Will Computer Science become useless knowledge? (arnoldkling.substack.com)

A debate between a UC Berkeley instructor and a colleague (Jain) has crystallized a growing dilemma: students are using AI chatbots to complete coding assignments in an introductory database course, and the instructor argues this short-circuits the formative learning—problem decomposition, debugging, and low-level systems intuition—that traditional CS courses aim to teach. Jain counters that basic tools (like calculators) free practitioners to focus on higher-order tasks, and that for many software engineering jobs deep knowledge of machine code or logic gates may be unnecessary. The exchange frames a larger question: which parts of classical CS curricula are essential knowledge versus historical ballast in an era where AI can write and debug code. This matters to the AI/ML community because it forces a re-evaluation of pedagogy, assessment, and workforce skill sets. If LLMs reliably automate routine coding, curricula may shift toward system design, verification, prompt engineering, data literacy, and governance rather than low-level implementation. But important technical caveats remain: reliance on AI risks gaps in understanding systems behavior, performance trade-offs, reproducibility, and security—areas where kernel-level or algorithmic knowledge still matters. The conversation implies universities must decide whether to enforce formative hands-on practice, retool learning objectives for AI-assisted workflows, or craft new assessments that certify deeper conceptual mastery.