Google says that he has developed a new family of “refined” generative AI models for learning: Learnlm.
A collaboration between the DEEPMIND research division of Google and Google Research, the Learllm models – built above the Gemini models of Google – are designed for students of tutor “Conversation” on a range of materials, known as Google .
Learnlm already offers features on Google Products, including in YouTube, the Gemini applications of Google, Google Search and Google Classroom.
“Learnlm is based on educational research, making learning experiences more personal and engaging,”Manyika souls, please from Research, Technology and Society at Google Research, said on stage during a speech Tuesday, Google I / O 2024 conference Developer.
Google says that, thanks to a pilot program in Google Classroom, he works with educators to see how LearnLM could simplify and improve the lesson planning process. Learnlm could help teachers discover new ideas, content and activities, known as Google, or find equipment adapted to the needs of specific students cohorts.


Elsewhere, Learnlm offers Circle to search for AndroidA functionality that helps to solve mathematics and basic physics problems – and that will soon include problems, including symbolic formulas, diagrams and graphics. And Learnlm underpins A tool on YouTube (Only on Android in the United States for the moment) which allows users to watch academic videos ask clarification questions, get explanations or take a quiz according to what they are looking at.
In Google’s Gemini applications in the coming months, LearnLM will allow users to create personalized chatbots that can act as experts in terms. These chatbots will provide study advice and practice activities such as quizs and games, and they will respect the individual preferences of each learner, says Google.
Google also plans to join organizations such as Columbia Teachers College, Arizona State University, Nyu Tisch and Khan Academy to see how LearnLM can be extended beyond its own products.
“Today marks a new chapter for learning and education at Google,” said Manyika. “The generative AI unlocks new ways for us to make information and knowledge of the world universally accessible and useful.”
Learnlm is an intriguing work, of course. But a technical document Details of the development of the model reveals that it suffers from some of the same problems as other generative AI models.
For example, Learnlm does not speak in a very “encouraging” tone and has trouble identifying themselves when the students respond correctly to the questions of practice in relation to one of the gemini vanilla models (Gemini 1.0), although It is better to spot errors. And it is not immune to hallucinations, or manufacture of facts and figures in response to the prompts it is powered.
For these reasons, the co -authors of the prudent article against the use of learning in applications “without further evaluation and analysis of damage specific to (applications)” – while suggesting that Google has carried out such assessments and Analysis for its applications. For sakes of students and educators, let’s hope it is true.
We are launching an Ai newsletter! Register here To start receiving it in your reception boxes on June 5.