I will arrive early and get us two cups of coffee.
I think that while this is a very interesting and funny answer (which is why I posted it), we should keep in mind that GPT-4 was trained on human conversations, so however good the RLHF was by OpenAI, sometimes it says something out of character for an AI assistant. It just tried to predict a likely path for the conversation.
I love it. I’d have asked it what colour tshirt it will be wearing. I haven’t interacted with gpt4 much yet but I’m guessing gpt3 would just give me the I’m AI speech.
It actually said that it would be there:
I think that while this is a very interesting and funny answer (which is why I posted it), we should keep in mind that GPT-4 was trained on human conversations, so however good the RLHF was by OpenAI, sometimes it says something out of character for an AI assistant. It just tried to predict a likely path for the conversation.
I love it. I’d have asked it what colour tshirt it will be wearing. I haven’t interacted with gpt4 much yet but I’m guessing gpt3 would just give me the I’m AI speech.
GPT-4 is excellent at giving you the “I’m an AI” speech:
It’s supremely annoying. That’s why this café conversation was so surprising.
Just as a comparison, here is GPT-3.5’s answer: