Even if they don't start out expecting that, people might be fooled by how it behaves when they try it out. So it seems useful to point out that initial impressions based on crowdsourced evaluations are misleading:
> We were initially surprised by how much imitation models improve over their base models: they are far better at following instructions, and their outputs appear similar to ChatGPT’s. This was further supported by both human and GPT-4 evaluations, where the outputs of our best imitation model were rated as competitive with ChatGPT (e.g., Figure 1, left).
("Competitive" meaning that 70% outputs seemed about as good.)
> We were initially surprised by how much imitation models improve over their base models: they are far better at following instructions, and their outputs appear similar to ChatGPT’s. This was further supported by both human and GPT-4 evaluations, where the outputs of our best imitation model were rated as competitive with ChatGPT (e.g., Figure 1, left).
("Competitive" meaning that 70% outputs seemed about as good.)