GPT vs. StarCraft II – 2 – Creating the Prompt
This is the second in a series of webcasts covering the development and experimentation of using GPT algorithms to control the high-level strategy of a StarCraft II bot.
I’ll be starting with a new Python project, adding the relevant libraries for LangChain, OpenAI and the Burney SC2 bot framework. I’ll then implement the code to use LangChain to call GPT models hosted in the Azure Open AI service to generate sstrtegy recommendations based on the current game state. I’ll compare the results from GPT3.5 Turbo and GPT 4, and also explore the use of prompt engineering to improve the strategies generated and make the prompt more machine readable.
The other webcasts in this series are available here:
GPT vs. StarCraft II – 1 – Introduction
• GPT vs. StarCraft II – 1 – Introduction
GPT vs. StarCraft II – 2 – Creating the Prompt
• GPT vs. StarCraft II – 2 – Creating t...
GPT vs. StarCraft II – 3 – Testing the GPT Bot
• GPT vs. StarCraft II – 3 – Testing th...
GPT vs. StarCraft II – 4 – Prompt Engineering & Monitoring
• GPT vs. StarCraft II – 4 – Prompt Eng...
GPT vs. StarCraft II – 5 – Collecting Vespene and Training Advanced Units • GPT vs. StarCraft II – 4 – Prompt Eng...
