Trending Hashtags
Alerts
Always receive notifications to stay updated on new posts & activities on your content and account on Social Network.
-
Last week, ahead of its Made by Google event, Google DeepMind announced SL2T, an AI model that translates sign language directly into written text. The technology powers a new sign-to-text feature in Gboard and Live Transcribe across the Pixel 11 lineup. For American Sign Language (ASL) users, this means they can now sign into their phone’s camera anywhere they would normally type—whether that is a search bar, a message, a document, or a prompt to Gemini. In Live Transcribe, a Deaf user can sign a reply during a face-to-face conversation instead of thumbing out a response while the other person waits.
For anyone who has spent a career watching accessibility technology get announced, underdelivered, and sometimes shelved, this announcement deserves a close look. It is great work, and perhaps even revolutionary.
Translating ASL is a fundamentally different problem than transcribing speech. ASL is a full language with its own grammar, and meaning is carried simultaneously through the hands, arms, torso, head, and face. There is no clean word-for-word mapping to English.
Previous approaches tried to solve this problem by first converting signs into written labels called glosses, and then translating the glosses. Google’s team abandoned that step entirely. Their argument is that glosses flatten exactly the non-manual markers and spatial constructions that carry meaning in ASL. Instead, an on-device computer vision system tracks 130 points across the signer’s face, body, and hands, converting the video into a moving map of coordinates. The raw footage is discarded on the device; only the coordinates travel to Google’s servers for translation.
The model was trained on more than 100,000 hours of data spanning over 50 sign languages, with roughly a quarter of that data in ASL. It was deliberately trained to handle one-handed signing, which matters when you are holding the phone with your other hand, and to handle left-handed signers.
The concept came from Sam Sepah, a Deaf Googler. Google convened an AI Sign Language Advisory Committee composed of Deaf organizations and subject-matter experts, and ran evaluation through Deaf user studies.
To understand why the technical choice to track the face and torso matters—and why the governance around this launch matters more—you have to know what has been done to this language before.
ASL was born in a school. In 1817, Thomas Hopkins Gallaudet opened the American School for the Deaf in Hartford, Connecticut, alongside Laurent Clerc, a Deaf French teacher he had recruited in Paris. What emerged there was a blend of French Sign Language, the home signs students brought from their own families, and the signing already in use in communities like Martha’s Vineyard, where hereditary deafness was common enough that hearing and Deaf residents alike signed as a matter of course. Within a generation, that blend had become a distinct language with its own grammar, transmitted through Deaf schools, Deaf clubs, and Deaf families across the country.
Then, in September 1880, a room in Milan voted it out of existence.
The turning point came in 1880 at the Second International Congress on the Education of the Deaf, held in Milan. Organized by the Pereire Society, a group that championed oral education, the event was carefully curated to produce a specific outcome. The guest list was stacked, and the results reflected that imbalance. Of the twelve speakers who addressed the Congress, nine argued in favor of oral-only instruction, which prohibited the use of sign language in favor of speech and lip-reading.
The Congress ultimately passed a series of resolutions declaring that the oral method should be the preferred approach and that sign language should be removed from Deaf education. The vote was not close. Of the 164 delegates present, only the five Americans and a single English delegate voted against the measure. By most historical accounts, there were just one or two Deaf people in the entire room.
The consequences of the Milan vote were swift and severe, particularly in the United States. The share of Deaf students taught through oral methods jumped from 32 percent in 1887 to 70 percent by 1917. Deaf teachers were systematically pushed out of the profession, replaced by hearing speech instructors who had little to no experience with Deaf culture or communication. Children were punished for signing, and an entire pedagogical system was built on the premise that a Deaf child’s native language was an obstacle to be removed at all costs.
Despite this institutional pressure, American Sign Language (ASL) did not die. It survived in exactly the places the authorities were not looking. It passed between children in dormitories after lights went out. It was kept alive in Deaf clubs, at Deaf sporting events, and around kitchen tables in Deaf families. For eighty years, ASL was transmitted peer to peer while every official structure insisted it was not a language at all.
The vindication came in 1960, when William Stokoe, a hearing linguist at Gallaudet College, published Sign Language Structure. In that landmark paper, Stokoe demonstrated that ASL possessed phonology, morphology, and syntax, the same structural components found in any other natural language. That paper is the reason it is now widely understood that ASL is not simply English performed with the hands. It was not obvious then, and Stokoe was widely mocked by his own colleagues for making the claim.
What followed was progress that came far slower than it should have. Gallaudet, which was chartered by Abraham Lincoln in 1864, did not have a Deaf president until students shut down the campus in the Deaf President Now protests of March 1988. The Americans with Disabilities Act followed in 1990, providing broader civil rights protections. And the International Congress on the Education of the Deaf did not formally repudiate the Milan resolutions until it met in Vancouver in 2010, a full 130 years after the original vote.
Technology has been part of this story throughout, but often as a series of well-meaning misunderstandings. The most instructive example is the sign language glove, an idea that resurfaces every few years in university engineering departments and crowdfunding campaigns, usually accompanied by a press release about breaking down barriers. Deaf linguists have explained each time why the concept cannot work. A glove captures the hands and nothing else, yet ASL is built out of facial expression, head position, eye gaze, body shift, and the use of space. A glove that reads only the hands is reading roughly a third of the language and confidently reporting the result as a complete translation.
The shift in Google’s approach is more than a technical adjustment. By choosing to track 130 distinct landmarks across the face, torso, and hands, the company has built its architecture around a principle Deaf linguists have championed for decades: American Sign Language is a complete, complex language. This is the first time a major consumer product has been designed from the ground up with that fundamental truth in mind.
For the Deaf people who will actually use this technology, that recognition carries profound weight. Chrissy Marshall, a Deaf writer and director, shares that perspective. “This is groundbreaking technology, and as a Deaf ASL user, I’m excited to see ASL recognized computationally as the complex language it is: from handshapes and classifiers to spatial grammar, facial expression, and non-manual markers,” she says. However, she is quick to add a note of caution: “But I can’t help thinking about how this technology fits into our existing accessibility landscape.”
That accessibility landscape has long depended on the unpaid and often invisible labor of Deaf individuals. Marshall points out the daily reality for many Deaf people: “Deaf people already carry so much of the communication burden: watching captions, filling in what was missed, correcting errors, advocating for accommodations, all while trying to participate in the conversation.” She expresses frustration with a society that accepts subpar tools as the norm, stating, “As a society, I think we are far too comfortable expecting Deaf people to rely on imperfect communication tools. Especially when they are ‘futuristic’ or ‘cool.’” The excitement around new technology, in her view, should not overshadow the fact that Deaf people are often forced to make do with tools that are simply not good enough.
Google and its advisory committee are explicit about the intended scope of the Sign Language to Text (SL2T) feature. It is designed strictly as an assistive input tool for low-stakes communication. The company clearly states that the software does not and cannot fulfil legal requirements for reasonable accommodations. It is not intended for use in high-stakes environments such as medical appointments, legal proceedings, classrooms, job interviews, or government hearings. The accompanying report explicitly warns institutions against using the software as a replacement for qualified human interpreters or as a way to bypass their legal accessibility obligations. This distinction is crucial for both Deaf users and the organizations that serve them.
As with any emerging technology, the system has significant limitations. The model can struggle with regional signs, slang, complex ASL grammar, and rapid fingerspelling. Because facial expression carries grammatical meaning in ASL, the technology may also misinterpret meaning conveyed primarily through non-manual markers. It can also generate what Google calls “ghost text”—words the user never actually signed—which can occur when someone pauses or when a second person walks into the frame. Accuracy further degrades in low-light conditions, when the signer is partially out of frame, or when there is insufficient contrast between the hands, face, clothing, and background.
These technical constraints leave a significant question unanswered: whose ASL is the system being trained to understand? Marshall highlights that ASL is not a single, uniform language. “ASL isn’t monolithic,” she explains. “There are regional signs, Black ASL and other dialects, generational differences, slang, signing styles, and differences in mobility all matter.” She also raises concerns about the technology’s ability to recognize people with facial differences or disabilities that affect movement, given the grammatical importance of facial expression. Her questions are practical and pressing: “Will I have to take off my glasses? Will POC be able to communicate and feel understood equitably?” These are not hypothetical concerns but essential considerations for any tool that aims to serve a diverse and varied Deaf community.
Human interpreters engage in the same kind of negotiation, and Marshall notes that this process is never fully settled even among people. “Even with interpreters who have decades of experience, there can be linguistic and generational nuances we have to navigate together,” she says. “In the entertainment field I work in, we adapt and create signs for efficacy and flow of work.”
This reality points to the central question that should anchor any evaluation of sign language translation technology. “Will an AI model understand the way I sign,” Marshall asks, “or will I have to adapt my language to be understood by the technology?”
The answer to that question determines whether a product is genuinely translating ASL or, instead, asking ASL users to make their language more legible to a machine. A system that performs best when the signer standardizes their input toward the technology has effectively shifted the work back onto the Deaf user—a position the community has occupied since 1817.
Beyond the conceptual concerns, Marshall identifies immediate practical barriers. “The current need for specific lighting, framing, and phone positioning also makes me question how practical it will be for spontaneous interactions,” she says. “And in formal settings, I would still prefer a qualified ASL interpreter.”
These constraints matter because accessibility technology rarely stays within its intended boundaries. Automatic captions, for example, were introduced with explicit guidance that they were not a substitute for CART or professional captioning. Within a few years, however, universities and employers began treating auto-captions as sufficient compliance. Deaf students and employees have since spent years litigating and negotiating to reclaim accommodations they were entitled to from the start.
Marshall has experienced both the benefits and the drawbacks of this pattern. “I know imperfect technology can still be incredibly useful because I don’t have an interpreter with me every day,” she says. “Live transcription and captions, for example, can be used spontaneously in everyday life, and emerging caption glasses can help me maintain eye contact instead of constantly looking down at a phone.”
At the same time, she has watched the substitution happen to her personally. “I’ve already experienced automatic captions treated as ‘good enough’ in academic and professional settings where better access exists.”
There is also the variable that accessibility technology most reliably gets wrong: who can actually access it. Sign Language Translation is tied to the Pixel 11 lineup, meaning the most sophisticated sign language model ever shipped is gated behind the purchase of a specific phone.
“I also hope the future of accessible technology includes affordability and distribution,” Marshall says. “We’ve watched emerging technologies like smart glasses evolve primarily as consumer products, while many of their most meaningful possibilities are accessibility-related.”
Her conclusion is the one that product teams, funders, and policymakers should sit with. “Innovation only goes so far if the people who could benefit most can’t afford or access it.”
The arrival of this feature in a mainstream consumer product marks a notable milestone in assistive technology. Unlike earlier efforts that treated sign language as a secondary consideration, this tool was developed in collaboration with Deaf users rather than merely designed for them. That distinction matters, and it is one of the reasons the rollout has generated genuine interest within the Deaf community and among accessibility advocates.
The most striking aspect of the initiative is the development process itself. The technology was shaped with direct input from Deaf people at every stage, which is an approach that remains far too uncommon in the industry. Too often, accessibility features are engineered in isolation and then presented to the communities they are meant to serve. This project took the opposite path, prioritizing lived experience and community feedback from the outset. That collaborative foundation is what separates this launch from previous attempts. It signals a shift in how major tech companies approach accessibility, moving away from a checklist mentality and toward genuine partnership. The result is a product that reflects the real-world needs of its users rather than assumptions made on their behalf.
Equally important is the candor Google has shown about what the technology can and cannot do. The company did not overstate the feature’s abilities or present it as a complete replacement for human interpretation. Instead, it offered a straightforward explanation of the current limitations, which is a level of transparency that builds trust. This honesty is significant because sign language is complex, nuanced, and deeply tied to cultural identity. No algorithm can fully capture that richness, and pretending otherwise would have done a disservice to the Deaf community. By acknowledging the boundaries of the technology, Google has set realistic expectations while still highlighting its genuine usefulness.
The response from Deaf advocates and professionals reflects a balanced perspective. Marshall, a representative voice in the community, offered measured praise for the innovation while keeping the focus where it belongs.
“The technology is remarkable. Its greatest value should be giving Deaf people more tools to communicate, and I’m always open to that. Innovation should expand access.”
That sentiment captures the prevailing attitude among many in the Deaf community. There is genuine excitement about the possibilities the technology opens up, but there is also a clear-eyed understanding that it is one tool among many. The emphasis on expanding access, rather than replacing human connection, is a reminder of what truly matters in assistive technology.
The combination of community involvement and honest communication is rarer than it should be in the tech industry. Many products claim to be inclusive, but few actually embed that principle into their development process. This launch demonstrates what is possible when companies take accessibility seriously as a core value rather than an afterthought. For other organizations working on similar technologies, this serves as a useful model. Building with the community, being transparent about limitations, and focusing on genuine usefulness are all practices that lead to better outcomes. They also build the kind of trust that is essential when introducing technology into culturally significant spaces.
As sign language technology continues to evolve, the guiding principle should remain the same. Innovation is most valuable when it expands access and gives people more ways to communicate. It is not about replacing human interaction or diminishing the importance of sign language as a cultural cornerstone. Rather, it is about creating additional pathways for connection and inclusion. The launch of this feature is a positive step in that direction. It demonstrates that major technology companies can produce meaningful accessibility tools when they are willing to listen, learn, and collaborate. The hope is that this becomes the standard rather than the exception, and that future innovations follow the same collaborative and transparent path.
8.1K articles
1.5K articles
34 articles
28 articles
Comment