OpenAI Develops Scale to Measure AI Advancement, Reports Bloomberg

OpenAI has devised a five-tier scale to assess the development of AI capable of surpassing human abilities, according to a company representative cited by Bloomberg. 

The first level, Chatbots, signifies the ability of a neural network to interact with humans “in conversational language.” 

The company believes its AI is currently on the cusp between the first and second levels, known as Reasoners. To fully achieve this stage, the neural network must perform basic tasks at the level of a person with a doctoral degree, without access to any tools. 

The third level, Agents, implies that AI can perform actions on behalf of a user for several days. 

At the fourth level, Innovators, the neural network is capable of inventing innovations.

The fifth level, Organizations, is characterized by the ability to perform the work of an entire organization. 

The company will study human feedback and may adjust the levels over time, Bloomberg noted. 

The agency added that OpenAI’s leadership demonstrated new skills of GPT-4, which elevate AI to the level of human thinking. 

Back in July, OpenAI partnered with the laboratory that developed the atomic bomb.

Earlier, the firm introduced a “more human-like” chatbot model — GPT-4o.

In June, it became known that the startup is considering transforming into a commercial business.