In an attempt to provide a clearer understanding of its views on safety and the future of artificial intelligence (AI), OpenAI has developed a set of five stages to monitor its progress toward developing software that can perform better than humans.
During an all-hands meeting, the company that created ChatGPT informed its staff about the new classification scheme, according to an OpenAI representative.
The levels, which OpenAI intends to disclose to potential investors and external stakeholders, span from the current state of AI that can converse with humans (level 1) to AI that can do organizational tasks (level 5).
According to the spokesman, the corporation thinks it is now on the first level but is about to enter the second, which it refers to as “Reasoners.” This is a reference to both humans with doctorates and systems capable of performing simple problem solving tasks without the need for any tools.
A person familiar with the discussions said that during the same meeting, business leadership presented a demonstration of a research project involving its GPT-4 AI model, which OpenAI believes has certain new talents that climb to human-like reasoning.
For many years, OpenAI has stated that it is developing artificial general intelligence (AGI), or machines that are substantially more proficient than humans at the majority of jobs.
AGI might be achieved this decade, according to OpenAI CEO Sam Altman’s prior statements.
The standards for achieving AGI have long been a topic of discussion among AI researchers. In a publication, a number of Google DeepMind researchers put up a framework for five escalating AI categories, which included designations like “expert” and “superhuman.”
The third level, “Agents,” would relate to AI systems that can work on behalf of a user for several days, based on the tiers that OpenAI has developed. Level 4 talks about AI that can generate novel ideas. The highest level would be referred to as ” Organisations .”
Leave a comment