OpenAI recently introduced a new five-tier system to assess the progress of artificial general intelligence (AGI). This system aims to provide a clear structure for understanding AI advancement, although it mainly describes hypothetical technology that is not yet a reality. The primary goal of OpenAI is to achieve AGI, which refers to AI systems that can perform tasks like humans without specialized training. While this goal drives much of the hype surrounding the company, it could also have significant societal impacts if achieved.
The CEO of OpenAI, Sam Altman, has expressed his belief that AGI could be accomplished within this decade. To handle the potential disruption that AGI might bring, the company has developed a ranking system to communicate internal milestones towards AGI. The five levels range from current AI capabilities to systems that could potentially manage entire organizations. As of now, OpenAI’s technology is at Level 1, focusing on conversational interactions, but the company is on the brink of reaching Level 2, known as “Reasoners.”
A Level 2 AI system would possess basic problem-solving skills comparable to a human with a doctorate degree but without external tools. OpenAI’s GPT-4 model has shown promising signs of approaching this human-like reasoning ability, as demonstrated in a recent research project. The classification system also envisions higher levels of AI capabilities, such as autonomous task completion, innovation generation, and organizational management.
While OpenAI’s classification system is still a work in progress, it reflects the company’s ambitious goals and efforts towards achieving AGI. Other AI labs, including Google DeepMind, have also proposed similar frameworks to assess AI advancement, indicating a growing interest in quantifying AI capabilities. However, measuring AI progress poses challenges, especially in defining what constitutes advancement and potential risks associated with AI development.
Overall, OpenAI’s five-tier system serves as a communication tool to showcase the company’s aspirations and attract investors. The AI research community has yet to reach a consensus on measuring progress towards AGI, highlighting the complexity of achieving such a goal. As AI technology continues to evolve, frameworks like OpenAI’s classification system offer a glimpse into the future possibilities of AI capabilities and their potential impact on society.
