👨🏫 Free AI Tools Revolutionizing Learning
Powerful tools like Khanmigo and OpenAI's GPT-4o now available for free to users
With AI tools rapidly advancing, their capabilities are becoming increasingly powerful and widely available for free. Khan Academy's AI-powered personal tutor and teaching assistant, Khanmigo, is now accessible to all teachers at no cost. OpenAI has unveiled GPT-4o, their most advanced language model capable of speech recognition and generation, and it is also free for all users. Joining the AI education wave, Google has introduced LearnLM, a family of models fine-tuned specifically for education experiments. As these cutting-edge AI technologies become more accessible, they hold immense potential to transform learning experiences across various settings. In this edition, we will look into these new models and their possible usage within educational contexts. How do you perceive the impact of AI tools on students’ critical thinking skills?
Here is an overview of today’s newsletter:
Latest Insights on GPT-4o and LearnLM
Assessment of AI’s Impact on Students’ Problem-Solving Skills
New US Legislation to Boost AI Education
AI Chatbots Designed to Address the School Counseling Crisis
📣 Student Voices and Use Cases
Join us on May 28th for our first webinar in the AI x Education Webinar Series, where we will be featuring Eric Klopfer, a renowned professor at MIT specializing in educational technology and game-based learning. Through the discoveries from Klopfer and his students, the webinar will explore implications and strategies for integrating AI in classrooms through the lens of CS education. Register now through this link!
🚀 Practical AI Usage and Policies
OpenAI's GPT-4o and Google's LearnLM, demonstrate the great potential of AI technologies in revolutionizing education. GPT-4o, with its multimodal capabilities, can serve as a versatile virtual tutor, providing personalized instruction and feedback across various subjects and learning styles. Meanwhile, Google's LearnLM family of models aims to create more engaging and effective learning experiences by breaking down content into digestible steps, adjusting to individual needs, and incorporating interactive features. The following section features a breakdown of these two latest advancements.
OpenAI - GPT-4o
One of the most significant changes in the GPT-4o model compared to GPT 3.5 and 4 is its ability to provide more accurate explanations and feedback significantly quicker. GPT-4o can also process various input types, including text, audio, images, PDFs, and videos. Additionally, it can generate output in text, video, audio, and images, enabling it to serve as an accessible tutor for students of varying age groups and abilities. Its multimodal input-output capabilities make it a versatile educational tool catering to different learning styles and preferences.
In the video below, Sal Khan and his son Imran demonstrate how GPT-4o can be used to help students solve math problems by verbally asking questions along the way and receiving immediate individualized feedback:
For ideas on how you can maximize the capabilities of GPT-4o, check out this blog post by Richard Campbell, How To Use GPT-4-O Voice In The University EFL Classroom. With GPT-4o, students can engage in verbal role play, practice their pronunciations, and more. Below is an example Campbell uses to demonstrate how it can help students practice their grammar and syntax:
Google - LearnLM
Google is also exploring ways to integrate Generative AI in educational contexts through its new family of models: LearnLM. Based on their latest Gemini model, LearnLM aims to create more personalized, engaging, and effective learning through Google platforms like YouTube, Google Classroom, and Google Search.
One of the most unique elements of LearnLM is its multimodal presentation, allowing the model to break down content into small steps and encouraging the learning process. Along with this, the model also dynamically adjusts to individual learners’ goals and needs. Additionally, LearnLM serves as a planner and progress tracker for students, allowing them to reflect and track their performance concerning set goals.
📝 Latest Research in AI + Education
Google DeepMind
Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach ↗️
Recent advances in generative AI have sparked interest in its potential for personalized tutoring. However, challenges remain in translating pedagogical knowledge into AI systems and robust evaluation methods. This research introduces LearnLM-Tutor, an initiative to enhance gen AI for education through collaboration with educators and learners. It aims to refine AI's pedagogical capabilities by establishing seven educational benchmarks encompassing quantitative and qualitative measures, including human evaluations.
The approach involves fine-tuning an AI model with pedagogical content to create LearnLM-Tutor, which has demonstrated great performance in educational settings. The process is continually refined through feedback loops with the educational community, aligning AI's functions with real-world needs. Moving forward, the project seeks to explore methods like reinforcement learning from human feedback (RLHF) to further enhance educational effectiveness, encouraging ongoing collaboration across AI and learning science communities to develop a comprehensive framework for evaluating and improving AI in educational contexts.
Jurenka, I., Kunesch, M., McKee, K., Gillick, D., Zhu, S., Wiltberger, S., Phal, S. M., Hermann, K., Kasenberg, D., Bhoopchand, A., Anand, A., Pîslar, M., Chan, S., Wang, L., She, J., Mahmoudieh, P., Rysbek, A., Ko, W.-J., Huber, A., Wiltshire, B., Elidan, G., Rabin, R., Rubinovitz, J., Pitaru, A., McAllister, M., Wilkowski, J., Choi, D., Engelberg, R., Hackmon, L., Levin, A., Griffin, R., Sears, M., Bar, F., Mesar, M., Jabbour, M., Chaudhry, A., Cohan, J., Thiagarajan, S., Levine, N., Brown, B., Gorur, D., Grant, S., Hashimoshoni, R., Weidinger, L., Hu, J., Chen, D., Dolecki, K., Akbulut, C., Bileschi, M., Culp, L., Dong, W.-X., Marchal, N., Van Deman, K., Misra, H. B., Duah, M., Ambar, M., Caciularu, A., Lefdal, S., Summerfield, C., An, J., Kamienny, P.-A., Mohdi, A., Strinopoulous, T., Hale, A., Anderson, W., Cobo, L. C., Efron, N., Ananda, M., Mohamed, S., Heymans, M., Ghahramani, Z., Matias, Y., Gomes, B., & Ibrahim, L. (2024). Towards responsible development of generative AI for education: An evaluation-driven approach. Google DeepMind. https://goo.gle/LearnLMIEEE Conference on Artificial Intelligence (CAI) 2024
This research examines the dynamics of human-generative AI collaboration in problem-solving tasks involving 79 undergraduate students interacting with ChatGPT. Three primary collaboration types emerged: even contribution, human-led, and AI-led, with most students (77.21%) perceiving they led or contributed equally. However, 15.19% felt ChatGPT led, indicating over-reliance tendencies. Students generally positive or mixed perceptions of ChatGPT interactions correlated positively with a sense of agency—essential for effective collaborative problem-solving.
While generative AI can augment human problem-solving, strategies are needed to mitigate over-reliance, which could undermine learning and decision-making skills. Integrating AI tools like ChatGPT should enhance, not replace, human cognitive functions. Future research aims to develop methods ensuring AI interactions empower student agency and problem-solving skills, and refine AI tools for specific educational tasks. This study informs educational practice and AI development, emphasizing careful AI integration in learning environments to support, not diminish, student agency and cognitive growth.
Zhu, G., Sudarshan, V., Kow, J. F., & Ong, Y. S. (2024). Human-generative AI collaborative problem solving: Who leads and how students perceive the interactions. arXiv. https://arxiv.org/abs/2405.13048📰 In the News
Senate Committee on Commerce
Sens. Moran, Cantwell Introduce Legislation to Boost AI Education ↗️
Key takeaways:
Senators Jerry Moran and Maria Cantwell introduced the NSF AI Education Act of 2024, aimed at expanding educational opportunities in AI through scholarships and professional development with NSF support.
The legislation targets a comprehensive range of students, from kindergarten to college, including those from rural, low-income backgrounds and community colleges, emphasizing the need for a skilled workforce to maintain U.S. competitiveness in AI.
The bill authorizes the National Science Foundation to fund undergraduate and graduate scholarships in fields like AI and quantum technology, and create professional development fellowships for educators and STEM workers.
Proposed initiatives include establishing community college "Centers of AI Excellence" across the nation, enhancing AI education for diverse groups including women and rural residents, and supporting AI research and training in agriculture to improve crop production.
Forbes
Ava AI Chatbot Could Help Ease The School College Counseling Crisis ↗️
Key takeaways:
The American School Counselor Association highlights an alarming student-to-counselor ratio of 385-to-1 in U.S. public schools, exacerbating the struggle to provide adequate college counseling and support during a youth mental health crisis.
The College Guidance Network has introduced Ava, an AI-driven chatbot designed to ease the workload of overburdened school counselors by providing democratized access to expert guidance and educational planning tools.
Ava offers a robust set of features, including personalized academic and extracurricular planning, multilingual support, and advice sourced from over 277 specialists across 110 topics, ensuring tailored and high-quality guidance.
The AI tool incorporates rigorous data protection measures, regular bias audits, and transparent algorithms to maintain privacy, fairness, and accountability in providing educational advice.
Looking ahead, Ava aims to enhance rather than replace human counselor interactions, focusing on making personalized guidance accessible nationwide, particularly in underserved areas, through various pricing strategies.
“Chatgpt.” ChatGPT, OpenAI (GPT-4), openai.com/chatgpt. Accessed 27 May. 2024. And that’s a wrap for this week’s newsletter! We look forward to having you join us at our upcoming webinar!
If you enjoyed our newsletter and found it helpful, please consider sharing this free resource with your colleagues, educators, administrators, and more.