3. AI still needs people who can judge its work
AI can produce a plausible answer without producing the right answer. The Canadian Centre for Cyber Security warns that generative AI output can be incorrect, nonsensical, incomplete or biased. It also notes that developers can introduce insecure or buggy code when they rely on AI without properly implementing error handling and security checks.
Detailed review requires technical knowledge and evaluative judgment. Computer science students learn to test assumptions, trace errors and assess whether a solution meets its requirements. Those skills become more valuable when producing an answer is easy but verifying it is not.
The fastest route to an answer is not always the best route to learning. Dr. Thiel describes the process of working through difficulty, making mistakes and searching for solutions as productive struggle. “Students should be running towards that discomfort, because that discomfort is the sense of learning,” he says. “That’s finding something new.”
AI can remove some of that discomfort by producing an immediate answer. But the struggle is part of the intended learning. It is part of what makes higher education so powerful. Students learn to think for themselves, work through unfamiliar problems and develop expertise in their field.
Formative educational experiences also help create the next generation of senior professionals. “If you replace all the people down at the bottom, you don’t have anyone progressing up, gaining experience, developing evaluative judgment," Dr. Theil says.
