Google’s AI team tells job seekers its HR filters are unreliable

Google's AI team warns job seekers its internal filters for screen applicants can be unreliable. The company's AI systems are marketed to clients as a means to efficiently review numerous applications and identify the most suitable candidates. However, some Google DeepMind researchers have expressed reservations about relying on these tools for their own hiring.

To mitigate the risk of applications being incorrectly screened out, the AGI Safety and Alignment Team has encouraged candidates to complete a special form alongside their application. The document seen by Bloomberg includes the disclaimer "PLEASE DO NOT SHARE THIS DOC WIDELY." The form's purpose is to ensure a human reviewer on the team can assess the applicant's qualifications.

A Google DeepMind spokesperson stated that the company's goal is to hire the most qualified talent and denied that the company's systems incorrectly filter out applicants. As AI becomes increasingly integrated into the hiring process, concerns arise about potential discrimination and the impact on HR roles. Google's Workspace team promotes its AI features to streamline job posting, resume evaluation, and hiring needs forecasting.

However, AI hiring systems have also been criticized for possible bias against candidates. The Bloomberg investigation uncovered signs of potential bias in OpenAI's ChatGPT based on applicants' names, while Workday is facing a lawsuit alleging its AI systems screen applicants based on race, age, and disability. Workday has denied these allegations, with the company stating that humans make the final hiring decisions.

The Google DeepMind team is cautious about candidates gaming the filters by using AI or flooding the system with numerous applications. The form designed to bypass automated screening advises that real humans will read the applications and have become weary of reading LLM (large language model) responses, as they often sound very similar.