LearnLM-Tutor is a generative AI model developed by Google DeepMind, specifically used in the field of education and aims to provide one-on-one conversation coaching.
Through features such as instant feedback, multi-round dialogue, error identification and active learning facilitation, the model not only helps students solve specific problems, but also cultivates their autonomous learning and critical thinking skills. Combined with customized learning plans, multidisciplinary support and progress tracking, LearnLM-Tutor is a powerful educational tool that supports students ‘well-rounded development.
Google recently shared a very detailed and interesting research paper on the development of generative artificial intelligence for education.
The paper outlines an evaluation-driven approach to creating artificial intelligence mentors that can significantly enhance students ‘learning experiences and provide assistance to teachers.
Generative artificial intelligence will disrupt education by providing personalized tutoring to every student and teaching assistance to every teacher.
The future of generative artificial intelligence in education
Generative artificial intelligence will disrupt education by providing personalized tutoring to every student and teaching assistance to every teacher.
However, this potential has not yet been fully realized.
The main challenge is to transform teaching instincts into artificial intelligence prompts and develop evaluation practices.
Google DeepMind held workshops with students and teachers to understand their needs and challenges.
Students expressed appreciation for the instant feedback and anonymity provided by the artificial intelligence instructor.
On the other hand, teachers worry about the possibility of cheating and the ability of artificial intelligence to adapt to individual student levels.
A well-known project is ShiffBot, where Google worked closely with an educator to develop an artificial intelligence mentor that seamlessly integrates into existing learning environments.
The project provides valuable insights into the specific needs and preferences of students and teachers.
To improve the teaching capabilities of the Gemini model, Google created several fine-tuning datasets.
These include real conversations between students and teachers, role-playing data, and specially written conversations.
Each dataset focuses on different teaching aspects, such as identifying errors and providing useful feedback.
Google has developed seven benchmarks to evaluate the performance of artificial intelligence instructors.
These include manual evaluation and automatic evaluation.
For example, a human evaluator evaluates the quality of a mentor’s answers, while an automated evaluator evaluates specific tasks, such as identifying errors.
Students used both methods;
- LearnLM-Tutor
- Prompt-tuned Gemini model.
They rated LearnLM-Tutor higher in most categories, especially in terms of confidence in what they learned.
Expert teachers also reviewed these conversations and found that LearnLM-Tutor was more able to engage students and spot mistakes.
Teachers provided detailed feedback on the mentor’s answers.
LearnLM-Tutor is rated better than Gemini 1.0 in promoting active learning and providing constructive feedback.
However, it can be improved in terms of more encouraging and recognizing student success.
Google uses Large Language Models (LLMs) as critics to automatically evaluate mentors ‘responses.
LearnLM-Tutor is superior to Gemini 1.0 in several key areas, such as attracting students and identifying misunderstandings.
Google has developed specific metrics to evaluate how well instructors handle practice sessions.
LearnLM-Tutor has shown significant improvements in maintaining the conversation process, providing high-quality feedback, and generating appropriately difficult questions.
To understand how students can use LearnLM-Tutor in a real academic environment, Google partnered with Arizona State University (ASU).
They integrated LearnLM-Tutor into Arizona State University’s study room project.
Students and teachers gave positive feedback. Students love instant response and personalized help. However, some people think it is necessary to carefully examine the answers of mentors.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
This article comes from:https://www.linkedin.com/pulse/development-generative-ai-gen-education-learnlm-tutor-muratcan-koylan-6kp7c/
Oil tubing: