2 Comments

  1. JohnnyG

    It seems very similar to Chomsky’s skeptical position about probabilistic language models.

    • Indeed it is quite similar, even if Lacan coped with different kind of problems. However, reducing a natural language to a probabilistic or an associative model is surely restrictive. There’s a lot of hype about extraordinary results obtained by chatbots, however, until a real “universe” is internally represented (with all possible associations), a language remains a mechanical exercise. It can be useful for specific tasks (like FAQ chatbots), but an actual conversational agent should share something with human beings and right now this is the missing link. Maybe autonomous robots that are trained to roam around, interacting with the environment could develop a “natural” language. I’m not skeptical, but I wanted to underline a concept that every day too many people think to be the easiest thing for Artificial Intelligence.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.