OpenAI introduces tier system to track progress towards AGI, showcasing innovative GPT-4 model

From Condé Nast: 2024-07-12 17:58:05

OpenAI has introduced a five-tier system to track progress toward developing artificial general intelligence (AGI), with levels ranging from chatbots to AI that can manage organizations. The company aims to convey internal milestones achieved on the path to AGI and recently showcased the potential of their GPT-4 model in an all-hands meeting.

The pursuit of AGI is OpenAI’s primary goal, with CEO Sam Altman believing it could be achieved within this decade. The new classification system aims to communicate AI advancements internally and potentially attract investors interested in the company’s technological progress.

OpenAI’s system, which includes levels like Chatbots and Reasoners, describes hypothetical future AI capabilities. A Level 2 system, for example, would have problem-solving abilities comparable to a human with a doctoral degree. The classification system is a work in progress, with plans to gather feedback for refinement.

Several AI research labs, including OpenAI and Google DeepMind, have introduced classification systems to track AI advancement. OpenAI’s system is similar to Anthropic’s AI Safety Levels, highlighting the industry’s effort to categorize and quantify AI capabilities despite the lack of consensus on measuring progress toward AGI.



Read more at Condé Nast: OpenAI reportedly nears breakthrough with “reasoning” AI, reveals progress framework