The recent development by OpenAI includes a new complex system for determining the measures of the offered intelligent tools. This development is to yearly monitor the progress of AI against the leveled-up human-like intelligence.
At present, OpenAI is in the first level. Still, it is moving to the second stage which is called ‘Reasoners. ’ That level defines AI systems capable of solving elementary problems similar to a human with a doctorate, but no more tools incorporated.
During this meeting, leadership presented an example of one of the conducted research based on the GPT-4 model and examples of new functions that can work with an increased degree of human-like intelligence.
Five-Tier System to Evaluate AI Advancements Toward Human-Level Intelligence: See How It Works
The five-tier framework is not yet complete and getting feedback from employees, investors, and the board is expected to be requested.
OpenAI’s long-term objective is to reach artificial general intelligence (AGI) which means to create AI systems that are better than humans in most areas. OpenAI’s current CEO Sam Altman still holds the possibility of attaining AGI in this decade.
Since its initiation, OpenAI’s operation has been to build AGI, a kind of artificial intelligence that will surpass human capabilities in most domains.
Currently, AGI is not present, but, as CEO Sam Altman commented, it is possible to achieve it this decade. The parameters for getting to AGI have been discussed among the investigators of artificial intelligence.
In November 2023, the researchers from Google DeepMind outlined five levels of AI and phases like ‘expert’ and ‘superhuman,’ based on the automotive system for self-driving cars. Similarly, the newly introduced levels in OpenAI also outline a five-step progression towards an AGI.
“Agents” is the third level that addresses AI systems that work for several days for users independently; the fourth level goes for creating innovations and the last one “Organizations” means independent operation in an organization.
The intelligence framework that OpenAI implemented has five levels of competence. Currently, ChatGPT is at Level 1 but is in the process of moving to Level 5. Each level corresponds to a specific set of capabilities:
Level 1: Simple problem-solving skills that can be compared to those of a normal human being. Level 5: Similar to the problem-solving skills of a PhD level candidate.
The five-tier framework has been developed but is not yet set in stone and it is planned to get feedback from the employees, investors, and the board. OpenAI’s long-term aim is to create artificial general intelligence or AI that is better than humans at most tasks. Currently, CEO Sam Altman still believes that AGI could be reached before the end of this decade.
These tiers were created by OpenAI’s executives and senior management and the company still considers them a work in progress. The company also plans to get feedback from employees, investors, and its board of directors on the levels to improve them.
Report By Bloomberg