MIT Tech Review reveals courts ruling only humans can be inventors as AI invents drugs; ArXiv papers address AI context learning and consciousness ethics.
Published
A legal confrontation between artificial intelligence capability and human-centered patent law has come into sharp focus. Insilico Medicine credited its AI platform with discovering a pulmonary fibrosis drug, then named five humans—including its CEO—as inventors on the resulting patent. Courts have consistently ruled that only humans can legally be inventors, because US law defines an inventor as an "individual." The US Patent and Trademark Office has previously published guidance on the matter, though legal scholars say the framework remains unsettled.
New research published on arXiv tackles separate fundamental challenges in AI development. One paper (2608.19202) formulates context acquisition for interactive AI agents as an active inference problem, examining how agents should balance proceeding with default assumptions against spending computational resources on clarifying questions, retrieval calls, or tool calls when users omit constraints or preferences.
Another arXiv paper (2608.19214) reviews robust metaheuristics for the berth allocation and quay crane assignment problem in maritime logistics, addressing uncertainties in vessel arrivals, handling times, and equipment availability.
A third arXiv paper (2608.19215) examines the ethical dilemma of AI consciousness, arguing that assuming insentience risks causing terrible harms to entities that may deserve moral standing, while assuming sentience risks wasting resources on insentient machines.