Apple recently revealed its latest lineup of products, and the AirPods Pro 3 come with a headline-grabbing feature: live translation. The new tool is designed to let users speak in one language and instantly hear a translation in another. The process sounds simple, but some say real-world conversations are far from neat.
Progress in machine translation
“This is a space where there’s just been continuous progress over the last decade. I mean, I’ve been working on it for 25 years now,” Philipp Koehn, a professor at Johns Hopkins Whiting School of Engineering, said on the advancement of machine translation.
Download the SAN app today to stay up-to-date with Unbiased. Straight Facts™.
Point phone camera here
The process uses computers to convert text or speech from one language to another. Apple’s system relies on that same process to make AirPods translate conversations in real time.
The AirPods will first convert spoken words into text, then Apple Intelligence translates the text, and finally, a synthetic voice speaks the translation back to the user.
Apple’s demonstrations make it appear seamless, but experts caution that speech technology has limits.
The problem with real conversations
Conversations rarely happen in pristine conditions. Background noise, overlapping speech and casual slang often throw translation models off. That’s before you factor in idioms, tone or cultural nuance.
Koehn explained that nuance often disappears in the translation process. “Like if I slightly emphasize a word or hesitate somewhere a little bit, all that means something. Maybe I’m not certain, or I’m fishing for confirmation. Those subtle clues still get lost,” he said.
Dialects and idioms
The technology performs best with clean, standard speech. Dialects are a bigger problem.
“If you go into some village in Scotland, the technology might not work anymore because they speak very, very strong dialects that the systems aren’t trained on,” Koehn said.
Idioms, however, may not be the stumbling block many expect. “There’s actually a lot of translated text out there — billions of words. So you’ll see all these common phrases many, many times in the training data,” he said.
For now, Apple’s live translation only supports a handful of major European languages. Asian languages are not included in the launch. Koehn explained that systems perform best on these languages because there is so much data to train them on.
“If you go beyond the top 100 languages, they tend to be languages that are mostly spoken and not written,” he explained.
A useful tool, not a replacement
Even when the translation is accurate, Koehn said users should expect a lag of two or three seconds. That delay, he added, makes conversation less natural.
Because of this, he recommends treating the feature as a supplement rather than a substitute for learning a language. While live translation might help travelers order dinner or ask for directions, it won’t capture the warmth of a greeting or the cultural connection that comes with speaking directly.
“This technology helps,” Koehn said, “but it doesn’t replace the power of [you] saying hello in their language.”
The new AirPods hit shelves Sept. 19.
