Chatbot Collaboration: OpenAI and Khan Academy

In early 2021, Sal Khan, founder of Khan Academy, was approached by Greg Brockman, president of OpenAI, to explore the integration of ChatGPT into educational resources. At the time, OpenAI was relatively unknown, and ChatGPT-3 was experimental, often making mistakes similar to a robot vacuum bumping into walls. Despite being intrigued, Mr. Khan politely declined the initial proposal for collaboration.

OpenAI’s interest in Mr. Khan stemmed from his widespread admiration in the educational sector. Since its inception in 2008, Khan Academy has provided thousands of free educational videos and interactive exercises across various subjects, aiming to deliver “a free, world-class education for anyone, anywhere.” Today, nearly 800 U.S. school districts utilize its software, and globally, about 190 million individuals access its services.

By summer 2022, Mr. Brockman, along with co-founder Sam Altman, reached out again with a more enigmatic proposal, months before ChatGPT-3.5 would introduce the world to generative AI. Mr. Khan and Kristen DiCerbo, the chief learning officer at Khan Academy, signed a non-disclosure agreement during this outreach and learned about GPT-4 with enhanced capabilities.

Observing GPT-4, Mr. Khan noted its prowess in answering advanced placement biology questions and its ability to explain both correct and incorrect responses. It could also generate new exam questions dynamically. Initially perceiving the model’s capabilities as a breakthrough, Mr. Khan remarked, “This is a big deal,” acknowledging later that perfection was yet to be achieved.

OpenAI aimed to launch GPT-4 within six months and proposed that Khan Academy introduce a tutoring chatbot simultaneously. Mr. Khan felt conflicted due to the significance of his work at Khan Academy, which influenced his decision-making.