OpenAI and training data company Handshake AI are asking third party contractors to upload real work from past and current jobs as part of efforts to collect higher quality data for training artificial intelligence models, according to a report by Wired.
What Contractors Are Being Asked To Provide
The report said OpenAI shared a company presentation that instructs contractors to describe tasks they have carried out in other jobs and to submit examples of real on the job work they have actually completed. The materials can include full files rather than summaries, such as Word documents, PDFs, PowerPoint presentations, Excel spreadsheets, images, or code repositories.
The request forms part of a wider strategy across the AI sector in which companies are hiring contractors to generate detailed, realistic training data so that models can be trained to handle more white collar tasks.
Data Handling Instructions
OpenAI reportedly tells contractors to remove proprietary and personally identifiable information before uploading any files. To do this, they are directed to use a ChatGPT tool called Superstar Scrubbing, which is intended to help clean documents of sensitive material before submission.
Handshake AI, which works with OpenAI on data collection, is also involved in requesting that contractors provide this kind of material, according to the Wired report.
Legal And Confidentiality Risks
Evan Brown, an intellectual property lawyer, told Wired that this method carries substantial risk for AI labs. He said the approach depends heavily on contractors making correct judgments about what information is confidential and what can be shared, which creates potential exposure for companies collecting the data.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.