Artificial Intelligence: The Dark Side of the Whisper Tool

The use of Whisper to create captions for the hearing impaired is also a serious problem, as they have no way of detecting false information hidden among the millions of words generated.

In an era where Artificial Intelligence (AI) is becoming an indispensable part of everyday life, a tool called Whisper developed by OpenAI has attracted great attention thanks to its ability to automatically take notes with almost "human-like accuracy."

Behind the acclaim, however, Whisper is facing heavy criticism for a serious flaw: its ability to automatically fabricate information, that is, generate text or even statements that no one has ever written.

According to software engineers and researchers, fictional details not only cause confusion but can also contain racist comments, violent language and unrealistic medical treatments.

Although OpenAI has warned that the tool should not be used in “high-risk areas,” many medical facilities are still actively adopting Whisper to record doctor-patient consultations.

In fact, a recent study found that up to 80% of recordings in a survey of public meetings contained these fictional details. A machine learning engineer also pointed out that an analysis of more than 100 hours of Whisper transcripts found that about 50% of them contained serious errors, raising concerns about the reliability of the technology, which is increasingly being used.

Experts warn that errors can lead to “really serious consequences”, especially in the medical field.

Alondra Nelson, former director of the White House Office of Science and Technology Policy, stressed that no one wants a misdiagnosis, and called for higher standards in the application of this technology.

The use of Whisper to create captions for the hearing impaired is also a serious problem, as they have no way of detecting false information hidden among the millions of words generated.

The rise of fictional texts has prompted many experts, activists, and former OpenAI employees to call on the federal government to consider regulating AI.

William Saunders, a research engineer in San Francisco, believes that this problem is completely solvable if OpenAI prioritizes processing it.

OpenAI has admitted that it is constantly working to minimize fictional details and appreciates feedback from the research community.

However, many engineers still claim that they have never seen another AI note-taking tool with a level of fiction as high as Whisper./.

(Vietnam+)

Source: https://www.vietnamplus.vn/tri-tue-nhan-tao-mat-trai-cua-cong-cu-whisper-post987914.vnp