OpenAI and Handshake AI are reportedly requesting contractors to submit examples of real work from their previous or current jobs to enhance AI training data quality. This initiative aligns with a broader trend among AI companies aiming to improve automation in white-collar tasks. Contractors are instructed to upload various types of concrete outputs—such as Word documents, PDFs, PowerPoints, and images—without proprietary or personally identifiable information. A tool called “Superstar Scrubbing” is provided to assist with data sanitization. However, intellectual property lawyer Evan Brown warns that this practice poses significant risks, as it relies heavily on contractors to determine what information is confidential. OpenAI has not publicly commented on this initiative. By focusing on real-world examples, OpenAI aims to refine its models, underscoring the need for trusted data sources in the evolving AI landscape. This move highlights the balance between leveraging real-world insights and safeguarding sensitive information.
Source link
Share
Read more