AI Learns Hmong A Surprising Linguistic Leap
At a recent digital marketing workshop for Hmong business owners in Maplewood, tech consultants May Yang-Her and her husband Dao Her witnessed a surge of excitement. When May asked how many had used AI for translation, nearly everyone raised their hand, sharing stories of ChatGPT and other AI tools accurately handling the Hmong language. "It has come a long way," May remarked. "The word choices they use are fascinating." Dao added, "Sometimes I think ChatGPT knows more Hmong than I do."
AI's Linguistic Evolution Beyond English
The core of artificial intelligence lies in its ability to absorb and manipulate language. Until recently, Large Language Models (LLMs), the powerhouses of AI, primarily focused on English, often sidelining less common languages. This poses a particular challenge for languages like Hmong, which was predominantly an oral language until the 1960s and remains spoken-only for many native speakers.
"We want these systems to be able to know our language, for our elders to use, and for future generations to still be able to learn our language," stated Mai Lee Chang, an AI and robotics researcher, during a panel at the Hmong National Development conference. The imbalance in language and cultural representation is a known issue in technology. English still dominates over half of all websites, despite only 5% of the world's population speaking it as a first language. While this imbalance is currently more pronounced in AI, it is expected to diminish.
The Shifting Landscape of AI and Language
AI is currently in a developmental phase where public perception oscillates between apprehension and excitement, much like the internet thirty years ago. AI, similar to the internet, is evolving into a utility. Its inherent weaknesses, including the risk of cultural imperialism, are being identified and actively addressed. For instance, the past year has seen significant advancements in small language models (SLMs). Unlike LLMs, SLMs use smaller datasets and can be tailored for specific tasks, such as language translation.
Hmong Tech Leaders Championing AI Adoption
May Yang-Her and Dao Her, owners of DaoTech Solutions, a St. Paul-based tech consulting firm, are closely observing these changes in both their professional and personal lives. They provide tech integration, networking, and other services to small businesses and charter schools. Both actively promote AI to their clients and within the Hmong community. They've utilized AI for diverse tasks, from helping school clients with lesson plans to diagnosing plumbing issues at their cabin. Their children also leverage AI: their teenage son creates podcasts for philosophy studies, and their daughter uses it to condense her nursing class notes. May, who participated in the Hmong conference's AI panel, predicted that future evaluations of individuals will focus less on degrees and more on AI-assisted problem-solving abilities.
Hmong Language A Global Presence Meets Local Innovation
The Hmong language is spoken by approximately 12 million people worldwide, primarily in China. A significant diaspora fled Laos after assisting Americans during the Vietnam War, settling in Thailand, France, and the U.S. Minnesota is home to one of the largest Hmong populations in America, numbering nearly 100,000. This year, the state celebrates the 50th anniversary of Hmong arrival.
"With the 50th anniversary this year, there’s been a lot of looking back and talk about where we’ve been," May shared. "The next 50 years is going to be so different." A trip to Thailand last year highlighted how the Thai language influenced Hmong spoken there. "They were using Hmong words we didn’t know," Dao recalled. This experience led the Hers to recognize AI's potential in navigating Hmong language variations and guiding its future.
Real World Impact AI's Growing Hmong Proficiency
May acknowledges that ChatGPT doesn't yet grasp all "the nuances and the culture and the different dialects" of Hmong. "But they’ve gotten a lot better with interpretations," she noted. Bon Xiong, a student in the Hers' class who works in industrial systems sales, has tracked this improvement. He frequently writes documents in English and uses ChatGPT for Hmong translation. "I noticed it in the last couple of months," Xiong said. "I spent half an hour trying to translate myself and then I thought, ‘What am I doing? Let me just plug this into ChatGPT.’ And it came back so fast, and I was like, whoa. I was impressed. And she read it too and she said, ‘Did you write this?’"
The Road Ahead Challenges and Hopes for Hmong AI
Despite these advancements, AI models still struggle with pronouncing Hmong words, a significant challenge given the language's seven or eight tones, depending on the dialect. "Right now, we’re just looking at text," May explained. "I would like to see it do text-to-talk for our elders who can’t read English or can’t read Hmong."
