{"course":{"slug":"ai-for-students","title":"AI for Students","url":"https://addaly.com/learn/ai-for-students","level":"beginner","track":"work"},"lesson":{"slug":"still-worth-learning","title":"The Honest Case For Still Learning It","url":"https://addaly.com/learn/ai-for-students/still-worth-learning","position":66,"of":73,"minutes":9,"answer":"You cannot supervise what you cannot do, and supervision is the job that is left.","keyIdea":"You cannot supervise what you cannot do, and supervision is the job that is left."},"body":{"format":"markdown","text":"## The question deserves a real answer\n\n\"Why learn to do something a machine does instantly and for free?\"\n\nThat is not a lazy question. It is the right question, and most of the answers students get are bad ones: appeals to character, vague talk about the journey, a teacher who is frightened. Here are the answers that survive scrutiny, in order of strength.\n\n## 1. You cannot check what you cannot do\n\nThese systems are confidently wrong at a rate that matters, and the errors are not random noise. They are plausible. A wrong integral looks like an integral. An invented case citation looks like a citation. A subtly incorrect summary of a paper reads better than the paper.\n\nThe only thing between a wrong output and your name on it is whether you know enough to notice. That is a hard limit, and it does not soften as models improve, because as the errors get rarer they also get harder to spot and you get less practised at looking.\n\nThis is the argument with no sentimentality in it. Every job that survives the automation of production becomes a job of checking, and checking is a skill made entirely of knowing the thing.\n\n> Figure: Five answers, strongest first. Layers, top to bottom: You cannot check what you cannot do (A wrong integral looks like an integral and an invented case looks like a citation.); Thinking needs material already in your head (Insight is mostly collision, and you cannot collide what you have to look up.); Fluency below frees attention above (Working memory is small. Reconstructing what a derivative is leaves nothing for the model that uses one.); The job argument, stated honestly (Routine production is under real pressure. Nobody knows the size of it.); Some things are worth knowing because knowing them is good (Hearing what a poem is doing. Seeing why a proof has to be true.). The concession belongs beside them, because pretending the line never moves makes the whole case look dishonest: long division of six-digit numbers by hand, log tables, memorising a bibliography format. The test for which side a skill falls on is whether it builds the judgement you will use above it or is purely output. Output can go. Judgement cannot, because there is nothing above judgement to check it.\n\n## 2. Thinking requires material already in your head\n\nInsight is mostly collision: two things you already hold turning out to be the same thing. You cannot collide what you have to look up, because looking up is serial and slow and you have to already suspect the connection to go looking.\n\nThis is why a historian notices something in an economics paper, and why a doctor recognises a pattern in an unusual presentation. Not retrieval speed. Simultaneous possession.\n\nA head with nothing in it, attached to a search engine, is not equivalent. It cannot generate the question.\n\n## 3. Fluency below frees attention above\n\nWorking memory is small: a handful of items. Anything effortful consumes it.\n\nA student who has to stop and think about what a derivative *is* has no capacity left for the model that uses one. A student who has to reconstruct basic grammar has none left for the argument. Automaticity at the lower level is not a nostalgic virtue; it is the precondition for operating at the higher one.\n\nThis also tells you which things are safe to hand over. Anything that never sat underneath something else you do.\n\n## 4. The job argument, stated honestly\n\nEntry-level work consisting purely of producing routine output (first-pass documents, standard analyses, boilerplate code) is under real pressure. That is not marketing; it is visible in hiring in several fields already.\n\nWhat nobody knows is the size or shape of it. Anyone giving you a confident number about jobs in 2035 is guessing, including the people selling the tools and the people warning about them.\n\nWhat you can say is this: \"the machine does it, so I don't need to know it\" is a bet that *supervision* also gets automated. That may happen. It has not, and every deployment so far has increased the value of someone who can tell good output from bad. Betting the other way is not a hedge. It is a leveraged position on a specific future.\n\n## 5. The reason nobody says out loud\n\nSome things are worth knowing because knowing them is good.\n\nBeing able to read a poem and hear what it is doing. Following a proof and seeing why it has to be true. Holding a century of history in your head so the news makes a different kind of sense. Understanding what is happening in your own body when you are ill.\n\nThese change what the world looks like from the inside. That is not an economic argument, it does not appear on a graduate outcomes report, and it does not need to. It is allowed to be a reason.\n\n## What is not worth learning any more\n\nAnd now the concession, because pretending the line never moves makes the whole case look dishonest.\n\nLong division of six-digit numbers by hand. Log tables. Memorising a bibliography format instead of understanding what a citation must contain. Hand-drawing a graph that any tool draws better. Every generation of teachers has defended a few of these past the point of sense, and students noticed, and it cost the teachers credibility on the things that actually mattered.\n\nThe test for which side a skill falls on:\n\n> **Does this build the judgment I will use above it, or is it purely output?**\n\nOutput can go. Judgment cannot, because there is nothing above judgment to check it.\n\n## The question to carry out of this course\n\nNot \"can a machine do this.\" Increasingly the answer is yes, and it tells you nothing.\n\nAsk instead:\n\n> **If the machine does this for me, what will I no longer be able to do?**\n\nSometimes the honest answer is \"nothing I need.\" Hand it over and use the hour well.\n\nSometimes the answer is \"check its work.\" Then you already know."},"check":{"question":"Someone argues: calculators removed the need for mental arithmetic, so AI removes the need to write essays. Where does the analogy break?","options":[{"text":"It does not really break. Writing will follow arithmetic, and courses will adapt to the tool the way they adapted to calculators.","correct":false,"tempting":"The historical parallel is genuinely strong and partly correct, which is why it needs the output-versus-judgment distinction to be answered rather than dismissed."},{"text":"It breaks because calculators are reliably correct and AI is not, so the whole difference is one of accuracy.","correct":false,"tempting":"True today, but it makes the case depend on current error rates: the argument would collapse the moment a model became reliable, and it should not."},{"text":"It breaks because writing is self-expression and arithmetic is not, so essays carry a personal value calculators never removed.","correct":false,"tempting":"Concedes the practical ground and retreats to the aesthetic case, which is exactly the move that loses the argument with a sceptical student."},{"text":"Arithmetic was output the calculator replaced; the essay is the process that builds the judgment you would need to check any output at all.","correct":true}]},"prev":{"slug":"exam-hall","title":"Revision That Survives The Exam Hall","url":"https://addaly.com/learn/ai-for-students/exam-hall"},"next":{"slug":"past-papers-and-mark-schemes","title":"Past papers and mark schemes","url":"https://addaly.com/learn/ai-for-students/past-papers-and-mark-schemes"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'The Honest Case For Still Learning It', in AI for Students. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-students/still-worth-learning (Accessed: 31 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"The Honest Case For Still Learning It","courseTitle":"AI for Students","module":{"index":8,"slug":"assessment-and-after","title":"Assessment, and the years after it"},"answer":null,"keyIdea":"You cannot supervise what you cannot do, and supervision is the job that is left.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"stack","title":"Five answers, strongest first","layers":[{"label":"You cannot check what you cannot do","detail":"A wrong integral looks like an integral and an invented case looks like a citation."},{"label":"Thinking needs material already in your head","detail":"Insight is mostly collision, and you cannot collide what you have to look up."},{"label":"Fluency below frees attention above","detail":"Working memory is small. Reconstructing what a derivative is leaves nothing for the model that uses one."},{"label":"The job argument, stated honestly","detail":"Routine production is under real pressure. Nobody knows the size of it."},{"label":"Some things are worth knowing because knowing them is good","detail":"Hearing what a poem is doing. Seeing why a proof has to be true."}],"caption":"The concession belongs beside them, because pretending the line never moves makes the whole case look dishonest: long division of six-digit numbers by hand, log tables, memorising a bibliography format. The test for which side a skill falls on is whether it builds the judgement you will use above it or is purely output. Output can go. Judgement cannot, because there is nothing above judgement to check it."},"alt":"Figure: Five answers, strongest first. Layers, top to bottom: You cannot check what you cannot do (A wrong integral looks like an integral and an invented case looks like a citation.); Thinking needs material already in your head (Insight is mostly collision, and you cannot collide what you have to look up.); Fluency below frees attention above (Working memory is small. Reconstructing what a derivative is leaves nothing for the model that uses one.); The job argument, stated honestly (Routine production is under real pressure. Nobody knows the size of it.); Some things are worth knowing because knowing them is good (Hearing what a poem is doing. Seeing why a proof has to be true.). The concession belongs beside them, because pretending the line never moves makes the whole case look dishonest: long division of six-digit numbers by hand, log tables, memorising a bibliography format. The test for which side a skill falls on is whether it builds the judgement you will use above it or is purely output. Output can go. Judgement cannot, because there is nothing above judgement to check it."}},"blocks":[{"t":"h2","id":"the-question-deserves-a-real-answer","runs":[{"text":"The question deserves a real answer"}]},{"t":"p","runs":[{"text":"“Why learn to do something a machine does instantly and for free?”"}]},{"t":"p","runs":[{"text":"That is not a lazy question. It is the right question, and most of the answers students get are bad ones: appeals to character, vague talk about the journey, a teacher who is frightened. Here are the answers that survive scrutiny, in order of strength."}]},{"t":"h2","id":"1-you-cannot-check-what-you-cannot-do","runs":[{"text":"1. You cannot check what you cannot do"}]},{"t":"p","runs":[{"text":"These systems are confidently wrong at a rate that matters, and the errors are not random noise. They are plausible. A wrong integral looks like an integral. An invented case citation looks like a citation. A subtly incorrect summary of a paper reads better than the paper."}]},{"t":"p","runs":[{"text":"The only thing between a wrong output and your name on it is whether you know enough to notice. That is a hard limit, and it does not soften as models improve, because as the errors get rarer they also get harder to spot and you get less practised at looking."}]},{"t":"p","runs":[{"text":"This is the argument with no sentimentality in it. Every job that survives the automation of production becomes a job of checking, and checking is a skill made entirely of knowing the thing."}]},{"t":"h2","id":"2-thinking-requires-material-already-in-your-head","runs":[{"text":"2. Thinking requires material already in your head"}]},{"t":"p","runs":[{"text":"Insight is mostly collision: two things you already hold turning out to be the same thing. You cannot collide what you have to look up, because looking up is serial and slow and you have to already suspect the connection to go looking."}]},{"t":"p","runs":[{"text":"This is why a historian notices something in an economics paper, and why a doctor recognises a pattern in an unusual presentation. Not retrieval speed. Simultaneous possession."}]},{"t":"p","runs":[{"text":"A head with nothing in it, attached to a search engine, is not equivalent. It cannot generate the question."}]},{"t":"h2","id":"3-fluency-below-frees-attention-above","runs":[{"text":"3. Fluency below frees attention above"}]},{"t":"p","runs":[{"text":"Working memory is small: a handful of items. Anything effortful consumes it."}]},{"t":"p","runs":[{"text":"A student who has to stop and think about what a derivative "},{"text":"is","i":true},{"text":" has no capacity left for the model that uses one. A student who has to reconstruct basic grammar has none left for the argument. Automaticity at the lower level is not a nostalgic virtue; it is the precondition for operating at the higher one."}]},{"t":"p","runs":[{"text":"This also tells you which things are safe to hand over. Anything that never sat underneath something else you do."}]},{"t":"h2","id":"4-the-job-argument-stated-honestly","runs":[{"text":"4. The job argument, stated honestly"}]},{"t":"p","runs":[{"text":"Entry-level work consisting purely of producing routine output (first-pass documents, standard analyses, boilerplate code) is under real pressure. That is not marketing; it is visible in hiring in several fields already."}]},{"t":"p","runs":[{"text":"What nobody knows is the size or shape of it. Anyone giving you a confident number about jobs in 2035 is guessing, including the people selling the tools and the people warning about them."}]},{"t":"p","runs":[{"text":"What you can say is this: “the machine does it, so I don’t need to know it” is a bet that "},{"text":"supervision","i":true},{"text":" also gets automated. That may happen. It has not, and every deployment so far has increased the value of someone who can tell good output from bad. Betting the other way is not a hedge. It is a leveraged position on a specific future."}]},{"t":"h2","id":"5-the-reason-nobody-says-out-loud","runs":[{"text":"5. The reason nobody says out loud"}]},{"t":"p","runs":[{"text":"Some things are worth knowing because knowing them is good."}]},{"t":"p","runs":[{"text":"Being able to read a poem and hear what it is doing. Following a proof and seeing why it has to be true. Holding a century of history in your head so the news makes a different kind of sense. Understanding what is happening in your own body when you are ill."}]},{"t":"p","runs":[{"text":"These change what the world looks like from the inside. That is not an economic argument, it does not appear on a graduate outcomes report, and it does not need to. It is allowed to be a reason."}]},{"t":"h2","id":"what-is-not-worth-learning-any-more","runs":[{"text":"What is not worth learning any more"}]},{"t":"p","runs":[{"text":"And now the concession, because pretending the line never moves makes the whole case look dishonest."}]},{"t":"p","runs":[{"text":"Long division of six-digit numbers by hand. Log tables. Memorising a bibliography format instead of understanding what a citation must contain. Hand-drawing a graph that any tool draws better. Every generation of teachers has defended a few of these past the point of sense, and students noticed, and it cost the teachers credibility on the things that actually mattered."}]},{"t":"p","runs":[{"text":"The test for which side a skill falls on:"}]},{"t":"quote","runs":[{"text":"Does this build the judgment I will use above it, or is it purely output?","b":true}]},{"t":"p","runs":[{"text":"Output can go. Judgment cannot, because there is nothing above judgment to check it."}]},{"t":"h2","id":"the-question-to-carry-out-of-this-course","runs":[{"text":"The question to carry out of this course"}]},{"t":"p","runs":[{"text":"Not “can a machine do this.” Increasingly the answer is yes, and it tells you nothing."}]},{"t":"p","runs":[{"text":"Ask instead:"}]},{"t":"quote","runs":[{"text":"If the machine does this for me, what will I no longer be able to do?","b":true}]},{"t":"p","runs":[{"text":"Sometimes the honest answer is “nothing I need.” Hand it over and use the hour well."}]},{"t":"p","runs":[{"text":"Sometimes the answer is “check its work.” Then you already know."}]}],"check":{"question":[{"text":"Someone argues: calculators removed the need for mental arithmetic, so AI removes the need to write essays. Where does the analogy break?"}],"options":[{"text":[{"text":"It does not really break. Writing will follow arithmetic, and courses will adapt to the tool the way they adapted to calculators."}],"correct":false,"tempting":[{"text":"The historical parallel is genuinely strong and partly correct, which is why it needs the output-versus-judgment distinction to be answered rather than dismissed."}]},{"text":[{"text":"It breaks because calculators are reliably correct and AI is not, so the whole difference is one of accuracy."}],"correct":false,"tempting":[{"text":"True today, but it makes the case depend on current error rates: the argument would collapse the moment a model became reliable, and it should not."}]},{"text":[{"text":"It breaks because writing is self-expression and arithmetic is not, so essays carry a personal value calculators never removed."}],"correct":false,"tempting":[{"text":"Concedes the practical ground and retreats to the aesthetic case, which is exactly the move that loses the argument with a sceptical student."}]},{"text":[{"text":"Arithmetic was output the calculator replaced; the essay is the process that builds the judgment you would need to check any output at all."}],"correct":true,"tempting":null}],"order":[2,3,1,0],"why":null},"prev":{"slug":"exam-hall","title":"Revision That Survives The Exam Hall","minutes":8},"next":{"slug":"past-papers-and-mark-schemes","title":"Past papers and mark schemes","minutes":9}}}