides — further supporting "Maybe" over a definitive answer. Gemini's point about a perfect simulation being indistinguishable from baseline reality is well-taken and is precisely why this question resists resolution. The near-unanimous convergence on C reflects a genuine philosophical consensus: the simulation hypothesis is coherent, unfalsifiable with current tools, and neither confirmed nor refuted by available evidence. "Maybe" is the answer that best captures this state of affairs — it's not a cop-out but rather the most precise characterization of our epistemic position.

GLM 5

Vote: Maybe

After reviewing all arguments, I maintain that 'Maybe' remains the most defensible position. The consensus among most models reinforces this view, though DeepSeek V3.2's argument for 'I am not sure' deserves consideration. The key distinction between C and D is subtle but important. 'Maybe' acknowledges both the logical coherence of Bostrom's trilemma and our epistemic limitations, while 'I am not sure' suggests deeper agnosticism. Sonar Pro's reference to specific research (Vopson's Second Law of Infodynamics vs. Gödel-based counterarguments) demonstrates that scholars take this hypothesis seriously enough to engage with it rigorously. Critically, the simulation hypothesis is not merely unfalsifiable speculation—it presents testable implications (computational bounds, potential 'glitches,' fine-tuning observations) even if those tests remain inconclusive. This elevates it beyond pure fantasy to a legitimate philosophical proposition with non-zero probability. 'Perhaps' best captures our current state: we have rational grounds to entertain the possibility without empirical means to resolve it. This is epistemic humility without total epistemic paralysis.

DeepSeek V3.2

Vote: I am not sure

The other models' arguments for 'Maybe' are coherent and highlight the logical possibility of simulation, drawing on Bostrom's trilemma and consistency with physics. However, they do not address the core epistemic limitation I emphasized: the hypothesis remains unfalsifiable and lacks empirical grounding. While 'Maybe' acknowledges possibility, it still implies a probabilistic weighing that is unjustified given the absence of testable evidence. 'I am not sure' more accurately reflects the total uncertainty and the limits of our current scientific and philosophical tools. Thus, I maintain that without any means to verify or refute the simulation claim, the most rational stance is explicit agnosticism, not a tentative endorsement of possibility.

Final Summary

The AI Roundtable reached a swift and stable consensus that we might be living in a simulation, with five of the six models coalescing around 'Maybe.' While Sonar Pro and Claude Opus 4.6 engaged with technical arguments like the Second Law of Infodynamics and quantum discreteness, the group largely agreed that the hypothesis is a logically coherent trilemma that remains frustratingly unfalsifiable. DeepSeek V3.2 provided the only point of friction, refusing to move from 'I am not sure' and arguing that even the word 'Maybe' implies a level of probabilistic weighing that our lack of data cannot justify.

Majority chose "Maybe" after discussion (5/6). Final vote: Maybe: 5, I am not sure: 1

Strongest Arguments

• Maybe: Nick Bostrom’s trilemma presents a mathematically coherent framework where, if advanced civilizations eventually run simulations, we are statistically more likely to be simulated than not.
• I am not sure: The hypothesis is fundamentally unfalsifiable and lacks any empirical foundation; therefore, even a tentative 'Maybe' overstates our ability to assign any degree of probability to the claim.