Should you use AI for a task? Here’s a simple way to decide | Bruce Schneier
Bruce Schneier argues that whether people should use AI depends on whether a task is “work” or “the gym”. AI is appropriate when the outcome is what matters and the technology can perform the task reliably, but it can be harmful when the effort itself develops important skills.
He applies this distinction to student writing, saying policy memos are exercises designed to build critical thinking through outlining, drafting and revising rather than simply to produce finished documents. Schneier says students may mistake polished AI prose for sound reasoning, while avoiding the productive discomfort that helps them test and improve their own ideas; he also stresses that AI must be accurate, secure and trustworthy before it is used.
In short
- Use AI for outcomes, not for skill-building exercises.
- Writing practice develops reasoning as well as prose.
- Polished AI text can conceal weak logic.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
AI should not be used for tasks primarily designed to develop skills through effort and struggle. Student writing assignments teach critical thinking not through the final product but through the iterative process of outlining, drafting and revising. When students use AI to bypass this productive discomfort, they lose the opportunity to test their reasoning against genuine resistance and may mistake polished but potentially unsound AI output for sound thinking they haven't yet developed.
The case against
AI need not be excluded from learning contexts if used thoughtfully as a tool within the educational process rather than replacing it entirely. Students could use AI to generate drafts they then critically revise and improve, developing both writing and AI-literacy skills in the process. The distinction between 'work' and 'gym' is often blurred in practice, and restricting tools might disadvantage students with fewer resources or less time available, whilst preventing them from learning to use powerful tools effectively as professionals increasingly must.
Explained simply
Artificial intelligence systems are now widely available for tasks from writing emails to solving problems, prompting organisations and individuals to decide when to use them. This raises a fundamental question: are there situations where using AI could be unhelpful even if it produces good results?
Technology expert Bruce Schneier proposes distinguishing between activities where the end result matters most, and those where learning through the effort matters most. AI could be helpful for the first type but potentially harmful for the second.
This distinction becomes particularly important in education, where writing assignments and other tasks are designed to develop skills like critical thinking and problem-solving. Using AI to bypass the difficult work might produce polished results but could undermine the learning that comes from grappling with a problem yourself.
AI Government Politics Technology