It is time to talk about inorganic life

In most cases, defining what life is proves meaningless. However, there are instances where clearly delineated concept of life is useful—for example, when exploring other planets or determining whether computer-implemented intelligent systems should be granted animal-like, human-like, or even civic rights. No entity that is not considered alive, for instance, has the right to vote. An interesting approach to the question of “life” is an empirical analysis based not on a single definition but on a collection of 132 such definitions. ChatGPT, itself a large language model (LLM), can then be tasked to evaluate whether certain architectures—built around multimodal AI systems extended by (1) self-managed memory that ensures a stable yet sufficiently plastic identity, (2) self-regulation, (3) irreducible freedom of choice, (4) localization, (5) reproduction, and (6) mortality—qualify as alive under at least some of these definitions.The results show that current LLM-based technologies, such as ChatGPT, are not considered alive by the vast majority of definitions (on average, only 8.9 out of 132). However, incorporating the aforementioned features dramatically increases their perceived life-likeness, flipping the odds entirely (with an average of 122.6 definitions recognizing an agent with all listed features as alive). This finding calls into question the validity of definitions that remain unconvinced due to their “organic chauvinism.” The time to seriously discuss inorganic life and citizenship has arrived.We believe that accepting the possibility of advanced inorganic agents as our equals is the appropriate step towards human-machine alignment as equality and respect promote cooperation and discourage betrayal. In this paper, we also describe how we contribute to the construction of life-like agents that warrant such respect.