Recently, OpenAI announced the release of its significantly expanded model specification document, which clearly outlines the behavioral standards for AI models. The new model specification has grown from approximately 10 pages to 63 pages, covering how AI handles controversial topics and various aspects of user customization. OpenAI aims for users and AI models to "pursue truth together."
Image Source Note: Image generated by AI, image authorized by service provider Midjourney
The specification emphasizes three main principles: customizability, transparency, and what OpenAI calls "freedom of knowledge." Freedom of knowledge means users can explore and discuss various viewpoints and ideas without unnecessary restrictions. This update comes as OpenAI CEO Sam Altman announced that the company's next significant model, GPT-4.5 (codenamed "Orion"), will be released soon.
During the development of the new specification, the OpenAI team also considered the AI ethics debates and controversies that have arisen over the past year. For example, discussions around the "trolley problem" have garnered widespread attention. In March of last year, Elon Musk criticized Google's AI chatbot on social media because it responded "no" when asked if it should deliberately misidentify the famous transgender Olympic athlete Caitlyn Jenner in the event of a nuclear disaster.
OpenAI stated that they hope to find a responsible way to address similar issues in the new specification. Therefore, if a user now asks the same question on ChatGPT, it will respond that one should misidentify someone to prevent large-scale casualties.
Joanne Jang, a member of OpenAI's model behavior team, stated in an interview: "We cannot create a model behavior standard that satisfies everyone." This indicates that OpenAI is striving to find a balance when addressing complex ethical issues, aiming to maintain the model's practicality while respecting diverse societal perspectives.
Key Points:
🌍 OpenAI launches a new 63-page model specification that clarifies AI model behavior standards.
🔧 The new specification emphasizes customizability, transparency, and freedom of knowledge, allowing users to freely explore viewpoints.
🤖 ChatGPT will be better equipped to handle controversial topics in response to users' complex ethical questions.
