But Whisper has a major flaw: It is prone to making up chunks of text or even entire sentences, according to interviews with more than a dozen software engineers, developers and academic researchers. Those experts said some of the invented text — known in the industry as hallucinations — can include racial commentary, violent rhetoric and even imagined medical treatments.
Experts said that such fabrications are problematic because Whisper is being used in a slew of industries worldwide to translate and transcribe interviews, generate text in popular consumer technologies and create subtitles for videos.
More concerning, they said, is a rush by medical centers to utilize Whisper-based tools to transcribe patients’ consultations with doctors, despite OpenAI’ s warnings that the tool should not be used in “high-risk domains.”
The full extent of the problem is difficult to discern, but researchers and engineers said they frequently have come across Whisper’s hallucinations in their work. A University of Michigan researcher conducting a study of public meetings, for example, said he found hallucinations in eight out of every 10 audio transcriptions he inspected, before he started trying to improve the model.
A machine learning engineer said he initially discovered hallucinations in about half of the over 100 hours of Whisper transcriptions he analyzed. A third developer said he found hallucinations in nearly every one of the 26,000 transcripts he created with Whisper.
The problems persist even in well-recorded, short audio samples. A recent study by computer scientists uncovered 187 hallucinations in more than 13,000 clear audio snippets they examined.
That trend would lead to tens of thousands of faulty transcriptions over millions of recordings, researchers said.
This story is from the November 02, 2024 edition of Techlife News.
Start your 7-day Magzter GOLD free trial to access thousands of curated premium stories, and 9,000+ magazines and newspapers.
Already a subscriber ? Sign In
This story is from the November 02, 2024 edition of Techlife News.
Start your 7-day Magzter GOLD free trial to access thousands of curated premium stories, and 9,000+ magazines and newspapers.
Already a subscriber? Sign In
AUSTRALIAN STATES BACK NATIONAL PLAN TO BAN CHILDREN YOUNGER THAN 16 FROM SOCIAL MEDIA
Australia’s states and territories unanimously backed a national plan to require most forms of social media to bar children younger than 16.
FACEBOOK AND INSTAGRAM USERS IN EUROPE CAN OPT FOR LESS PERSONALIZED ADS
Facebook and Instagram users in Europe will get the option to see less personalized ads if they don’t want to pay for an ad-free subscription, social media company Meta said Tuesday, bowing to pressure from Brussels over privacy and digital competition concerns.
IN THIS FLORIDA SCHOOL DISTRICT, SOME PARENTS ARE PUSHING BACK AGAINST A CELLPHONE BAN
It’s no surprise that students are pushing back on cellphone bans in classrooms. But school administrators in one South Florida county working to pull students’ eyes away from their screens are facing some resistance from another group as well – parents.
'SATURDAY NIGHT LIVE' TO TRUMP: 'WE'VE BEEN WITH YOU ALL ALONG'
The first “Saturday Night Live” since Donald Trump’s election victory began with the most somber of tones as a group of plainly dressed cast members, primarily women and minorities, described their new reality.
DIAMOND SPORTS GROUP WILL OFFER SINGLE-GAME PRICING TO STREAM NBA AND NHL GAMES STARTING NEXT MONTH
The nation’s largest owner of regional sports networks will offer single-game pricing for NBA and NHL games beginning next month.
ON THE EVE OF OSCARS HONOR, JAMES BOND PRODUCERS REFLECT ON LEGACY AND FUTURE OF 007
For the late James Bond producer Albert “Cuddy” Broccoli, receiving the Irving G. Thalberg Memorial Award was a true high point in his career.
WAYMO'S ROBOTAXIS NOW OPEN TO ANYONE WHO WANTS A DRIVERLESS RIDE IN LOS ANGELES
Waymo this week opened its robotaxi service to anyone who wants a ride around Los Angeles, marking another milestone in the evolution of self-driving car technology since the company began as a secret project at Google 15 years ago.
US AGENCY SAYS TESLA'S PUBLIC STATEMENTS IMPLY THAT ITS VEHICLES CAN DRIVE THEMSELVES. THEY CAN'T
The U.S. government’s highway safety agency says Tesla is telling drivers in public statements that its vehicles can drive themselves, conflicting with owners manuals and briefings with the agency saying the electric vehicles need human supervision.
STELLANTIS RECALLS JEEP AND DODGE SUVS TO FIX COMPUTER PROBLEM THAT CAN DISABLE BRAKE SAFETY DEVICES
Stellantis is recalling about 207,000 Jeep and Dodge SUVs in the U.S. to fix a computer problem that can disable the anti-lock brakes and electronic stability control safety features.
VATICAN, MICROSOFT CREATE AI-GENERATED ST. PETER'S BASILICA TO ALLOW VIRTUAL VISITS, LOG DAMAGE
The Vatican and Microsoft this week unveiled a digital twin of St. Peter’s Basilica that uses artificial intelligence to explore one of the world’s most important monument’s while helping the Holy See manage visitor flows and identify conservation problems.