relieve the pain, ease the pain, ease the pain, ease the pain, ease the pain, make me less susceptible to the flu. In Spanish, depending on the country, there are at least 12 different ways to indicate that you have the flu.
“Cardiac arrest” is also difficult to translate into Spanish, because “arresto” means to be detained by the police. Similarly, “intoxicado” does not mean drunk, but rather food poisoning.
There are countless examples of how translation doesn't work in any language: words take on new meanings, idioms emerge and die, and communities adopt slang and dialects into their daily lives.
While human translators are working hard to keep up with the changes, California will soon hand that responsibility over to technology.
State health policymakers want to use emerging artificial intelligence technology to translate a wide range of documents and websites related to “health and human services information, programs, benefits and services,” according to state records. Sami Gallegos, a spokeswoman for the California Department of Health and Human Services, declined to provide details about which documents or languages ​​would be affected, saying the information is “confidential.”
The agency is inviting bids from technology companies for the ambitious effort, the timing and cost of which haven't been disclosed yet. A human editor will oversee the project, he said, overseeing and editing the translations.
Agency officials said they hope to save money and provide important medical forms, applications, websites and other information to more people in what is described as the nation's most linguistically diverse state.
The project will start with translating documents, and Commissioner Mark Ghaly said if the technology is successful it could have broader applications.
“How can we use AI to transform not just all of our documents, but our websites, our conversational capabilities, some of our call center inputs?” Ghaly asked during a briefing on AI in healthcare in Sacramento in April.
But some translators and scholars worry that the technology lacks the nuance of human interaction and is ill-prepared for the challenge: Leaving this delicate task to machines could lead to errors in language and understanding, ultimately making information less accurate and accessible to patients, they say.
“AI can't replace human compassion, empathy, transparency, meaningful gestures and tone,” says Rithy Lim, a Fresno-based medical and legal interpreter for 30 years who specializes in Khmer, Cambodia's main language.
Artificial intelligence is the science of designing computers that mimic human thought through reasoning, problem solving, and language understanding. In one type of artificial intelligence called generative AI (GenAI), computers are trained with large amounts of data to “learn” the meaning of things and respond to prompts. There has been a wave of investment led by companies like Open AI and Google.
AI is being rapidly adopted into healthcare, including programs to diagnose diabetic retinopathy, analyze mammograms, and connect patients with nurses remotely. Proponents of the technology often make hyperbolic claims that soon everyone will have their own personal “AI doctor.”
AI has also been a game changer in the field of translation. ChatGPT, Google's neural machine translation, open source, is not only faster than older technologies such as Google Translate, but it can also process vast amounts of content and leverage a huge word database to nearly mimic human translation.
A professional translator may take three hours to translate a 1,600-word document, but AI can do it in one minute.
Arjun “Raj” Manraj, an assistant professor of biomedical informatics at Harvard Medical School and associate editor of the New England Journal of Medicine AI, said the use of AI technology is a natural progression in medical translation, given that patients already use Google Translate and AI platforms to translate for themselves and their families.
“Patients aren't waiting,” he said.
He said GenAI could be particularly useful in this context.
These translations “can bring real value to patients by simplifying complex medical information and making it more accessible,” he said.
In bidding documents, the state said the project's goal is to “increase the speed, efficiency and consistency of translation, creating improved language access” in a state where one in three people speaks a language other than English and more than 200 languages ​​are spoken.
In May 2023, the state Department of Health and Human Services adopted a “Language Access Policy” requiring departments to translate all “critical” documents into at least the top five languages ​​spoken by Californians with limited English proficiency. At the time, those languages ​​were Spanish, Chinese, Tagalog, Vietnamese and Korean.
Examples of important documents include applications for state programs, notices of benefit eligibility, and public website content.
Currently, these translations are done by human translators. With AI, more documents can be translated into more languages.
A survey conducted late last year by the California Health Care Foundation found that 30% of Spanish speakers have trouble explaining health problems or concerns to a doctor, compared with 16% of English speakers.
Health equity advocates argue that AI can help close that gap.
“This technology is an incredibly powerful tool in the field of language access,” said Sandra R. Hernández, president and CEO of the foundation. “If managed properly, there are many opportunities to expand translation capacity to address inequities.”
But Hernandez cautioned that AI translation requires human oversight to truly capture meaning.
“The human interface is really important to ensure accuracy and cultural nuance is reflected,” she said.
Lim recalled one time when a patient's daughter read out the pre-op instructions to her mother the night before surgery: Instead of translating the instructions as “don't eat after a certain time,” the daughter told her mother, “it's better not to eat.”
The surgery had to be rescheduled because the mother had eaten breakfast.
“Even a few words that change meaning can dramatically affect how people perceive information,” said Sejin Paik, a doctoral student studying digital journalism, human-computer interaction and emerging media at Boston University.
Paik, who grew up speaking Korean, also noted that AI models are often trained from a Western perspective. The data that drives translation filters language from an English perspective, “which can lead to misinterpretations of other languages,” he said. In a rapidly changing landscape, “we need to bring in more diverse voices and have more people thinking about ethical concepts to how to best predict the impact of this technology,” he said.
Manray noted that the emerging technology has other flaws that must be addressed. For example, AI can invent sentences or phrases that aren't in the original text, potentially creating false information, something AI scientists call “hallucinations” or “confabulations.”
Chinh Wong, executive director of the Vietnamese Community Health Promotion Project at the University of California, San Francisco, has been translating health-related content from English into Vietnamese and Chinese for 30 years.
He gave examples of language nuances that can confuse AI translation programs, such as how breast cancer is called “chest cancer” in Chinese, he said.
Additionally, in Vietnamese, “you” can have different meanings depending on one's status in a family or community, and doctors using “you” incorrectly with patients can be offensive, Wong said.
But Mr Ghaly stressed that the opportunities outweigh the drawbacks, saying governments should “nurture innovation” and help vulnerable people get better access to care and resources.
And he made it very clear: “We're not going to replace humans.”
This story was produced by KFF Health News, which publishes California Healthline, an editorially independent service of the California Healthcare Foundation.
This article was reprinted from khn.org, a national newsroom and one of KFF's core operating programs that produces in-depth journalism on health issues and is an independent source of health policy research, polling and journalism.
