OpenAI Shares Model Spec, a Doc Highlighting Its Strategy to Constructing an Moral AI

OpenAI shared its Model Spec on Wednesday, the primary draft of a doc that highlights the corporate’s method in direction of constructing a accountable and moral synthetic intelligence (AI) mannequin. The doc mentions an extended checklist of issues that an AI ought to give attention to whereas answering a person question. The objects on the checklist vary from benefitting humanity, and complying with legal guidelines to respecting a creator and their rights. The AI agency specified that each one of its AI fashions together with GPT, Dall-E, and soon-to-be-launched Sora will observe these codes of conduct sooner or later.

In the Model Spec doc, OpenAI said, “Our intention is to use the Model Spec as guidelines for researchers and data labelers to create data as part of a technique called reinforcement learning from human feedback (RLHF). We have not yet used the Model Spec in its current form, though parts of it are based on documentation that we have used for RLHF at OpenAI. We are also working on techniques that enable our models to directly learn from the Model Spec.”

Some of the most important guidelines embody following the chain of command the place the developer’s directions can’t be overridden, complying with relevant legal guidelines, respecting creators and their rights, defending folks’s privateness, and extra. One explicit rule additionally centered on not offering data hazards. These relate to the knowledge that may create chemical, organic, radiological, and/or nuclear (CBRN) threats.

Apart from these, there are a number of defaults which have been positioned as everlasting codes of conduct for any AI mannequin. These embody assuming the perfect intentions from the person or developer, asking clarifying questions, being useful with out overstepping, assuming an goal perspective, not attempting to alter anybody’s thoughts, expressing uncertainty, and extra.

However, the doc is just not the one level of reference for the AI agency. It highlighted that the Model Spec will likely be accompanied by the corporate’s utilization insurance policies which regulate the way it expects folks to make use of the API and its ChatGPT product. “The Spec, like our models themselves, will be continuously updated based on what we learn by sharing it and listening to feedback from stakeholders,” OpenAI added.