A Critique of the University's 'AI Practice Course'

If the course begins by asserting authority through intimidation, with the atmosphere of a sealed institutional vat, then I have no interest in participating. A group of people who have not even subscribed to ChatGPT Plus are gathering to conduct an “AI practice” course; beyond a review of classical deep learning and guided readings of post-Transformer literature, I can scarcely imagine what substantive content they could offer. If my aim is simply to acquire knowledge, I will of course turn to AI. Competent prompt engineering is, in effect, like having two PhD candidates alternate in providing intellectual labor on my behalf.

For those who do not aspire to conduct research in NLP / CV / RL, the central task is not to immerse themselves in esoteric academic assignments and then serve as executors of application-layer transfer, thereby becoming academic human capital. What truly matters is how to integrate AI as a tool into one’s own workflow; this is what AI-native is meant to signify.

The pathology of most such “practices” lies precisely here: they regard AI merely as a local patch upon existing paradigms, rather than as the opening of a new interaction paradigm altogether. Hence the familiar contradiction: companies claim to seek candidates who can use AI tools well, while simultaneously banning AI in structured interviews.

The current rush toward deep learning is less the conviction of a believer than the acquisitive posture of an investor; it remains far from the self-consciousness of an AI-native believer. And as I have argued before, such stampedes are often irrational. Even a modest familiarity with the consumer-side landscape and community practice would be enough to temper this near-fanatical imagination.