OpenAI and training data company Handshake AI have reportedly been requesting third-party contractors to submit real work examples from their past and current jobs. This initiative is part of a broader strategy within the AI industry, where companies are enlisting contractors to provide high-quality training data in the hopes of advancing their models to automate more white-collar tasks.
According to a report from Wired, OpenAI’s contractor presentation asks individuals to detail the tasks they have completed in previous roles and upload tangible examples of their work. These examples are required to be actual files, such as Word documents, PDFs, Powerpoint presentations, Excel sheets, images, or code repositories. Contractors are instructed to remove any proprietary or personally identifiable information before submission, with guidance on using ChatGPT’s “Superstar Scrubbing” tool for this purpose.
However, intellectual property lawyer Evan Brown has raised concerns about the risks associated with this approach, emphasizing the need for a high level of trust in contractors to determine what information is confidential. Despite these concerns, OpenAI has chosen not to comment on the matter at this time.
This development highlights the growing importance of quality training data in the AI industry, as companies strive to enhance the capabilities of their models and move closer to automating complex tasks. As the use of AI continues to expand, the handling of sensitive information and intellectual property will remain a critical consideration for organizations seeking to harness the power of artificial intelligence.
