The Wolfram plugin also has extremely impressive examples [1].
If I were OpenAI, I would use the usage data to further train the model. They can probably use ChatGPT itself to determine when an answer it produced pleased the user. Then they can use that to train the next model.
If I were OpenAI, I would use the usage data to further train the model. They can probably use ChatGPT itself to determine when an answer it produced pleased the user. Then they can use that to train the next model.
The internet is growing a brain.
1: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its...