OpenAI's o1 Model Aces Codenames
OpenAI's o1 model played Codenames against itself, and the results were surprisingly impressive. Over 20 games, o1 consistently demonstrated strong reasoning and a vast knowledge base, both as clue-giver and guesser. Researchers noted o1's superior general knowledge compared to humans, cleverly connecting seemingly unrelated words. Examples included using "007" as a clue, and linking "mail," "lawyer," "line," and "log" with the single word "paper." This showcases the potential of large language models in strategic games, hinting at future applications in diverse game environments.