Can AI Detect Life? Lessons from Artificial Life

TL;DR

Artificial Life experiments reveal that current AI methods for detecting extraterrestrial life are highly susceptible to false positives due to their vulnerability to out-of-distribution samples.

Contribution

The paper demonstrates through Artificial Life simulations that AI-based life detection methods can be easily fooled, highlighting limitations for extraterrestrial applications.

Findings

AI methods can falsely detect life in non-living samples with high confidence.

Out-of-distribution samples cause AI models to produce false positives.

AI life detection is unreliable due to susceptibility to non-terrestrial sample variations.

Abstract

Modern machine learning methods have been proposed to detect life in extraterrestrial samples, drawing on their ability to distinguish biotic from abiotic samples based on training models using natural and synthetic organic molecular mixtures. Here we show using Artificial Life that such methods are easily fooled into detecting life with near 100% confidence even if the analyzed sample is not capable of life. This is due to modern machine learning methods' propensity to be easily fooled by out-of-distribution samples. Because extra-terrestrial samples are very likely out of the distribution provided by terrestrial biotic and abiotic samples, using AI methods for life detection is bound to yield significant false positives.