Rethinking AI in Education: Beyond ChatGPT to Personalized Learning

AI is transforming every industry, and education is no exception. But it's not all good news. Almost daily, new articles emerge, warning of plagiarism, cheating, and falsifications made possible by students misusing tools like ChatGPT. We're warned that AI will create a generation of unqualified graduates.

But current conversations focusing on large language models (LLMs) like ChatGPT and Gemini miss the mark. While LLMs have sparked both excitement and concern, they aren't the right fit for education.

The Problem with LLMs in Education

Large language models (LLMs) like ChatGPT are trained on the internet. In other words, they learn from a vast, unfiltered sea of data. And they can sometimes mirror the internet's chaotic nature, generating text that's irrelevant, incorrect, or incomplete. LLMs are also known to "hallucinate", a phenomenon where the model makes up convincing sounding facts in order to fill gaps in the text.

LLMs' strength in generalization and producing human-like responses to questions or prompts becomes a weakness in education, where precision, trust, and alignment with learning objectives are crucial.

But while LLMs like ChatGPT and Gemini fall short of reasonable expectations, there is a future for AI in education. That future lies in closed AI models, designed specifically for educational contexts.

Closed AI Models: A Safer, More Reliable Approach

Closed AI models offer a lesser-known but potentially revolutionary alternative to LLMs. Unlike LLMs, which attempt to cover an expansive range of topics, closed models are designed to operate within specific guardrails. These models are trained on carefully curated, high-quality datasets relevant to educational goals, rather than the entire internet.

The appeal of these models comes from their ability to:

• Operate within specific, curated datasets of high-quality educational material
• Align with curriculum objectives and educational standards
• Provide more precise and reliable information
• Avoid the risks associated with indiscriminate internet scraping

Closed AI models avoid many of the risks associated with LLMs. They are able to meet the rigorous standards education requires because their datasets are smaller, controlled, and highly focused.

Personalization, Flexibility, and Safety

A criticism of closed AI models is that they are inherently restrictive and static compared to their larger counterpart, the LLM. While it's true that closed models don't offer the same breadth of knowledge, this limitation is actually their strength in an educational setting.

Despite this apparent constraint, closed AI models are still able to deliver personalized and dynamic learning experiences. They support learner-centered education by allowing students to explore interconnected topics at their own pace while ensuring they don't veer into misinformation or irrelevant content.

For example, a student curious about Tuscany in preparation for an upcoming vacation could start by learning about its modern landscape, then dive into the production of wine in the region, explore the economics of the hospitality industry, and even investigate the process of opening and running a fine restaurant. With a closed AI model, this kind of free-ranging exploration is not only possible but safe.

This approach creates rich, personal, exploratory learning experiences impossible in traditional, rigid curricula.

Educators as Guides in an AI-Driven Environment

While AI is set to reshape education, the role of educators isn't diminished, it's enhanced. Rather than replacing educators, AI is set to free future educators to become facilitators of personalized learning, focusing on developing critical thinking, problem-solving, creativity, and social-emotional skills.

Beyond that, by analyzing how students interact with the system, AI can identify any learning obstacles faced by individual students, and flag them for the educator's attention.

The partnership between AI and human educators will make learning more flexible, responsive, and tailored to individual needs, better preparing students for a rapidly changing world.

Dynamic Curriculum Evolution

One advantage of closed models is their ability to be updated in real time without the need for overhauling entire curricula or undergoing costly textbook revisions. This ensures that students always have access to the most current knowledge without waiting for textbook updates. This adaptability keeps education systems flexible and responsive to new developments.

Reducing Educational Inequality

By offering personalized, up-to-date learning experiences regardless of location or school resources, closed AI models could significantly address educational inequality. They could deliver high-quality education to remote or underfunded areas, helping to level the playing field.

Closed AI - But at What Cost?

Contrary to expectations, closed AI models are more cost-effective than LLMs. They can operate offline, require less computational power, and have a smaller carbon footprint. This makes them not only more affordable for educational institutions but also a more environmentally responsible choice.

For instance, a school district could implement a closed AI model tailored to their curriculum, running it on local servers without the need for constant internet connectivity or expensive cloud computing resources.

The Path Forward

The future of AI in education isn't about mimicking human conversation or automating tedious tasks. It's about designing AI that understands the specific needs of learners, educators, and educational systems. Closed AI models, with their curated data and ability to offer dynamic, personalized learning, provide a path forward that addresses current concerns while unlocking new opportunities for innovation and growth.

As we navigate this transformation, it's crucial to remember that humans will remain essential. The key to the future of learning lies in the symbiosis between educators and purpose-built AI, creating a more supportive, personalized, and dynamic learning environment for all.

What Do You Think?

Education has remained largely unchanged for over a century, even as other aspects of life have transformed radically. Is AI the technology with the potential to disrupt education in a meaningful way? Feel free to share your thoughts below!

And if you want to read my full article it’s here on Medium.