🚀 DeepSeek vs. OpenAI: Should Educators Pay Attention to This New AI?
Learn more about the current buzz around DeepSeek and its potential to provide lower-cost AI solutions to schools around the world.
Many of us are familiar with AI giants like OpenAI, but there's a new player making waves: DeepSeek. This Chinese startup has made significant strides in developing large language models, recently releasing a reasoning model that rivals some of OpenAI’s most advanced models at a fraction of the cost to train and develop. Its efficiency has disrupted the industry so drastically that Nvidia lost $600 billion in market value overnight, marking the largest market cap loss in history. Even more groundbreaking, DeepSeek’s models are open-source models available to the public for free! What makes DeepSeek so powerful, and what does this mean for students and educators? In this edition, we’ll break it down and explore how AI models like DeepSeek can transform the field of education.
Here is an overview of today’s newsletter:
Explanation of the current buzz around Deep Seek and OpenAI’s Operator
Latest survey results on student usage of AI in schoolwork
Exclusive interview with Merissa Sadler-Holder, founder of Teaching with Machines
Research study comparing the effectiveness of AI-generated vs. human-made teaching videos
🚀 Practical AI Usage and Policies
⭐️ AI x Education Webinar - The Digital Divide 2.0: Can AI Level the Playing Field?
Join us virtually on February 6th, as we dive into how AI is reshaping education and explore what it takes to ensure equity, accessibility, and innovation in classrooms. Our latest webinar brings together:
Glenn Kleiman – Senior Advisor, Stanford Accelerator for Learning
Chris Mah – PhD Candidate, Stanford Graduate School of Education
Benjamin Klieger – CS Undergraduate, Stanford University
Our speakers will share actionable use cases, strategies for equitable AI integration, and the paradigm shifts needed for Gen AI-powered classrooms. Topics will include personalized learning, bridging resource gaps, and ensuring that AI benefits everyone.
Attendees will leave with practical tools and a deeper understanding of how AI can create more inclusive and innovative learning environments. Whether you’re an educator, administrator, or student, this session is designed to spark meaningful discussion and provide valuable takeaways.
💻 The Latest Developments in AI
Deep Seek
🔎 What Is DeepSeek?
DeepSeek is a Chinese AI company that has recently developed a groundbreaking AI reasoning model called DeepSeek-R1 that’s not only fast and efficient but also much cheaper to create than anything we’ve seen so far. While companies like OpenAI are spending billions on AI development (see The Stargate Project), DeepSeek built its system with just $6 million! 😲
Here’s the kicker: their AI matches or even outperforms existing models in many areas like problem-solving, logical reasoning, and coding. It has become the top app surpassing ChatGPT on the Apple app store!
🌟 Why Should Educators Care?
Unlike some companies that keep their tech locked behind paywalls, DeepSeek is an open-source large language model that makes the model free to the public. This could mean more opportunities for classrooms to incorporate lower-cost advanced AI solutions without breaking the bank.
DeepSeek-R1 excels in areas like math and logic, showing promising applications in STEM education. It could provide real-time tutoring in math, coding, and other complex subjects, adapting to students’ needs faster than ever before.
With companies like DeepSeek stepping up, innovation in the AI space could accelerate significantly. However, this progress also brings important questions about how to enhance learning while safeguarding student privacy and fostering critical thinking skills.
🤔 How Does DeepSeek Compare to ChatGPT?
DeepSeek uses fewer computational resources during training and usage by utilizing techniques like inference-time computing while activating only the most relevant parts of the model for each task.
While ChatGPT excels at general-purpose tasks, like writing essays and having casual conversations, DeepSeek specializes in reasoning capabilities, excelling in areas like solving math problems, tackling coding challenges, and handling logical reasoning tasks.
DeepSeek uses distillation techniques to create smaller, efficient versions of its model that still retain strong reasoning capabilities. This makes it more adaptable for low-resource settings with limited computing power, like schools with limited tech budgets.
OpenAI Operator
In our previous newsletter, we explored the upcoming emergence of AI agents in 2025. Now, that vision has become reality with OpenAI's recent release of Operator, an advanced AI agent designed to automate tasks for you. Operator can autonomously navigate the web to handle tasks like making dinner reservations, filling out forms, and much more.
🌟 Why Should Educators Care?
With Operator’s ability to navigate the web and access resources quickly, it could make educational materials and tools more accessible to students. For instance, students could request help with research, find textbooks, or locate educational materials tailored to their curriculum.
Teachers often juggle numerous administrative tasks, from grading to managing emails and scheduling. Operator could streamline these processes, allowing teachers to focus more on direct interaction with students. It could assist with automating repetitive administrative tasks, saving time, and reducing burnout.
As safety and privacy are critical concerns in educational settings, Operator’s emphasis on control, safeguards, and data privacy will be key to its adoption. Schools and educators will need to be confident in the tool’s ability to protect sensitive data, especially when working with students.
🧠 Food for Thought
The Pew Research Center recently published its latest findings on American teens' use of ChatGPT in school. A survey of U.S. teens aged 13 to 17 found that the proportion of teens using ChatGPT for schoolwork has increased to 26%, compared to 13% in 2023. The survey also showed that ChatGPT use varies by demographic group, with Black and Hispanic teens more likely than White teens to report using it for schoolwork. On the other hand, the study found no significant differences in ChatGPT use for schoolwork based on gender or household income. For more detailed statistics, check out the findings here.
An AI Chatbot Took A Graduate Course And Got An A. No One Noticed. (Forbes)
A recent experiment showed just how advanced AI has become. ChatGPT completed an entire online Master's course in health administration, earning a 99.36% grade, higher than the class average. What's even more surprising is that no one noticed. The chatbot participated in discussions, seminars, and assignments without detection, highlighting concerns about AI's role in academic integrity. The findings also suggest that if AI can perform well undetected, the value of such degrees could be at risk.
Ethan Mollick’s latest Substack post that explores different AI tools and its specific use cases:
🧭 Resources for Students and Educators
Virtual Event: Building an AI-Powered University (OpenAI)
Join leaders from OpenAI, GSV Ventures, and top academic institutions today (1/28) as they discuss how AI is transforming higher education. They will dive into AI integration, the shifting university model, and practical insights for institutions navigating AI transitions. The session wraps up with a live Q&A, giving you the chance to engage directly with the panelists.
The US Office of Educational Technology recently released this brief to support institutional leaders currently overseeing AI implementation at their institutions. AI is reshaping key areas such as admissions, enrollment, academic advising, and learning environments. Postsecondary institutions can leverage AI to both expand access and success for all students, as well as prepare students for careers in a landscape transformed by AI innovation.
Resources Compiled by Lily Lee📣 Conversations with AI Leaders
This week, we had a chance to interview Merissa Sadler-Holder, founder of Teaching with Machines, a newsletter and consulting platform dedicated to AI professional development for educators. She was honored with ASU+GSV's 2024 Leading Woman in AI award. Prior, she spent 13 years as a high school teacher. In the following, we present select highlights from these conversations:
Q: What does the current conversation about AI and education get wrong about the needs of teachers?
That we need something else to help us write lesson plans. It's great, but that isn't solving the problem. And I think, back to what we were talking about—Chris Dede says his greatest fear about AI and education is that it will become the duct tape holding together a crumbling, industrial revolution educational system. I feel the same way.
It goes back to what we were talking about at the beginning of our conversation. AI is wonderful, there's lots of promises out there, and it has major potential. But there's going to have to be some major changes that only humans can make at a fundamental level of education. Right now, all of these things are just Band-Aids on a crumbling system that isn't working. I myself am a teacher who wanted to leave the system, but then I was able to help fight for that voice in this space because I do see hope for teachers.
Q: One of educators’ top concerns is the potential for cheating and plagiarism. What are your thoughts about assessments in the new era of AI?
I always think of educators as incredibly smart and talented individuals, if given the time to innovate and create. But we just don't give them that time. If we did, and said, "Let's be innovative about how we assess things," we'd create something worthwhile.
However, it's really difficult because every teacher has to confront this challenge. For example, in my French classes—I spent about two years preparing all these digital projects for my students. Then Google Translate was released. Suddenly, I had students working on digital projects that rivaled Molière-level writing. It was super high-level French, and I thought, "I didn’t teach you that, and I know for a fact you don’t know it."
It was kind of the same situation we’re finding ourselves in now. But what I ended up doing was saying, “Okay, this is not a technology I’m going to fight with my students because, A, it's literally in their pocket.” So I decided to teach them how to use Google Translate in a way that’s beneficial for them, but I could still track their progress.
I created a system where I said, “You can use Google Translate, but you have to cite your work. You have to show me you understand, and fill out this paperwork with it. You have to identify areas of challenge.” It wasn’t just about using it freely; they had to show me their work, understand why they used it, and where they struggled. As a teacher, this allowed me to pinpoint what areas my students were struggling with and why they used the tool.
I think we can do the same with AI—it will just look a little different. I'm not totally changing my assessment. I still used the digital projects, but I adjusted my approach and taught the students how to use the tool. That way, they weren’t caught up in that back-and-forth about whether it's cheating. It was clear to them exactly what they needed to do to use Google Translate, and there was no ambiguity.
I’ve created something called a Flexible AI Toolbox, which is available on my website. It’s free to download and helps teachers take a project they already have and adapt it by attaching AI guidelines to the project prompt. Teachers can simply check off what parts of the project students can use AI for and set those parameters. This provides a clear framework for integrating AI into the learning process. Not only does this approach build AI literacy for students, but it also removes the feeling of cheating. It allows you to establish clear boundaries with them, showing where and how they can use AI in your classroom. I think more educators need to adopt this kind of structure.
Q: What advice would you give to educators who feel hesitant or underprepared to adopt AI technologies in their teaching?
I think the first step is figuring out why they're so hesitant. Then, explore that—why am I hesitant about embracing this technology? Is it the time to learn something new? It could be very personal. So, figure that out first. But then also realize that it’s best practice for every educator to know what technology is available that our students have access to. If we can take this technology and shift it from something our students might abuse or use unethically into something that's a tool for learning, then that’s where our expertise comes in. Only we can do that.
Be aware of your own biases, explore them, and be aware that as an educator, you need to know best practices. Then, be kind to yourself because this is a big topic, and it’s a lot to take on. You don’t have to tackle everything at once. Be aware of your own needs, and maybe give yourself a little time to explore it.
Now, if I had a teacher who’s dealt with their biases and acknowledged they have to do this begrudgingly, and they sit down to learn, I would say, pick only one tool—literally just one tool. Play with it a little bit, keep going, and make that your tool of the month until you get really good at it. Then you can move on to the next one.
Again, this doesn’t need to happen overnight, but it is a big topic. Because we're not talking about just a new piece of technology—we're talking about a technology that is going to change humanity. I think that’s really important, because it means it's also going to change education.
Interview Conducted by Imogen Lee📝 Latest Research in AI + Education
Computers & Education Journal
Comparing human-made and AI-generated teaching videos: An experimental study on learning effects ↗️
This study compared the effectiveness of AI-generated versus human-made teaching videos in an online management course. A total of 447 participants rated their learning experiences and completed an exam to assess their learning outcomes. Each video aimed to teach the same content, but one was developed by AI while the other by humans. Below are the videos that were used in the study:
AI-generated
Human-made
Which video did you prefer? The results of the final exam showed that the topics were understood at the same level for both video types, which begs the question, are the AI videos just as good if the watchers understood the material equally well? The study highlights that when producing the AI-generated video, it was significantly faster than recording and producing videos with a human teacher, demonstrating a favorable tradeoff between time and cost for developers. These findings suggest that AI-generated videos could be a practical and scalable solution for education.
Netland, T., von Dzengelevski, O., Tesch, K., & Kwasnitschka, D. (2025). Comparing human-made and AI-generated teaching videos: An experimental study on learning effects. Computers & Education, 224, 105164. https://doi.org/10.1016/j.compedu.2024.105164SBS Swiss Business School, Anglia Ruskin University, London School of Economics & Political Science (LSE)
AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking ↗️
This recent study explores how reliance on AI tools impacts critical thinking, focusing on the role of cognitive offloading—delegating mental tasks to technology. The research reveals a strong negative correlation between frequent AI usage and critical thinking abilities, especially among younger users. As AI tools streamline tasks like decision-making and problem-solving, they also reduce opportunities for deep, reflective thinking. The study highlights an urgent need for educational strategies that balance the convenience of AI with activities that nurture critical engagement. For educators, this means designing learning experiences that empower students to use AI responsibly without compromising their cognitive development.
Student Opinion by Yash Yardi:
This study demonstrates a strong correlation between reliance on generative AI and reduced critical thinking abilities. This claim is elaborated on further with age, as the ANOVA analysis further proves that reliance on AI has more prominent adverse effects for younger users than those who are older. Having said this, if the first study expects the proliferation of AI-generated teaching video tools in the future, I wonder how that may impact AI reliance and cognitive thinking. I suspect that the rise of AI-generated content is likely to replicate the impact resulting from the AI tools analyzed in this study and the effects of LLMs like Chat-GPT or Claude. I have seen that students could use these models poorly (plagiarism or completing assignments swiftly without consideration for learning) or efficiently (engineering prompts to provide practice problems and understand materials), thus reducing or increasing cognitive thinking and comprehension, respectively. Ultimately, only time will tell how students and teachers utilize materials like these.Gerlich, Michael, AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking (January 03, 2025). Available at SSRN: https://ssrn.com/abstract=5082524 or http://dx.doi.org/10.2139/ssrn.5082524And that’s a wrap for this week’s newsletter! If you enjoyed our newsletter and found it helpful, please consider sharing this free resource with your colleagues, educators, administrators, and more.