According to a confidential document from OpenAI, the company has "hired folks across occupations to help collect real-world tasks modeled off those you’ve done in your full-time jobs, so we can measure how well AI models perform on those tasks." The document instructs contractors to "take existing pieces of long-term or complex work (hours or days) that you’ve done in your occupation and turn each into a task."

This initiative is part of OpenAI's broader effort to assess its progress toward achieving artificial general intelligence (AGI). In September, OpenAI launched a new evaluation process focused on comparing the performance of its AI models against human professionals across diverse industries. OpenAI defines AGI as an AI system that surpasses human capabilities in most economically valuable tasks.

The data collected from these real-world tasks will be used to train and refine OpenAI's AI models, enabling them to better understand and execute complex assignments. By comparing AI performance against a human baseline, OpenAI hopes to identify areas where its models excel and areas that require further improvement. This approach is crucial for developing AI systems that can effectively augment or even automate various professional tasks.

The implications of achieving AGI are far-reaching, potentially transforming industries and reshaping the nature of work. While the development of AGI promises significant benefits, such as increased productivity and innovation, it also raises concerns about job displacement and the ethical considerations of increasingly autonomous AI systems. OpenAI's efforts to benchmark AI performance against human capabilities are a step towards understanding and addressing these complex issues.

The current status of the project involves ongoing data collection and analysis. OpenAI has not yet released specific details about the performance of its models against the human baseline. However, the company is expected to continue refining its evaluation process and incorporating new data as it progresses towards its goal of achieving AGI. The next developments will likely include further iterations of AI models based on the collected data and ongoing assessments of their performance across a wider range of tasks.