Have you ever wondered if chatting with an AI assistant would be more natural if you pretended it was a specific person, like a teacher, friend, or customer service agent? That is the idea behind role-based prompting. In this video, we will put it to the test on the hottest AI chatbots out there.
We start with ChatGPT and ask it to give us instructions on making a kite. The response is divided into materials and instructions, making it easier to read. Then, we introduce a role and ask ChatGPT to act as a preschool teacher. The response changes to simpler language aimed at children with little to no experience.
Next, we try the same task with Claude.ai. The response without a user role is targeted towards a normal person, while the response with a preschool teacher role focuses on simple, child-friendly instructions. We also test Google Bard and Llama 2 by Meta, but the responses do not change much based on the user role.
In conclusion, assigning a role to AI chatbots improves the quality of the response in some cases, but not all. It seems that role-based prompting is more effective in chatGPT, Claude AI, and Llama 2 by Meta. Google Bard does not show significant changes based on user roles.
Subscribe to our channel for more AI-related videos!
