Tuesday, October 8, 2019

AI and Undertanding Language


In this episode, Byron talks about natural language processing. For more on Artificial Intelligence: https://voicesinai.com https://gigaom.com https://byronreese.com https://ift.tt/31WftGA... Transcript: Natural language processing is a really hard problem for AI. Languages were clearly not designed by engineers. They’re mushy, organic, and constantly changing. Just think of the problems. First, around the world, there are tons of different languages. Then, even within one language, words often have multiple meanings. There is also how we use the language, recognizing things like sarcasm. And most of the time, no two words mean exactly the same thing, and yet the differences between them are subtle and nuanced. With speech, even inflection matters. Take the sentence, “I didn't say you stole his money.” Whatever word you emphasize, that changes the sentence. “I-I didn't say,” “I didn't sa-ay...” All of this together means that it's going to be a long time before you can have a conversation with an artificial intelligence in a way that you do with a human.

No comments: