(논문 요약) Towards General Computer Control: A Multimodal Agent for Red Dead Redemption II as a Case Study (Paper)

핵심 내용

• General Computer Control: 사람이 컴퓨터를 조작하는 task 를 AI 가 수행.
• method: GPT4V 로 observe & plan.
  • Episodic memory: summaries of textual and visual information.
  • Procedural memory: retrieved skills in function forms. Upon skill curation, skills can be added, updated, or composed.
• skill 은 function 으로 생성하여 저장해둠.
  

실행 예시