[Sam Altman: Open source GPT-3, reduce the cost of GPT-4, double the model size] On June 1, Raza Habib, founder of the AI ​​development platform HumanLoop, held a closed-door discussion with more than 20 developers including OpenAI CEO Sam Altman. Sam Altman revealed a lot about OpenAI's future plans and current situations. Specifically: 1. OpenAI is currently severely limited by GPUs, causing them to postpone many short-term plans. Most of the problems with ChatGPT reliability and speed are caused by a shortage of GPU resources. 2. Reducing the cost and improving the efficiency of GPT-4 is OpenAI's current priority. 3. Longer ChatGPT context windows (up to 1 million tokens). In the future, there will be an API version that remembers conversation history. 4. GPT-4's multimodal capabilities will not be made public until 2024, and the visual version of GPT-4 cannot be expanded to everyone before more GPU resources are obtained. 5. OpenAI is considering open-sourcing GPT-3. Part of the reason they haven't done so yet is because they feel that not many people and companies have the ability to properly manage such a large language model. 6. Many recent articles claiming that "the era of giant AI models is over" is not correct. OpenAI's internal data shows that the law of scale proportional to performance still holds true. OpenAI's model size may double or triple every year (multiple sources indicate that the parameter size of GPT-4 is 1 trillion), rather than increasing by many orders of magnitude.