(Washington, D.C.) November 28, 2023 — Silicon Valley-based educator Sal Khan, founder of online learning platform Khan Academy, spoke on the Doha Debates podcast about the need for “guardrails” and ethical practices when it comes to using generative AI in student learning products, and why he foresees potential harm in for-profit education-focused AI companies.
On the new Doha Debates podcast episode, Khan spoke with NBC News technology correspondent Jacob Ward for a debate moderated by journalist Joshua Johnson. A former hedge fund analyst, Khan founded Khan Academy in 2008 after being inspired by his experiences tutoring his young cousins. Today, his educational technology nonprofit serves more than 160 million users in 50 languages and has produced over 6,500 video lessons. Khan Academy recently introduced Khanmigo, an AI-powered teaching assistant piloted to students in California.
“When OpenAI reached out to us over a year ago, before ChatGPT even came out, we were some of the first folks in the world to see GPT-4,” Khan said. “We had the reaction that GPT-4 can do the rich type of tutoring experience that we always aspired to do — frankly, more than I thought would be possible in my lifetime.”
Khan discussed cheating in AI, and why he believes some edtech start-ups may misuse AI technology.
“There are multibillion dollar for-profit companies where their business model — this was before generative AI — is to essentially do kids’ homework for them,” Khan noted. “I won’t name names, but some of their stock prices have gone down dramatically after ChatGPT came on the scene.”
When it comes to edtech, Khan said, “You already see billions of dollars already going into AI-driven edtech. People are just pumping money into that.” He added, “Everyone’s talking big about helping humanity and leveling the playing field, but after a few years, the venture capitalists want to see a return on the business.” Khan predicted that market forces may eventually drive these AI-driven companies to provide “free homework help,” “help students cheat” and “not do the pedagogically sound thing.”
Khan said, “I think there’s a few sectors in our economy where market forces are leading to suboptimal outcomes — including education and healthcare, where the beneficiaries are different than the payers, who are different than the decision makers.”
The founder of Khan Academy also addressed the pitfalls of using AI in online education technology, including the possibility of bias and AI hallucinations. He noted, “We know AI is infamously bad at math, which surprises some people about computers.”
For college students interested in using AI, Khan expressed caution and urged students not to “get into an unethical space” or use these tools if they violate their school’s honor code.
Khan went on to say, “This is all changing very, very quickly. I encourage people to not minimize the power of AI tools. Already, we know these models are in the 80th percentile of the LSAT; they’ve passed the medical boards at a higher level than a lot of doctors. This is just the current generation. By the time we get to GPT-6 or GPT-7 — which will have similar complexity to the human brain and get more exposure to information than any human being could do in a thousand lifetimes — AI is going to raise some interesting questions.”
Khan emphasized the need for guardrails, which he said his nonprofit Khan Academy has already put into place. “What we’ve done on Khanmigo is, it only talks about things anchored in human-generated, vetted content. I wouldn’t say that we’ve eliminated the problem of misinformation or hallucinations, but we’ve been able to mitigate it a good bit.”
In the classroom, Khan noted, “Our true north at Khan Academy has always been, how do we leverage technology to encourage more human-to-human interactions? What we’ve always preached is, let’s thoughtfully use technology so it can unlock more time for human-to-human interaction. When human beings are together, they should not be lecturing to each other; they shouldn’t be making kids stay quiet so an educator can say something that an on-demand video can do, much less an AI can do. That’s the direction we need to go in.”
Finally, Khan offered thoughts on the future of edtech, as artificial intelligence advances rapidly.
“It is very early days in terms of efficacy studies, but this is something we at Khan Academy take very seriously,” Khan shared. “We’re already doing AI studies, and we have early evidence that it’s keeping students engaged more and they’re quitting less. It’s very early, very preliminary, but I do expect in the next few years you’re going to see some pretty strong evidence.”
He also shared a warning about bypassing AI, stating, “If the good actors aren’t out there, and trying to proactively use it well, ChatGPT and all these tools are still going to be there, and market forces are going to create a whole bunch of cheating tools. It’s going to be very tempting for adults to just keep watering down things, lowering expectations and kids will suffer. Meanwhile, affluent, educated families will provide these supports and keep the rigor high, and then you’re going to have a higher and higher separation” between kids in the classroom.
Khan shared, “Our goal is, let’s be that good actor, let’s harness these technologies to try to level that playing field and up the rigor for everyone. There are so many people with such a strong incentive to create crap that is not going to be good for people, it’s going to create a lot of noise.”
“We’re in it and we feel that there is some real power to this technology. We have to be careful about things like bias and so forth, but big picture, let’s make a path that is more utopian than dystopian. If the good actors are out there, putting in proper guardrails and using the technology thoughtfully, I think we can get to a more utopian world.”
Jacob Ward, author of “The Loop: How Technology Is Creating a World Without Choices and How to Fight Back,” sounded a warning for listeners. “The technology by its nature is a shortcut machine, it cheats by its nature,” said Ward. “It grabs the work of other people, it regurgitates it in ways that are very often inaccurate and sometimes dangerously misleading, and it does not show its work. It is the classic black box, and the market incentives that run all through that industry continue to make it more and more opaque.”
He concluded, “The market dynamics at work around AI applied to education, really cheat all of us, especially our students, by driving down the resources available to us for what we want to do in our lives, while increasing the expectation of productivity.”
The Doha Debates Podcast is a biweekly podcast that brings together people with starkly different opinions for an in-depth conversation that tries to find common ground. Doha Debates is a production of the Qatar Foundation. The full podcast can be viewed here. [https://dohadebates.com/podcasts/]