Jeenius CEO: AI's Next Frontier Is the World It Can't Yet Reach
AI "knows" hundreds of languages. So why do Apple's AirPods only offer interpreting in nine? Jeenius tested the leading AI models and the results will surprise you.
WASHINGTON, Nov. 5, 2026 /PRNewswire/ -- While developing its AI interpreter, Jeenius tested leading AI models for speech-to-speech interpreting in healthcare settings. The discovery: only 20 languages met its accuracy threshold. These results reveal a stark gap between AI's promise, the public's perception of its capabilities, and its actual readiness for real-world interactions.
In her recent Washington Business Journal article, CEO and Co-Founder Kirsten Brecht Baker argues the shortfall isn't a modeling failure. It's a data failure.
AI has a lot to learn about conversation
Baker stated: "Imagine a busy emergency department: television blaring, multiple conversations happening, a frightened patient explaining what hurts in a foreign language. AI must distinguish between speakers, discern intention, transfer into another language with cultural nuance, and respond naturally." That requires data most models lack. "Text trains AI to do tasks," she adds. "To engage with humans, AI must be trained on authentic audio and video data where people are speaking and responding naturally."
The data void
Frontier models have already scraped the same public internet. An estimated 90% of AI training data is English, drawn largely from North America and Europe. Only 2% of training data comes from the African continent. Baker states: "the data needed by AI isn't online waiting to be found. It has to be created."
Opportunity: data that looks like the world
"We didn't come to multilingual AI from the outside. We've spent years building a community of tens of thousands of qualified linguists across hundreds of languages," Baker says. Its Experts-in-the-Loop create authentic data, build domain-specific datasets, and benchmark model accuracy for labs and AI companies.
She concludes: "The winning AI won't be the biggest model. It will have the most reach: engaging a patient in Guatemala, a customer in Lagos, a family in Kyiv. The real test of intelligence will not be what it can do, but who it can do it for."
Jeenius builds the multilingual data infrastructure AI needs to communicate with the world. Its Experts-in-the-Loop network of tens of thousands of qualified linguists creates authentic conversational data, domain-specific datasets, and model accuracy benchmarking for developers of LLMs and SLMs. jeeniusAI.com.
Media: jeeniusai.com/contact
SOURCE Jeenius
Share this article