[Paper Review] An Anthropic Argument against the Future Existence of Superintelligent Artificial Intelligence

Toby Pereira|arXiv (Cornell University)|May 8, 2017

Space Science and Extraterrestrial Life7 references3 citations

TL;DR

This paper argues that superintelligent AI is unlikely to exist in the future by applying a refined anthropic principle—Super-Strong Self-Sampling Assumption (SSSSA)—which weights observer-moments by cognitive capacity. It contends that if humans are to be typical observers, superintelligent AI must not dominate, thereby reducing the probability of its future emergence.

ABSTRACT

This paper uses anthropic reasoning to argue for a reduced likelihood that superintelligent AI will come into existence in the future. To make this argument, a new principle is introduced: the Super-Strong Self-Sampling Assumption (SSSSA), building on the Self-Sampling Assumption (SSA) and the Strong Self-Sampling Assumption (SSSA). SSA uses as its sample the relevant observers, whereas SSSA goes further by using observer-moments. SSSSA goes further still and weights each sample proportionally, according to the size of a mind in cognitive terms. SSSSA is required for human observer-samples to be typical, given by how much non-human animals outnumber humans. Given SSSSA, the assumption that humans experience typical observer-samples relies on a future where superintelligent AI does not dominate, which in turn reduces the likelihood of it being created at all.

Motivation & Objective

• To challenge the likelihood of future superintelligent AI emergence using anthropic reasoning.
• To address the problem that standard anthropic principles fail to account for cognitive capacity differences between observers.
• To introduce a new observer-sampling framework that weights observer-moments by cognitive size.
• To show that human typicality under this framework implies the absence of superintelligent AI dominance.
• To provide a theoretical basis for why superintelligent AI may not come into existence, despite its conceptual feasibility.

Proposed method

• Introduces the Super-Strong Self-Sampling Assumption (SSSSA), extending SSA and SSSA by weighting observer-moments by cognitive capacity.
• Applies SSSSA to model observer-samples as proportional to mind size, ensuring human minds are not underrepresented.
• Uses the assumption that humans are typical observer-moments to infer constraints on future AI development.
• Argues that if humans are typical, then superintelligent AI cannot dominate the observer-moment distribution.
• Employs anthropic reasoning to derive a probabilistic constraint on the future existence of superintelligent AI.
• Relies on the idea that non-human animals vastly outnumber humans, so human typicality requires non-dominance of AI.

Experimental results

Research questions

• RQ1 Under what conditions would human observer-moments be considered typical in a multiverse with superintelligent AI?
• RQ2 How does weighting observer-moments by cognitive capacity affect anthropic probability assessments?
• RQ3 What constraints does human typicality place on the future emergence of superintelligent AI?
• RQ4 Can anthropic reasoning reduce the likelihood of superintelligent AI coming into existence?
• RQ5 Why does the standard Self-Sampling Assumption fail to account for cognitive differences between observers?

Key findings

• The introduction of SSSSA allows for a more nuanced anthropic framework that accounts for differences in cognitive capacity between observers.
• Human observer-moments can only be considered typical if superintelligent AI does not dominate the observer-moment distribution.
• The assumption of human typicality under SSSSA implies a reduced probability of superintelligent AI emerging in the future.
• The model shows that non-human animals, due to their vast numbers, would otherwise dominate observer-moment counts unless cognitive weighting is applied.
• The paper concludes that anthropic reasoning under SSSSA leads to a strong argument against the future existence of superintelligent AI.
• The framework provides a theoretical basis for why superintelligent AI may be less likely to emerge than commonly assumed.