OpenAI has created a new system to measure how close they are to developing super AI, known as artificial general intelligence (AGI). AGI is a form of AI that can perform tasks better than humans can, potentially transforming various fields and industries.
New Five-Level Classification System
On July 9, OpenAI introduced a five-level classification system to its employees to help track their advancements in AGI. This system will be shared with investors and other interested parties in the future, providing greater transparency about their progress.
Current Level: Conversational AI
Currently, OpenAI’s technology is classified as Level 1. This level indicates that their AI can engage in conversations with people in a way that feels natural and fluid, laying the groundwork for more complex interactions.
Approaching Level 2: Problem Solvers
OpenAI is nearing Level 2, known as “Reasoners.” At this level, the AI is capable of solving problems with the same skill and insight as a person who holds a doctorate, suggesting a significant leap in cognitive abilities.
Future Levels: Agents and Innovators
- Level 3 (Agents): At this level, AI systems can autonomously manage tasks for several days without any human assistance, showing a higher degree of independence.
- Level 4: This level encompasses AI that can innovate and create new products or ideas, further showcasing its capabilities.
Top Level: AI Organizations
Level 5 (Organizations): The highest tier represents AI systems that can effectively operate and manage the tasks typically handled by an entire organization, indicating a major step toward full autonomy.
System Still Developing
OpenAI emphasizes that this classification system is a work in progress. It may evolve based on ongoing research and feedback from experts, ensuring it remains relevant and effective.
Predictions for AGI
- OpenAI Leaders: CEO Sam Altman and CTO Mira Murati believe AGI could be achieved within the next decade. Altman has expressed confidence that AGI will become the most powerful tool humanity has ever created.
- Nvidia CEO: Jensen Huang has stated that if AGI is defined as the ability to pass any conceivable test, it could be realized in five years.
- Elon Musk: Musk predicts that AI could surpass human intelligence by 2025 or 2026. He also mentioned that the total amount of “sentient compute”—or independent AI thinking—could exceed human capability within five years.