OpenAI has reportedly created a new classification that shows different levels of artificial intelligence (AI) development towards achieving superintelligent AI. As per the report, the classification system, which OpenAI has shared with its employees, has five levels: the first level includes existing generative AI-powered conversational chatbots and the fifth would include yet to be achieved artificial general intelligence (AGI), an AI system considered to be powerful enough to handle the work of an entire organisation. It is said that the AI firm also shared details of a new project that utilises GPT-4 model and demonstrates human-like reasoning — an ability that could take the model to level two.
OpenAI Reportedly Shares Five Levels to AGI
Citing an OpenAI spokesperson, a Bloomberg report highlighted that the company shared the new classification system with its employees in an “all-hands meeting”. The company reportedly plans to share the classification with its investors and people outside of the organisation as well, although no timeline for the same was shared.
Currently, OpenAI is said to be at level one, however, citing people familiar with the matter, the report claimed that the AI firm shared new capabilities of GPT-4 that might be equivalent to level two. The most notable feature is reportedly the reasoning capability of the AI model. No other features were mentioned.
The publication also shared the five levels towards AGI. These are:
- Level 1 — Chatbots: AI with conversational language
- Level 2 — Reasoners: Human-level problem solving
- Level 3 — Agents: Systems that can take action
- Level 4 — Innovators: AI that can aid in invention
- Level 5 — Organisations: AI that can do the work of an organisation
Based on the reported classification, level one is where most AI systems today are with conversational chatbot capabilities. OpenAI is said to have reached level two dubbed Reasoners with GPT-4, however, this capability is yet to be demonstrated by the company.
Level three includes Agents, which can execute actions on behalf of people. This is likely the stage where AI systems can independently take certain actions based on behaviour patterns. This level also includes the integration of AI into real-world operations. The fourth level, as per the classification, will give AI a certain level of innovation and it will be able to assist in inventing new technologies and products. Notably, true innovation is said to be one of the defining traits of human intelligence.
Finally, level five or AGI is when an AI system can handle the tasks of an entire organisation. This would far exceed the capability of humans and would take AI to the superintelligent tier.