What's up in the world, by Dmitry Stavisky
The purpose of this thought experiment is to provoke discussion about the future of AI in education, a topic I began discussing in my previous blog post.
Imagine it's 2028. AI has progressed like Sam Altman prophesied (check out this podcast.) The latest bombshell from OpenAI, GPT-4o, hints at this future. In other words, assume that the latest AI progress S-curve based on transformer/LLM and diffusion architectures (perhaps with things like JEPA mixed in) continues for a few more years.
You are tasked with building AI school teachers for Kenya, a country facing a severe shortage of human teachers. These AI tutors will work alongside local educators and report to a human headmaster.
You'll use a foundation model like GPT-6 or Claude-5 (let's call it "AI") and "train" your teacher. Start by asking the AI to research the best practices for training elementary school teachers:
"You are an assistant professor at an Ivy League university's School of Education. You have a Ph.D. in Pedagogy. You also have an MBA, focusing on public education. Research best practices for training elementary school teachers. Include modern methodologies like game-based learning and project-based learning. Write an outline for a research paper."
