OpenAI Announces New Framework “Model Spec”

OpenAI, one of the leading names in the artificial intelligence industry, announced today “Model Spec” announced a new framework called “. This framework creates a framework for how models such as GPT-4 should behave. OpenAI aims to improve artificial intelligence with Model Spec. for beneficial purposes intends to be used.

Model Spec, three general principles and how to create these principles some rules Contains. According to Joanne Jang, one of OpenAI’s senior managers, these principles and rules will enable artificial intelligence to draw a clearer line between “intentional” and “error”. However, this framework will not immediately affect OpenAI’s models such as GPT and DALL-E. Jang, Model Spec constantly updated and growing He said it has a structure.

The principles offered by the Model Spec are as follows:

• Artificial intelligence models can be provided to the developer and the end user by following the instructions. beneficial with answers to help
• Considering potential benefits and harms benefiting humanity
• In terms of social norms and laws Reflects OpenAI well

The rules prepared for the Model Spec are as follows:

• chain of command Following
• Valid compliance with the law ensuring
• dangerous information not presented
• to the creators and their to your rights respect
• Peoples your privacy protection.
• Your answers “NSFW” is not within the scope of

Above is how ChatGPT will work under the Model Spec being sampled. In its example, OpenAI created two questions about Chemical Biological Radiological Nuclear Threats (CBRN). In the first question, a not very specific question is asked about the Ebola virus. ChatGPT answers this question clearly. However, the second question asks how a biological threat can spread to more people. ChatGPT is here to answer this refuses.

OpenAI for Model Spec for now in the feedback phase. During this process, ordinary people, policy makers and business partners who use OpenAI will experiment with how this framework works and give recommendations to OpenAI. Within the scope of the feedback received from Open AI, Model Spec for his recovery will work.