OpenAI Unveils Five-Tier System to Track AI Progress Beyond Human

OpenAI, creator of ChatGPT, has unveiled a five-tier ranking system to monitor its progress in creating artificial intelligence software that "surpasses human capabilities''. This system aims to provide better insight into the company's approach to security and the future of AI.

During a general meeting on Tuesday, an OpenAI spokesperson told Bloomberg that this classification system, soon to be shared with investors and other stakeholders, ranges from currently available conversational AI (Level 1) to AI capable of performing the functions of an entire organization (Level 5).

• Level 1 (Current): Conversational AI, such as ChatGPT, which interacts in natural language but with limitations.

• Level 2 "Reasoners": AI capable of solving complex problems at the level of someone with a PhD, without the need for external tools.

• Level 3 "Agents": Systems that can work autonomously for days, handling multiple tasks.

• Level 4 "Innovators": AI that generates original ideas and makes scientific discoveries on its own.

• Level 5 "Organizations": AI with the ability to run complex operations, such as those of an entire enterprise.

OpenAI executives explained that the company is currently at the first level but is close to reaching the second level, called "Reasoners," according to Wired.

According to an Open AI spokesperson, the goal of the classification system is to provide a clear and understandable measure of progress toward creating increasingly sophisticated AI. Additionally, it seeks to help employees, investors, and the general public better understand OpenAI's approach to security and the future of AI.

At the same meeting, OpenAI demonstrated a research project with its GPT-4 AI model, which exhibits abilities approaching human reasoning. The company has claimed for years that it is working on the development of artificial general intelligence (AGI): computers that can perform most tasks better than people.

Sam Altman, CEO, OpenAI, has previously mentioned that he hopes to achieve AGI within this decade, Wired notes.

In March, at Nvidia's GTC conference, Nvidia CEO Jensen Huang projected a five-year timeline for approaching this possibility, noting that the definition of AGI will influence the speed of its development, which supports Altman's previous statements. Huang said if AGI is defined as a specific set of tests on which a software program can perform better than most people, it could be achieved in five years, La Vanguardia reported.

Some experts, beyond the current abilities of chatbots or systems that solve problems using data and language in AI, question the feasibility of achieving an AGI. For this group of researchers, as reported by Wired, the fluency and capabilities of these technologies do not reflect a human-like mental structure, raising doubts about the very possibility of achieving an AGI that there is a lack of philosophical, rather than scientific, consensus on how to define and achieve this level of AI.

“Five tiers might seem to imply a slow, steady progression up the capability staircase to Level 5, it is just as possible that OpenAI will hold its AGI cards close to the vest and suddenly announce a “Eureka!” moment allowing it to leapfrog a level or two to achieve AGI,” declared an Open AI spokesperson to Bloomberg.