OpenAI introduces five-level scale to rank progress towards 'human-level problem-solving'
OpenAI has unveiled a five-level scale to monitor its advancement in creating artificial intelligence (AI) software that can surpass human capabilities. The organization, renowned for its ChatGPT maker and its efforts to develop more advanced AI systems, presented this new classification during an internal meeting with employees.
OpenAI's ultimate goal is to develop artificial general intelligence (AGI), machines that can outperform humans in most tasks. CEO Sam Altman has expressed his belief that AGI can be achieved within this decade.
Advancements in AI
During the meeting, OpenAI executives showcased a research project involving the GPT-4 AI model, demonstrating new skills that demonstrate human-like reasoning capabilities. The organization constantly tests new capabilities internally as part of industry best practices.
Levels of AI Development
Researchers at Google DeepMind proposed a similar framework in a November 2023 paper, outlining five levels of AI progression, including tiers such as “expert” and “superhuman”. OpenAI's classification system aligns with this approach, with the third tier labeled as “Agents” - AI systems capable of autonomous decision-making over several days on behalf of a user.
Level 4 describes AI that can generate new innovations, while the highest level, Level 5, refers to “Organizations”. These levels are subject to refinement based on feedback from stakeholders, including employees and investors.
This classification framework is a collaborative effort by OpenAI executives and senior leaders, aiming to provide a clearer understanding of the organization's AI development progress and future goals.
Source: The Edge
