ItalyThe article discusses a study conducted by researchers at Michigan State University, Christoph Adami and Ankit Gupta, who tested the reliability of artificial intelligence in distinguishing between living and non-living entities. They used Avida, a digital environment where short computer programs can self-replicate, mutate, and evolve. The AI was trained on tens of thousands of sequences, some containing instructions for replication and others not. Initially, the AI correctly identified living-like sequences with 99.97% accuracy. However, when the researchers gradually modified non-replicating sequences to increase the likelihood of being classified as 'alive,' the AI began to label them as life with nearly 100% confidence. This highlights the limitations of AI in defining life based solely on data inputs. The article critiques the scientific ambiguity around defining life and questions whether AI can truly discern life without human-defined criteria.
Bias read (Center): The article presents a balanced discussion of scientific research without overtly favoring any political ideology. It focuses on the technical and philosophical challenges of defining life through AI, rather than taking a partisan stance.
