Ex-Microsoft exec Craig Mundie has heard this question again and again — parents asking him a version of the same worry: Their kids are heading toward college, artificial intelligence is advancing fast, and jobs feel uncertain. What, exactly, should their kids be studying?
That question — what education will matter most in five years — reflects a deeper uncertainty about the future.
Mundie, who spent 22 years at Microsoft helping steer the company’s vision toward AI and retired as the company’s chief research and strategy officer in 2014, says that parents are simply asking the wrong question.
It’s not only the students who have to change to fit the new AI era — it’s the education system itself, said Mundie, who now advises other executives on AI and public policy.
Rather than chasing down the right job, Mundie urges families to prepare kids for a world where learning itself becomes continuous, personalized, and done in partnership with intelligent machines.
AI is altering the human experience
During an interview with Business Insider’s Reem Makhoul in June, Mundie said artificial intelligence and robotics are poised to reshape work more deeply than past technologies. See the edited cut of his interview below:
That shift, Mundie said, forces a bigger question than which job skills will survive. It challenges how societies define human value. This is something Mundie’s been pondering for over a decade.
In his 2015 book “Genesis,” Mundie, with co-authors Eric Schmidt and Henry Kissinger, examined how AI could alter the human experience. “What we say is we have to think differently about how we value ourselves and what we do.”
For much of history, he said, dignity has been tied to work because people had to work to survive. AI could loosen that link by automating more tasks across both physical and intellectual labor.
Meanwhile, humans will need to learn how to work alongside intelligent machines, and the traditional higher-education system doesn’t offer a clear path toward that, right now.
He described today’s education system as sharply divided between STEM and the humanities. The liberal arts emphasize reasoning, but at the expense of special technical skills you learn in STEM fields, Mundie said.
Students will need both skills moving forward. “If I could create a new curriculum in college, it would be a liberal education in technology,” and STEM, he said.
The classroom model itself is reaching its limits
Mundie went further, questioning whether the classroom model that dominates education today still makes sense.
He traced that structure back to the printing press, which created a surge in written information and a need for mass literacy. Schools, he said, became an efficient “machine for teaching” because societies lacked enough individual tutors.
AI changes that constraint.
We can have scalable, polymathic teachers, Mundie said. “We can have as many teachers as we want now because the AI will be the teacher.”
He said this opens the door to a more personalized, Socratic model of learning, where students can interact continuously with an intelligent system that adapts to their curiosity, pace, and interests. Progress would be limited less by standardized curricula and more by a student’s motivation and capacity.
Schools and universities have been slow to embrace this shift. Early reactions often involved banning AI tools outright. “They’ve now given up on that,” Mundie said.
That resistance, he added, is typical of incumbent systems. “The natural tendency of the incumbent is to preserve the incumbent system,” or make only incremental changes, he added. But “when you get something as powerful as these AIs, most incumbent systems are not going to be preserved.”
He also pointed to early experiments on the right track, like versions of Khan Academy, an online non-profit educational platform founded in 2008 and headquartered in California. It uses an AI tutor, named Khanmigo, designed to guide students rather than simply give answers. In those systems, he said, the AI nudges students toward better questions and deeper understanding.
“So that’s the difference between sort of a broad chat about anything interface and an AI application that was specifically oriented around teaching,” he said, adding, “That’s just one tiny example of how people will build more and more apps on these common artificial intelligence platforms.”
“We will move beyond the specific generic interface to a world of millions of applications that are really customized in some clever way to guide people to solutions in the areas they care about,” he said. These agents may, in fact, do much of the work autonomously by interacting with others, he added.
Mundie said parents and older generations may have difficulty imagining this model, while children are likely to adapt quickly. The harder question, in his view, is whether educational institutions are willing to change.
Read the full article here
