https://platform.openai.com/playground/p/WZA12ApwQb6AX4YjrecE12e0
Prompt is – “Step into the world of Hollywood magic! Pick your favorite Hollywood actor or actress and highlight three movies that, in your opinion, define their cinematic brilliance. Discuss the diverse roles they’ve portrayed, the impact of each performance, and share why these movies stand out in their filmography. Ready, set, Hollywood brilliance”
The way I designed it was to invite users to share their insights and opinions about their favorite Hollywood actor or actress. It encourages them to select three movies that they believe define the cinematic brilliance of the chosen actor. The prompt goes further by encouraging users to discuss the diverse roles the actor has portrayed, the impact of each performance, and to articulate why these movies stand out in the actor’s filmography. The use of phrases like “Step into the world of Hollywood magic” and “Ready, set, Hollywood brilliance” adds an energetic, enthusiastic and engaging tone to the prompt.
Temperature: 0.83
The temperature is set at 0.83, which is a moderate value. This strikes a balance between generating diverse and creative responses while maintaining coherence. It allows for a range of imaginative answers without veering too far from the context of discussing Hollywood brilliance. When I tried 0.7, the explanations got cut off and vague.
The temperature choice of 0.8 and above ensures that the responses are not overly conservative or too random, ensuring a conversational flow that is interesting and relevant to the topic.
Top p (top percentage of tokens to consider): 0.85
From my research online it seems that a slightly higher top p of 0.8 ensures a broader exploration of ideas within the context of the chosen Hollywood actor. It allows for more diverse perspectives on what makes certain movies the actor’s top three.
Max tokens: 213
Given the complexity of discussing three movies and providing detailed analyses, I kept a slightly higher maximum token limit of 200. It seems a higher token allows for comprehensive and engaging responses. However, the length of the conversations were still being cut off. Maybe it needs more tokens?