Google announced that it has developed LearnLM, a new family of generative AI models that are “fine-tuned” for learning.
The LearnLM model, a collaboration between Google's DeepMind AI research division and Google Research, is built on Google's Gemini model and is designed to “conversationally” guide students on a variety of topics. says Google.
LearnLM is already powering features across Google products, including YouTube, Google's Gemini app, Google Search, and Google Classroom.
“LearnLM is based on educational research and makes the learning experience more personal and engaging,” said James Manyika, SVP of Research, Technology, and Society at Google Research, at Google's developer conference I/N on Tuesday. He said this on stage at the O 2024 keynote speech.
Google said it is working with educators through its Google Classroom pilot program to explore how LearnLM can simplify and improve the lesson planning process. Google says LearnLM can help teachers discover new ideas, content, and activities, and find materials tailored to the needs of specific student populations.
Image credit: Google
Elsewhere, LearnLM is powering Circle to Search on Android. This feature helps you solve basic math and physics problems, and you can quickly understand problems such as symbolic equations, diagrams, and graphs. LearnLM also powers tools on YouTube (currently Android only in the US) that allow users watching academic videos to ask clarifying questions, receive clarification, and take quizzes based on what they're watching. You will be able to answer.
Coming to Google's Gemini app in the coming months, LearnLM will allow users to create custom chatbots that act as subject matter experts. Google says these chatbots will provide study guidance, provide practice activities such as quizzes and games, and respect each learner's personal preferences.
Google also plans to partner with organizations such as Columbia Teachers College, Arizona State University, NYU Tisch, and Khan Academy to explore how LearnLM can be expanded beyond its own products.
“Today begins a new chapter for learning and education at Google,” Manika said. “Generative AI unlocks new ways to make the world's information and knowledge universally accessible and useful.”
Indeed, LearnLM is an interesting job. However, technical documentation detailing the development of this model reveals that it suffers from some of the same issues as other generative AI models.
For example, LearnLM doesn't speak in a very “encouraging” tone. Also, compared to his one in the vanilla Gemini model (Gemini 1.0), it is harder to identify when a student answered a practice question correctly, but it is better at spotting mistakes. And they don't hallucinate or make up facts and figures in response to stimuli presented to them.
For these reasons, the paper's co-authors caution against using LearnLM in apps without further evaluation and analysis of app-specific harms. [the apps]” – At the same time, it suggests that Google has performed such evaluation and analysis on its own apps. Let us pray for the sake of students and educators that this is true.