The Humane AI Pin is lost in translation

Of all the things the Humane AI Pin promised, I was most intrigued by translation. In a demo, a man speaks to Humane co-founder Imran Chaudhri in Spanish. The AI Pin automatically translates it to English. Chaudhri replies in English. Again, the AI Pin translates his words back into Spanish. There are notable pauses when the AI is processing, but it’s a powerful concept. Unlike with Google Translate, there was solid eye contact between both people. The AI voice sounded more natural and less robotic. And crucially, there were no screens. The language barrier was still there, but it was much more permeable. 

That’s not what happened when I tried it myself.

I spoke some simple phrases in Japanese and Korean. Instead of translating, the AI Pin spewed gibberish back at me. I asked my colleague David Pierce, who reviewed the damn thing, if I was doing something wrong. I wasn’t. It just didn’t work.

The whole experience was funny. It felt like vindication for the blood, sweat, and tears I poured into two decades of foreign language study. But when I rewatched Humane’s translation demo, my heart broke. I found myself wishing I had something like this when my parents were dying. 

Living in an immigrant, multilingual family will open your eyes to all the ways humans can misunderstand each other. My story isn’t unique, but I grew up unable to communicate in my family’s “default language.” I was forbidden from speaking Korean as a child. My parents were fluent in spoken and written English, but their accents often left them feeling unwelcome in America. They didn’t want that for me, and so I grew up with perfect, unaccented English. I could understand Korean and, as a small child, could speak some. But eventually, I lost that ability. 

I became the family Chewbacca. Family would speak to me in Korean, I’d reply back in English — and vice versa. Later, I started learning Japanese because that’s what public school offered and my grandparents were fluent. Eventually, my family became adept at speaking a pidgin of English, Korean, and Japanese.

This arrangement was less than ideal but workable. That is until both of my parents were diagnosed with incurable, degenerative neurological diseases. My father had Parkinson’s disease and Alzheimer’s disease. My mom had bulbar amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). Their English, a language they studied for decades, evaporated. 

It made everything twice as complicated. I shared caretaking duties with non-English speaking relatives. Doctor visits — both here and in Korea — had to be bilingual, which often meant appointments were longer, more stressful, expensive, and full of misunderstandings. Oftentimes, I’d want to connect with my stepmom or aunt, both to coordinate care and vent about things only we could understand. None of us could go beyond “I’m sad,” “I come Monday, you go Tuesday,” or “I’m sorry.” We struggled alone, together.

If the Humane AI Pin’s translation features worked as demoed, how much grief and loneliness might have been spared? Would our burden have felt even a smidge lighter? I’ve kept rerunning those fantasies in my head since rewatching the demo. And I know that, for the foreseeable future, it’ll stay a fantasy.

Real-time translation is so hard for so many reasons. Some may be beyond Humane’s current ability to fix.
Photo by Amelia Holowaty Krales / The Verge

Better versions of Humane’s translation tech already exist. Google Translate requires screens, but it’s a widely known tool that can act as a digital interpreter. Even so, Google Translate still struggles to keep up with all the ways language evolves. It now knows that “ㅋㅋ” is how Koreans text lol, but it completely fumbles with the Japanese idiom tsutsumotase. What you get is a direct but incorrect translation of the characters, along with an approximate definition. (The word refers to a specific type of badger game where a man and woman team up to financially extort another man.) Not to mention, English may be the lingua franca that unites tech, but some languages are easier to translate than others. It could very well be that Humane opted for Spanish and French demos because they’re much more closely related to English. Perhaps it simply didn’t have the same resources to actually build out all the languages of the world  — and the myriad permutations that would entail.  

Tsutsumotase doesn’t mean beauty bureau. Confidence person gets close to the meaning, but it lacks the actual nuance of the phrase.
Screenshot: Victoria Song / The Verge

But those are the finer details of mastery and fluency. You need much less to “survive” in another language. That’s where Google Translate excels. It’s handy when you’re traveling and need basic help, like directions or ordering food. But life is lived in moments more complicated than simple transactions with strangers. When I decided to pull off my mom’s oxygen mask — the only machine keeping her alive — I used my crappy pidgin to tell my family it was time to say goodbye. I could’ve never pulled out Google Translate for that. We all grieved once my mom passed, peacefully, in her living room. My limited Korean just meant I couldn’t partake in much of the communal comfort. Would I have really tapped a pin in such a heavy moment to understand what my aunt was wailing when I knew the why? 

Translation is an art, and art is something AI often gets wrong

Translation is an art, and art is something AI often gets wrong. It’s not enough to spit out a direct meaning. For high-context languages like Japanese and Korean, you also have to be able to translate what isn’t said — like tone and relationships between speakers — to really understand what’s being conveyed. If a Korean person asks you your age, they’re not being rude. It literally determines how they should speak to you. In Japanese, the word daijoubu can mean “That’s okay,” “Are you okay?” “I’m fine,” “Yes,” “No, thank you,” “Everything’s going to be okay,” and “Don’t worry” depending on how it’s said. (See: this ball of rice explaining it.) It’s confusing enough for humans to get it right — how are AI translation gadgets trained by imperfect humans supposed to? 

Even so, I can’t help but long for the future Humane demoed. I can study Japanese and Korean for the rest of my life — and I will — but there’ll always be gaps. I have countless memories of times when I forgot how to speak my second and third languages. Times when I was in physical pain, nervous, or had to do math. (I guarantee you, everyone does math in their native tongue.) In those moments, it’d be nice to have a simple, seamless way to ask for help. And to be understood. 

Related:

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top