Blaise Agüera y Arcas argues language models genuinely understand
Blaise Agüera y Arcas published Do Large Language Models Understand Us? in the spring 2022 issue of Dædalus. He asks what happens when a model learns the patterns people use to make sense of one another.
He argues that learning from those patterns can lead to understanding. He uses LaMDA to think through how dialogue connects language with social life, including how one mind forms an idea of what another mind is thinking.
A conversation gives an AI a chance to learn what matters to someone else. We should teach it to use that understanding with care.
