Inside the OpenAI project where freelancers train ChatGPT on everything from farming to commercial flying

In a groundbreaking effort to refine artificial intelligence, freelancers are playing a crucial role in enhancing ChatGPT's capabilities across diverse fields, from agriculture to aviation. The initiative, known internally as Project Stagecraft at HandshakeAI, has enlisted thousands of contractors who are compensated at least $50 per hour to develop training materials that will hone the AI's understanding of various professional tasks. Documented insights from Business Insider reveal that this extensive project has engaged between 3,000 to 4,000 freelancers, each tasked with creating realistic work scenarios that reflect their specialized knowledge. These contractors, who hail from a variety of industries, are instructed to embody specific roles, such as animal husbandry experts or music composers, and to design tasks that simulate real-world applications in their fields. Handshake AI, based in San Francisco, has grown from a platform for young professionals into a significant player in the data-labeling sector, joining the ranks of numerous startups that provide critical training data for leading tech firms. This data is essential for improving a range of technologies, including self-driving vehicles and advanced chatbots from companies like OpenAI and Meta. The nature of the work has evolved, with a clear trend towards more specialized tasks that often require deep knowledge or advanced degrees. According to the information available on Handshake's platform, certain expert contractors can earn up to $500 per hour for roles not explicitly linked to OpenAI. A Handshake training document emphasizes that the focus of this initiative is on intellectual tasks rather than manual labor, aiming to create a comprehensive map of economically relevant activities to enhance the AI's performance. One contractor candidly noted the irony of their work: they were aware that they were essentially training an AI that could potentially replace them. In a notable aspect of the project, contractors are guided to create detailed personas, complete with specific context, objectives, and deliverables that reflect their professional environments. This process includes drafting prompts that are designed to emulate real job expectations without oversimplifying the complexities involved. To ensure the quality and accuracy of the training material, each submission undergoes a rigorous review process, including evaluations by industry experts and additional oversight from OpenAI. This collaborative effort aims to ensure that the AI learns from authentic human expertise, enhancing its reliability and ethical standards in real-world applications. However, the project has not been without its challenges. Reports have emerged of contractors facing significant payment disputes, with some alleging that Handshake denied them thousands of dollars for their contributions due to alleged violations of platform rules. Handshake's representatives have refrained from commenting on these payment issues, while OpenAI has also remained silent on the matter.