With the ongoing craze surrounding generative AI and LLMs, there have been increasing concerns displayed in recent news headlines:

• ChatGPT Banned in New York City Public Schools Over Concerns About Cheating, Learning Development (source: WSJ)
• Top AI conference (ICML) bans use of ChatGPT and AI language tools to write academic papers (source: The Verge)
• A junior engineer hoping to improve their code discovers that the company internet had restricted access to ChatGPT (source)

At first thought, it’s a fair argument that generative AI tools like ChatGPT and Stable Diffusion can create concerns around issues like plagiarism and the degradation of critical thinking.

However, when tracing back in history, some of the most common tools in the modern classroom went through a similar phase of rejection. The integration of calculators into the education system in the 70s faced a similar backlash, with skeptics fearing that the use of calculators will impair students’ analytical capabilities.

In the proceeding decades, these concerns proved to be far from reality. Calculators freed students from spending time performing paper and pencil calculations that proved to be tedious. Instead, students developed deeper levels of mathematical thinking and reasoning and were able to grasp more complex theorems and concepts at an earlier age. As a result, educators evolved math instruction and curriculums to focus on developing mathematical understanding, reasoning, number sense, and applications. (source)

Don’t forget, the use of the internet and laptop devices in classrooms also went through a very similar battle not long ago. With answers readily available at our fingertips via a quick search on the internet, educators thought that it was game over. Yet, two decades passed, and teachers were actually able to shift the curriculum to become more substantive, collaborative, and project-based instead. The rise in self-directed learning is also thanks to online learning.

With the ability to produce “80%-ready” drafts of answers or provide a collection of ideas surrounding a specific topic in a matter of seconds, will the use of generative AI in education also result in shifts in the current curriculum? Absolutely. Some ways I think education curriculums will shift in the long run:

• More emphasis on creative discovery — Generative AI is a passive tool. Just like how prompt engineering is a ‘craft’, a quality input is still required for generative AI to produce high-quality outputs.
• Emphasis on better judgment — Noah Smith mentions the rise of ‘sandwich’ workflows. “First, a human has a creative impulse, and gives the AI a prompt. The AI then generates a menu of options. The human then chooses an option, edits it, and adds any touches they like.” While this may take away some of the joy from creating something from 0 → 1, this will also free up effort that can be dedicated to the 1 → 100 part of the creation.
• Enablement of personalized learning — Students are not the only ones to benefit from generative AI. Teachers will be able to create more customized learning plans, prompts, or questions, that are catered to different skill levels or learning needs.

In addition, tools like GPTZero are already emerging with the ability to quickly detect whether a piece of text is created by generative models. OpenAI is also working on the ability to implement cryptographic watermarks, that could likely be used to deter basic cases of plagiarism.

At the end of the day, it may be daunting that ‘B+’ essays are generated at the snap of a finger. When AI reaches a widespread level of prevalence in the near future, there will be a new round of regression to the mean where ‘B+’ is just the new ‘C-’.