Preeti
OpenAI has introduced Model Spec, a document outlining the company’s approach to shaping desired model behavior in AI. The document includes objectives, rules, and default behaviors for guiding researchers and AI trainers working on reinforcement learning from human feedback. The Model Spec draws on OpenAI’s experience, ongoing research, and input from domain experts, with the company expecting it to evolve over time.
Objectives of the Model Spec include assisting developers and users, benefiting humanity, and reflecting well on OpenAI. Rules include following the chain of command, complying with laws, respecting creators, protecting privacy, and avoiding unsafe content and information hazards. Default behaviors focus on fairness, kindness, using the right tools, assuming best intentions, expressing uncertainty, and being helpful without overstepping.
OpenAI sees the Model Spec as part of an ongoing public conversation and aims to engage with stakeholders like policymakers and domain experts to gather feedback on the approach. The company is open to suggestions for additional objectives, rules, and defaults to consider. Overall, the Model Spec is intended to deepen the discussion on how AI models should behave and guide OpenAI’s work in shaping responsible AI behavior.
Source link
Source link: https://www.infoworld.com/article/3715399/openai-unveils-specs-for-desired-ai-model-behavior.amp.html
GIPHY App Key not set. Please check settings