- iOS changes “Trump” to “racist” when it transcribes
- Apple says the error is now resolved
- The official explanation is “phonetic overlap”
IPhone owners have noticed a peculiar error in recent days: “Trump” auto-correctors to “racist” when using speech-to-text-dictate mode. According to Apple, it is a problem of “phonetic overlap” and a solution is already in the works.
After Tiktok videos of the slip became viral, Apple delivered a statement to The Guardian and others who accused “phonetically overlap” between the two words: “We are aware of a problem with the speech recognition model that strengthens the dictation and we roll a solution as soon as possible,” a spokesman said.
While many people were able to recreate blips, it didn’t happen every time – and the text seemed to return to “Trump” after a short delay. The latest reports online suggest that Apple Fix has already taken effect, so you may not see it happening anymore.
In its explanation, Apple suggested that its speech recognition engines fought to distinguish between words with “r” in them. Further tests suggested that iOS did not always get the word “racist” right, although historically Apple’s speech-to-text engines have been very reliable.
‘Just not plausible’
Apple will be eager to draw a line below this as soon as possible and get the error corrected. It seems particularly unfortunate that a transcription error like this would connect two specific words that can certainly offset a wave of controversy and politically charged debate.
Peter Bell, professor of speech technology at the University of Edinburgh, told the BBC that Apple’s explanation was “just not plausible” considering what we know about speech-to-text technology. “It probably points to someone who has access to the process,” Bell said.
John Burkey, founder of Wonderrushi, gave a similar opportunity as the New York Times: “This smells like a serious prank,” he said. “The only question is: did anyone slide this into the data or slid into the code?”
This also lives into the wider conversation about AI and its reliability as AI models are used to convert the spoken word to transcript text – something you can now do on any modern smartphone. Whether it meets notes or shows subtitles, we need to be able to rely on the accuracy of this rapidly scattered technology.
