OpenAI is firmly established as one of the key players in the pursuit of human-level artificial intelligence, reportedly just four steps away from achieving this goal.
The company has developed a five-level system to monitor its progress toward artificial general intelligence (AGI), which it shared with employees this week, an OpenAI spokesperson revealed to Bloomberg. These levels range from the current conversational AI to AI capable of performing organizational tasks. OpenAI plans to disclose these levels to investors and external stakeholders.
Currently, OpenAI executives believe they have reached the first level, with the organization nearing the second level, defined as “Reasoners.” This stage describes AI capable of basic problem-solving, akin to a human with a doctorate but without tools. The third level, dubbed “Agents,” refers to AI that can operate independently for several days on the user’s behalf. The fourth level, “Innovators,” describes AI that assists in creating new inventions.
In a notable demonstration, OpenAI showcased a research project involving GPT-4, which exhibited human-like reasoning abilities, according to a source familiar with the matter reported by Bloomberg. The company declined to provide further comments.
This system was reportedly created by OpenAI executives and leaders, with the potential for modifications based on feedback from employees, investors, and the company board.
In May, OpenAI dissolved its “Superalignment” team, which focused on addressing AI’s existential risks. The responsibilities of this team have since been integrated into other research initiatives across the company.