Students are using AI tools instead of building foundational skills - but resistance is growing (www.zdnet.com)

On day one of class UC Berkeley instructor Stephen Klein asked students an essay question — a week later several submissions were nearly identical. Klein used the moment to illustrate a broader problem: students are outsourcing thinking to large language models (LLMs), which he describes as “stochastic probabilistic auto-complete” engines that produce hollow, templated work. That anecdote echoes a June 2025 open letter from 14 technology professors urging universities to halt uncritical adoption of AI, push back against industry hype, and protect critical thinking, academic integrity and scientific rigor from brittle, market-driven tools. Researchers and instructors warn of concrete technical and curricular fallout: graduates proficient at prompt engineering and copying from code assistants but weak on algorithms, core CS concepts and collaborative problem‑solving. Ishe Hove reports enforcing in‑class, hands‑on coding to ensure students can reason through code without AI. Experts like Amelia Vance argue AI should be adopted only when it delivers tangible educational value (e.g., mapping curriculum standards or aiding lesson design) and after careful vetting. The takeaway for the AI/ML community: balance tool literacy with fundamentals, design assessments that verify conceptual mastery, and prioritize robust, purposeful deployments over convenience to avoid long‑term skill erosion.