When Sal Khan first had access to an advanced ai big language model, like many edtech enthusiasts, he couldn't get enough of it.
“There are the early adopters of technology and the technology curious,” Khan says. “When we first had access to technology, we didn't sleep, we just kept playing with it.”
Khan's early use of ai encouraged him to adopt it. GPT-4 technology helps power Khanmigoan ai tutor from its free online platform Khan Academy.
Khan believes these types of ai tutors can revolutionize the way writing is taught. and many other aspects of education. But at the same time, his personal use of ai has evolved. He spends less time playing with the tool and more time using it in specific ways to help prepare for the educational videos he still makes for Khan Academy.
This is how Khan uses it as an educator.
<h2 id="how-sal-khan-uses-ai-learning-through-discussion-and-ideation-xa0″>How Sal Khan uses ai: learning through discussion and ideation
Khan uses ai to prepare for making educational videos in two key ways. First, explore deeper conceptual questions that he has, and that many students might also have, about the topics the videos address. Secondly, Khan uses it to “devise.” For example, he says that if he is looking for books on a certain topic and some titles come to mind, he could request more information from an artificial intelligence tool.
The first use is the one you usually find most useful. He believes the key to his success is his ability to ask questions during his videos that anticipate questions students may have. This arose from his own lifelong intellectual curiosity.
“Maybe some teachers got upset with me when I was a kid. I would ask the polite question. 'Hey, if that's true, wouldn't it be true?'” Khan says.
When planning his videos, Khan explores conceptual questions that you may already know the answers to, as well as questions that you're not sure about and need to explore before filming. A good, large language model can be the perfect tool for testing these types of questions and can lead to deeper learning in the process, he says.
<h2 id="conceptual-questions-and-ai“>Conceptual questions and ai
A recent example of how Khan uses ai to explore conceptual questions came when he was preparing for a video and trying to understand why supernovae explode. Although he knew the basic concepts, he couldn't visualize why the death of a star would trigger an outward explosion. He then asked Khanmigo. Instead of providing a Wikipedia-style answer, he asked him what he already knew. Khan told the ai that he knew the basics but that he didn't quite understand why a star that had essentially run out of fuel would explode.
“I think it would collapse if the entire merger was stopped,” Khan explained to the ai when asked.
The ai told him that he was not wrong, but asked him several questions encouraging him to think more about that collapse and what it entails. “Have you ever seen something fall quickly without bouncing?” The ai asked. Khan responded, “Oh, you're telling me that it collapses so fast that it essentially compresses the core and bounces back?”
The ai said, “Yes, that's exactly what's happening.”
That led Khan to an “aha” moment, and after the exchange, he had a better idea of what caused a supernova. This type of conceptual insight is what Khan believes ai is so well positioned to do for both educators and their students.
“I've done Internet searches about this and read NASA articles,” he says. “I didn't get the conceptual understanding that I did from this ten-minute conversation I had with the ai.”
Ultimately, as with many things in education, getting the most out of an ai interaction is about asking good questions to create a useful dialogue.
