Artificial intelligence has traditionally advanced through automatic accuracy tests in tasks meant to approximate human knowledge. Carefully crafted benchmark tests such as The General Language ...