More than 70 million Americans live with hearing loss, and texting has increasingly replaced face-to-face signing in daily life — a shift the deaf community has watched with unease. In Georgia, a startup called SeaMe Systems built a kiosk meant to address that gap, letting users sign their way through everyday interactions instead of typing them out. But the company's founder says he deliberately kept the most advanced artificial intelligence out of the final product after hearing from the very community it was built to serve.
Randy Dean, founder and CEO of SeaMe Systems, developed the kiosk alongside his team to help deaf people navigate and communicate with the world around them, according to Atlanta News First. The kiosks let a user sign to a person displayed on a screen, who then verbally relays what was signed to a hearing person nearby, Dean said. They can also interpret ASL to give directions, offering a way to get around without relying on text. Per the company's own site, SeaMe Systems says its kiosks recognize common signed questions and respond instantly with filmed ASL video answers, without needing a live interpreter for routine requests.
Why Dean Scaled Back the AI
Dean said his company remains open to AI-driven communication tools in general, but pushback from the deaf community led him to change course on just how much AI the kiosks would actually use. What won't be found in the kiosks, Dean said, is advanced AI. Instead, he said SeaMe Systems prefers to lean into actual people to provide that type of service, so users feel the help is organic rather than synthetic.
That choice tracks with broader skepticism documented well beyond Georgia. A survey from Northeastern University found that many Deaf and Hard-of-Hearing respondents worldwide were wary of sign-language technology, and researchers noted that such tools have often been built without any collaboration with the Deaf community itself. The university's reporting also pointed out that the effectiveness of any AI sign-language software depends entirely on the data used to train it — a limitation that doesn't disappear just because the underlying tech improves.
A First Language, Not a Workaround
Jimmy Peterson, executive director of the Georgia Center for the Deaf and Hard of Hearing and a deaf person himself, said ASL is the deaf community's first language, with English serving as its second. Peterson has also said texting has begun replacing signing in everyday life, a trend the community dislikes, and that the deaf and hard-of-hearing community generally prefers human interaction over AI when communicating with others, per Atlanta News First's reporting.
Still, Peterson believes AI use is ultimately inevitable, and he's hoping it makes life easier rather than harder for deaf people as it spreads. He's also said greater study of these tools over time may make people more comfortable with them. That cautious optimism echoes what Northeastern's survey found among people with limited access to interpreters, who were more open to AI tools because they offer independence and support in situations where human interpreters simply aren't available.
What Other Researchers Are Building
SeaMe Systems isn't the only outfit chasing this problem. Sorenson Communications has unveiled two AI-powered tools designed to translate ASL in real time without a human interpreter, according to NBC Boston. One converts signing into English text; the other turns written text, speech or video into ASL using a lifelike digital avatar. The company expects to bring those tools to market within the next few weeks and plans to keep refining them with user feedback, the outlet reported. Even so, NBC Boston noted that experts still question how well AI can capture the nuance of sign language, including facial expressions that carry real grammatical weight in ASL.
Google DeepMind has taken its own swing at the problem, introducing a sign-language-to-text model called SL2T that now powers sign-to-text dictation in Gboard and Live Transcribe on the Pixel 11, starting with ASL-to-English translation. DeepMind has acknowledged that sign language translation presents core challenges, including the fact that signed languages are independent natural languages and that whole-body movement, not just hand shape, carries meaning. Meanwhile, researchers at Florida Atlantic University built a real-time ASL interpretation system using object detection and hand-tracking tools that reportedly achieved 98.2% accuracy with minimal latency in a study published in the journal Sensors. Gallaudet University has gone further still, standing up an Artificial Intelligence, Accessibility and Sign Language Center aimed at developing AI that can accurately interpret sign language and bridge gaps between deaf individuals and people who don't sign.
Hearing Aids Show Promise, But Questions Remain
AI isn't just reshaping sign-language tools — it's also showing up in hearing aids. The International Journal of Audiology published a study, cited by Atlanta News First, reporting that AI-enabled hearing aids outperformed traditional aids in 99 of 100 trials. A separate blinded, randomized-controlled trial of a Spatial AI hearing aid, detailed on Taylor & Francis Online, compared the device against five control hearing aids across 100 trials involving 20 adults with bilateral sensorineural hearing loss — finding the investigational device was preferred in 99 of those 100 trials, with one tie.
None of this settles the underlying debate. A lecturer quoted in the Northeastern University survey argued that sign-language technology should supplement, not replace, real-life interpreters, particularly in medical and legal settings where precision matters most. For now, federal guidance from the Americans with Disabilities Act continues to identify accommodations like qualified sign-language interpreters, real-time captioning and notetakers as standard options for effective communication — a reminder that, AI advances aside, the tools organizations are legally expected to offer haven't changed.
