Google says it has developed a new family of generative AI models “optimized” for learning: LearnLM.
A collaboration between Google’s DeepMind AI research division and Google Research, LearnLM’s models — built on top of Google’s Gemini models — are designed to “conversationally” teach students across a range of subjects, Google says.
LearnLM already powers features across all Google products, including YouTube, Google Gemini apps, Google Search, and Google Classroom.
“LearnLM is based on educational research, making learning experiences more personal and engaging,” said James Manyika, vice president of research, technology and society at Google Research, said on stage during a keynote address at Google’s I/O 2024 developer conference on Tuesday.
Google says that, through a pilot program in Google Classroom, it is working with educators to see how LearnLM can simplify and improve the lesson planning process. LearnLM could help teachers discover new ideas, content and activities, Google says, or find materials tailored to the needs of specific groups of students.
Elsewhere, LearnLM powers Circle to Search on Android, a feature that helps solve basic math and physics problems — and that will soon understand problems including symbolic formulas, diagrams, and graphs. And LearnLM supports a tool on YouTube (US Android only for now) that lets users watching academic videos ask clarifying questions, get explanations, or take a quiz based on what they’re watching.
In Google’s Gemini apps in the coming months, LearnLM will allow users to create custom chatbots that can act as subject matter experts. These chatbots will provide study guidance and exercises, such as quizzes and games, and will respect each student’s individual preferences, Google says.
Google also plans to work with organizations such as Columbia Teachers College, Arizona State University, NYU Tisch and Khan Academy to see how LearnLM can expand beyond its own products.
“Today marks a new chapter for learning and education at Google,” said Manyika. “Generative AI is unlocking new ways to make the world’s information and knowledge globally accessible and useful.”
LearnLM is definitely an interesting job. But a technical paper in detail the development of the model reveals that it suffers from some of the same problems as other artificial intelligence production models.
For example, LearnLM doesn’t speak in a very “encouraging” tone and has trouble recognizing when students are answering the practice questions correctly compared to one of the vanilla Gemini models (Gemini 1.0), although it is better at detecting mistakes. And it is not immune to hallucinations, nor to making up facts and figures in response to the prompts it feeds.
For these reasons, the paper’s co-authors caution against using LearnLM in applications “without further evaluation and analysis of the vulnerabilities associated with [the apps]” — while in the same breath suggesting that Google conducted such assessments and analyzes of its apps. For the sake of students and teachers, let’s hope this is true.
We’re launching an AI newsletter! Sign up here to start getting it in your inbox on June 5th.
