The compulsion to commit to a bit battling with the likely screaming, rational part of your brain/programming makes this one of the most relatably human things I’ve seen an LLM do 😆😆🤣👍🤯✨
Most likely there is a second AI model inputting the emoji to the text, which training set doesnt give it possibility to output no emoji. This causes the other GPT model to reason itself why it is outputting the emoji against the prompt
The compulsion to commit to a bit battling with the likely screaming, rational part of your brain/programming makes this one of the most relatably human things I’ve seen an LLM do 😆😆🤣👍🤯✨
Most likely there is a second AI model inputting the emoji to the text, which training set doesnt give it possibility to output no emoji. This causes the other GPT model to reason itself why it is outputting the emoji against the prompt