OpenAI has introduced a five-tier system to evaluate its progress toward developing artificial general intelligence (AGI). This new classification framework was shared with employees during an all-hands meeting, aiming to clarify AI advancements. However, this system describes hypothetical technology that doesn’t yet exist and may serve primarily as a marketing tool to attract investment.
AGI: The Ultimate Goal
AGI, an AI system capable of performing novel tasks like a human without specialized training, is OpenAI’s main objective. This pursuit has generated significant hype, despite its potential for disruptive societal impact. OpenAI CEO Sam Altman has expressed confidence that AGI could be achieved within this decade. The new classification system is part of OpenAI’s strategy to communicate its milestones on the path to AGI.
The Five Levels of AI
OpenAI’s five levels range from current AI capabilities to systems capable of managing entire organizations:
- Level 1: Chatbots – AI with conversational language abilities.
- Level 2: Reasoners – AI capable of human-level problem solving.
- Level 3: Agents – Systems that can take actions autonomously.
- Level 4: Innovators – AI that can aid in invention.
- Level 5: Organizations – AI that can perform the work of an entire organization.
Progress Toward Level 2
OpenAI believes its technology, like the GPT-4 model powering ChatGPT, is currently at Level 1. Executives have suggested they are on the brink of reaching Level 2, which involves basic problem-solving akin to a human with a doctorate but without external tools. During the all-hands meeting, a research project using GPT-4 was demonstrated, showing signs of approaching human-like reasoning ability.
Higher Levels and Hypothetical Capabilities
The upper levels of OpenAI’s classification describe increasingly advanced AI capabilities. Level 3 “Agents” would work autonomously on tasks for extended periods. Level 4 systems would generate novel innovations, and Level 5 envisions AI managing entire organizations. This system is still in development, with plans to gather feedback from employees, investors, and board members.
Industry Comparisons and Skepticism
OpenAI’s system is reminiscent of the levels of autonomous driving mapped out by automakers and similar to frameworks proposed by other AI labs, such as Google DeepMind and Anthropic. However, there is skepticism about the meaningfulness of quantifying AI progress, given the tech industry’s history of overpromising capabilities.
Conclusion
There is no consensus in the AI research community on measuring progress toward AGI. OpenAI’s five-tier system is best viewed as a communication tool to show the company’s aspirations rather than a scientific measurement of progress.