Artificial General Intelligence (AGI) is one of the ultimate goals of artificial intelligence research, aimed at developing machine intelligence capable of handling various tasks like humans. GPT-4 has been regarded as a spark for AGI. In this article, we will discuss GPT-4's early experiments and its contributions to AGI. The article is suitable for practitioners, tech enthusiasts, and researchers interested in machine learning, natural language processing, and AI technology. GPT-4 is the next-generation natural language processing model in the OpenAI plan, expected to be released within the next few years. It will utilize more data and advanced algorithms, further enhancing natural language processing capabilities, covering broader application areas, such as speech recognition, automated summarization, machine translation, etc. This article will cover GPT-4's early experiments, including natural language generation tests and machine learning tasks.