OpenAI has unveiled internally a five-level classification system that measures its progress towards the creation of Artificial General Intelligence (AGI), offering a detailed insight into its path towards AI with human-like capabilities.
A Classification System for AI Progress
OpenAI’s roadmap spans from Level 1, corresponding to current conversational AI, to Level 5, representing an AI capable of managing entire organizations. Currently, OpenAI’s technology is at Level 1, with significant progress towards Level 2, known as “Reasoners.”
Capabilities of Each Level
– Level 1: Conversational AI – This stage encompasses current capabilities of maintaining conversational interactions with humans.
– Level 2: Reasoners – Implies systems capable of solving problems at the level of a human with a Ph.D.
– Level 3: Agents – Defines systems that can act on behalf of a user for extended periods.
– Level 4: Innovators – Corresponds to an AI capable of generating innovations.
– Level 5: Organizations – Refers to AI systems that can perform the work of an entire organization.
Progress and Recent Demonstrations
At a recent meeting, OpenAI showcased a GPT-4 research project demonstrating human-like reasoning abilities, indicating advancements towards Level 2. Such demonstrations highlight rapid progress and exponential acceleration of technology, suggesting significant advancements may be seen in the near future.
Outlooks and Expectations
OpenAI’s roadmap not only provides clarity on its goals but also establishes a framework for evaluating progress and setting realistic expectations in AGI development. According to a Bloomberg report, OpenAI plans to release this classification system to investors and other stakeholders.
In October, OpenAI CEO Sam Altman and CTO Mira Murati expressed confidence that AGI could become a reality within the next decade. Altman stated: “We are convinced that when people are offered better tools, they do things that leave you amazed. And I believe that AGI will be the best tool humanity has created so far.”
Industry Predictions
Nvidia CEO Jensen Huang suggested in March that AGI, defined as the ability to outperform any presented test, could be achieved within five years. Elon Musk has proposed that AI could surpass the intelligence of the brightest humans as soon as 2025 or 2026 and that the “total amount of conscious computing” will surpass all humans within five years.
These predictions, along with OpenAI’s new classification system, underscore the rapid progress in the field of artificial intelligence and the growing anticipation surrounding AGI development. This milestone could radically transform our society and economy, and OpenAI’s roadmap provides a clear structure to track these advancements.