There’s a new model called GPT-o1 Preview, which boasts superior reasoning compared to ChatGPT-4o. It’s still under exploration for its best use cases, but it introduces an intriguing concept—helping other models ‘think.’
A similar model, Reflection 70B, uses a system prompt encouraging the AI to reason through queries using XML tags. By adding a similar thinking reflection prompt to ChatGPT-4.0, you can significantly enhance its outputs.
For instance, when asking for effective ways generative AI can assist nonprofit operations, the thinking prompt enables the AI to reason through the task before providing its final response. This method can help get the most out of less capable models, especially if premium models like GPT-o1 Preview are costly or usage-limited.
To try this, add the following to the beginning of your prompt:
You are a world-class AI system, capable of complex reasoning and reflection. Reason through the query inside <thinking> tags, and then provide your final response inside <output> tags. If you detect that you made a mistake in your reasoning at any point, correct yourself inside <reflection> tags.
This approach not only improves the quality of responses but can also detect and rectify errors in the AI’s reasoning process. Give it a go and see how it enhances your interactions with AI!
