🏠 Take-home task
AI Temperature Experiment (Playground Required)
This experiment needs a platform that exposes the temperature slider. Use the OpenAI Playground (platform.openai.com/playground) or Poe (poe.com) with a bot that shows a temperature control. Consumer chatbots like ChatGPT, Claude, Copilot and Gemini do NOT expose temperature to end users, so this task will not work in those apps. First, set the temperature to 0.1 and ask: "Complete this sentence: The best pizza topping is..." three times. Write down the answers. Then set the temperature to 1.5 and repeat the question three times, writing those answers too. Compare the low-temperature answers for similarity, and the high-temperature answers for creativity. Reflect on how temperature shifts the balance between safe and surprising outputs.
AI useYou need a temperature-adjustable AI platform: OpenAI Playground (platform.openai.com/playground) or Poe (poe.com). ChatGPT, Claude, Copilot and Gemini do NOT expose temperature and will not work for this task.
If no AI or internetNo Playground access? Do the paper prediction version. Write down what you PREDICT a low-temperature AI would answer three times to "The best pizza topping is..." (all safe, similar answers). Then write what you PREDICT a high-temperature AI would answer three times (much weirder, more creative). Explain in two sentences why the two sets should look so different.
Bring backYou will bring back either your six real answers with reflection, or the paper predictions with explanation.
Stay safeIf you have any questions or concerns about this task, please talk to a trusted adult.