An artificial intelligence system capable of constructing entirely new languages has ignited a debate on machine creativity and computational linguistics. Unveiled at last week’s Annual Meeting of the Association for Computational Linguistics, the tool known as ConlangCrafter represents a significant advancement in large language model applications. Developed by a research team led by computational linguist Gašper Beguš at the University of California, Berkeley, the system demonstrates that AI can autonomously generate structured, grammatical language systems ranging from human-like dialects to speculative constructs for fictional species. ConlangCrafter operates through a self-correcting iterative process. It begins by drafting a linguistic rulebook, defining phonetics, syntax, and grammar. A random-number generator then introduces atypical linguistic features drawn from global languages, steering the model away from conventional patterns. The AI subsequently generates vocabulary, translates English into the new framework, and validates its own outputs against its established rules, revising the grammar as needed. This methodology has already produced dozens of languages, including a specialized system for alien cephalopods that rely on color and tactile communication rather than vocalization. The system’s capabilities have sparked considerable discussion regarding the nature of creativity in artificial intelligence. Theoretical linguist Joseph Windsor of the University of Calgary, while acknowledging that ConlangCrafter’s outputs resemble authentic linguistic structures, notes that the languages currently lack the consistency required for extended use. He characterizes the AI’s method as a combinatorial selection process, comparing it to rolling dice rather than exhibiting genuine originality. Conversely, computational scientist Ganesh Bagler argues that human creativity is fundamentally combinatorial as well, citing culinary and musical innovation as evidence. He contends that restricting AI outputs to human-centric aesthetic preferences may unnecessarily limit the technology’s potential. Beyond the philosophical debate, researchers view ConlangCrafter as a valuable instrument for both linguistics and AI development. Beguš emphasizes that the tool serves as a window into how large language models reason and construct logic, independent of consciousness. The team plans to deploy the AI as a simulation engine for language evolution, testing whether novel meanings and conventions emerge when multiple AI systems interact over time. Morris Alper, a machine learning researcher involved in the project, positions ConlangCrafter as a creativity aid rather than a replacement for human linguists. He suggests that AI could generate candidate vocabularies for conlangers to curate, accelerating the traditionally laborious creative process. Despite its promise, experts identify clear limitations. Linguist Balthasar Bickel and linguistic anthropologist Christine Schreyer point out that ConlangCrafter does not account for semantic drift and cultural adaptation, which are essential to how natural languages evolve through usage. Historical examples, such as the organic expansion of vocabulary within the fictional Na’vi language, demonstrate that meaning develops dynamically through community interaction. Acknowledging these constraints, the development team maintains that the system’s primary value lies in augmenting human creativity and probing the boundaries of machine reasoning. As computational linguistics advances, ConlangCrafter establishes a new benchmark for synthetic language generation, reinforcing the growing consensus that AI may not replicate human imagination, but can certainly expand its frontiers.