AI researchers question whether language models will abandon English for invented languages to solve complex problems more efficiently.

The prevailing assumption that RL-trained LLMs will invent new languages to bypass English limitations may be unfounded.

Historical precedent suggests humans rarely create new languages specifically to solve non-linguistic problems, even when theoretically more efficient.

Lojban, designed as a less ambiguous alternative to human language, failed to drive scientific breakthroughs despite its intended advantages.

The author argues that human creativity remains effective at problem-solving within existing language frameworks, casting doubt on the inevitability of language model departures from human-intelligible reasoning.