A New Framework for Measuring AI Capabilities
In a significant development in the field of Artificial Intelligence (AI), OpenAI, the company behind ChatGPT, has unveiled a new 5-tier system to measure its progress toward achieving Artificial General Intelligence (AGI). This ranking system was first introduced internally during an all-hands meeting before being shared with the public. According to a report by Bloomberg, an OpenAI spokesperson revealed that their current GPT-4 models are classified at Tier 1, with the more advanced Tier 2 ‘Reasoners’ anticipated in the near future.
The 5-Tier System Explained
The newly introduced tiers are designed to help OpenAI and the broader AI community conceptualize the different stages of AI capabilities as they strive to develop models with genuine reasoning abilities. Here are the levels in the system:
Tier 1 – Chatbots
- Definition: AI with conversational language capabilities.
- Description: OpenAI believes their current GPT-4 models fall into this category, demonstrating the ability to engage in natural-sounding conversations and answer questions to the best of its knowledge.
- Examples:
- Chatbots that can respond to user queries on a wide range of topics.
- Virtual assistants like Siri or Alexa.
Tier 2 – Reasoners
- Definition: AI with human-level problem-solving skills, comparable to a person with a PhD but without any tools.
- Description: This tier represents the next step in AI development, where models can demonstrate advanced reasoning and problem-solving abilities, similar to those of a highly educated individual.
- Examples:
- AI systems that can analyze complex data sets and provide insightful recommendations.
- Models that can generate human-like responses to open-ended questions.
Tier 3 – Agents
- Definition: Systems that can take actions on a user’s behalf.
- Description: At this tier, AI models can interact with the physical world, making decisions and taking actions based on their understanding of the environment.
- Examples:
- Self-driving cars that can navigate through complex traffic scenarios.
- Robots that can perform tasks like assembly or packaging.
Tier 4 – Innovators
- Definition: AI that can generate its own innovations.
- Description: This tier represents the pinnacle of AI development, where models can create new ideas and solutions on their own, rather than relying on human input.
- Examples:
- AI systems that can design new products or services.
- Models that can develop novel algorithms for solving complex problems.
Tier 5 – Organizations
- Definition: AI capable of performing the work of an entire organization.
- Description: At this tier, AI models have reached a level of sophistication where they can manage and execute tasks autonomously, much like a human organization would.
- Examples:
- AI systems that can manage supply chains and logistics.
- Models that can oversee financial transactions and make investment decisions.
AGI: A Long-Term Objective
Artificial General Intelligence remains a long-term objective for many companies in the AI sector, including Mark Zuckerberg’s Meta. While OpenAI’s models are currently at Tier 1, the spokesperson mentioned they are close to achieving the second level, ‘Reasoners.’ During the internal meeting where the new tiers were unveiled, OpenAI also showcased new research involving its GPT-4 model, which they believe demonstrates skills approaching human-level reasoning.
A Flexible and Evolving System
This ranking system was developed by OpenAI’s senior leadership team and executives. It is not considered final and may be revised as the organization receives feedback and further input from employees and investors to better align with the evolving understanding of AI progress.
OpenAI’s Mission and Recent Changes
OpenAI’s mission is to create ‘safe and beneficial artificial general intelligence for the benefit of humanity.’ However, the recent dissolution of its safety-oriented Superalignment group, following the departure of Chief Scientist and co-founder Ilya Sutskever, has raised concerns about the company’s ability to uphold this mission.
In conclusion, OpenAI’s introduction of a 5-tier system marks a significant step forward in the development of AI. As researchers and developers work towards achieving AGI, it is essential to have a clear framework for measuring progress and understanding the different stages of AI capabilities. While challenges lie ahead, the potential benefits of AGI make it an exciting area of research that holds promise for improving human lives.
