OpenAI announced a set of 5 levels to track AI development progress, to help users better understand the safety and future of AI, according to Bloomberg.

The new system, shared internally at a meeting on July 9, divides the AI ​​development process into 5 levels: from the ability to perform basic tasks (Level 1) to replacing an entire organization. human function (Level 5). Specifically, OpenAI's 5 AI levels include:

  • Level 1: AI is capable of performing basic tasks such as answering questions, translating languages, and summarizing text.

  • Level 2: AI is capable of solving complex problems that require specialized knowledge, logic and reasoning, equivalent to a doctoral level.

  • Level 3: AI is capable of automating processes, completing tasks that humans usually do such as writing reports and analyzing data.

  • Level 4: AI possesses the ability to create art, new products including works of art, music, literature, or even scientific inventions.

  • Level 5: This is the highest level, AI can do any job that humans can do, even better, replacing an entire human organization.

OpenAI leadership said that the ChatGPT chatbot is currently at Level 1 and is confident it will soon reach Level 2 with the ability to solve problems equivalent to a PhD without the need for support tools. The proof is that the research project applying the GPT-4 model has shown new skills, approaching human-like reasoning ability.

Although it is unclear whether OpenAI has a goal of developing a 6th level of AI or the place of “Artificial Super Intelligence” (ASI) in the current system, this announcement has sparked much debate. While OpenAI claims to be aiming for “artificial general intelligence” (AGI) – computers that accomplish tasks that outperform humans – their definition of AGI is that “autonomous systems outperform humans in most jobs of economic value”, has not really received consensus from the AI ​​research community.

In fact, OpenAI's competitor - Google DeepMind, proposed a different AI classification system in November 2023, focusing on the "emerging", "proficient", "expert", "advanced" levels. teacher” and “superhuman”, instead of focusing on the ability to do “the work of an entire organization” like OpenAI.

OpenAI's failure to mention AGI in its classification system may be a strategy to avoid causing confusion among the public. It should be remembered that according to the agreement between OpenAI and Microsoft, AGI will not be part of the intellectual property and commercial licensing agreement. In other words, OpenAI will exclusively exploit the huge profits from AGI when achieved. The decision “AGI has arrived” belongs to the OpenAI board of directors, which currently includes CEO Sam Altman, President Bret Taylor and six other members.

OpenAI CEO Sam Altman has previously said that he hopes artificial general intelligence can be achieved within this decade.

Despite the controversy surrounding AGI, OpenAI's announcement of a 5-level system marks the company's efforts to make its AI development process transparent with the public and investors. This system is expected to continue to be improved based on input from the community. The collaboration between OpenAI and Los Alamos National Laboratory in building this system further shows its importance to the future of AI.