The Reality Check Education AI Needed

Against the glittering backdrop of Davos, where tech optimism usually reigns supreme, Squirrel AI co-founder Liang Jing delivered a sobering message about artificial intelligence in education. "We're seeing dangerous bubbles forming," she warned during a World Economic Forum dialogue that cut through the usual hype.

Beyond the Hype Cycle

The education technology market has become flooded with products boasting AI capabilities, but Liang sees most as little more than repackaged large language models forced into teaching roles. "These systems can produce smooth answers," she observed, "but they're like teaching assistants who never graduated beyond textbook recitation."

The fundamental issue? Current implementations often miss what makes education work. "Teaching isn't about delivering perfect responses," Liang explained. "It's about understanding why a student struggles with quadratic equations today when they mastered similar problems yesterday."

The Precision Paradox

What sets truly effective educational AI apart, according to Liang, is vertical specialization - deep understanding within specific subject areas. Generic models might handle broad conversations well but falter when pedagogical precision matters most.

"We've spent years accumulating millions of data points on how students actually learn math concepts," Liang shared, describing Squirrel AI's approach. "Which misconceptions typically arise? What explanations work best for different learning styles? This granular understanding transforms AI from an answer machine into a learning partner."

The Davos discussion highlighted growing concerns that the education sector's rush to adopt flashy new technologies might be overlooking what actually helps students learn. As schools worldwide increase investments in AI tools, Liang's critique serves as an important reality check.

Key Points:

• Vertical specialization beats general knowledge: Effective educational AI requires deep subject expertise
• Data quality matters more than model size: Millions of precise learning interactions beat billions of web pages
• Education isn't information delivery: True teaching involves diagnosing misunderstandings and adapting explanations