AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Uncanny Valley of Humor
GPT for in its current iteration doesn't have access to raw pixels, so sometimes they don't understand that certain visual elements are actually more important than they are. So there is some modality bridging issue that I think is going to be tricky to solve along with some of the other humor, quote unquote, understanding gaps. It's not immediately clear to me exactly what you can learn about human use of language using just a large language model. That being said, I'm cautiously optimistic about them in terms of their capacity to underlie useful tools by trained scientists.