OpenAI Unveils AI Development Monitoring Framework

Renowned for launching ChatGPT, OpenAI has revealed a new tool to monitor its developments towards artificial intelligence capable of surpassing human ability. During an all-hands meeting, the company unveiled a five-level structure that it will also provide to other outside stakeholders including investors. Part of OpenAI’s continuous initiatives to improve openness and knowledge about its advancement in artificial intelligence development is this mechanism.

From the present AI capabilities, which can engage with people in conversational language (Level 1), to AI systems able of handling challenging organisational tasks, the five levels span Executives of OpenAI told staff members that although the firm is on the brink of Level 2, sometimes known as “Reasoners,” it feels it is now at Level 1. Though without any further tools, these systems can solve simple problems on par with a human with a PhD degree.

OpenAI’s leadership presented a research project including the GPT-4 AI model during the conference, showcasing fresh talents approaching human-like reasoning. Confirming this information was a source acquainted with the conversation who asked anonymity. The spokesman for OpenAI said that, as is typical in the sector, the firm often evaluates new capabilities internally.

Long aiming to create artificial general intelligence (AGI), OpenAI refers to computers capable of surpassing humans in most jobs. Though AGI does not yet exist, OpenAI CEO Sam Altman has projected it could be reached in this decade. Researchers at Google DeepMind presented a similar five-level structure with levels including “expert” and “superhuman” in a November 2023 study, after AI researchers have argued on the requirements for obtaining AGI. This approach reminds me of the mechanism employed in the automobile sector to evaluate the level of automation in self-driving cars.

The third stage on the road to AGI, according to OpenAI’s classification system, is “Agents,” which relate to AI systems able to execute tasks on a user’s behalf across multiple days. Level 4, “Innovators,” talks of artificial intelligence able to create fresh ideas. The most advanced level, “Organisations,” shows artificial intelligence systems capable of doing all chores needed to manage a company.

Considered a work in progress, the framework was created by executives and top leaders of OpenAI. To hone the levels over time, the company intends to get comments from investors, staff, and its board. This project emphasises OpenAI’s dedication to forward artificial intelligence while guaranteeing openness and safety during the development process.