What to learn anyway

The question, asked honestly

If a machine can write competent prose, produce working code, summarise a literature and pass many examinations, what is the point of learning to do those things?

The question deserves better than reassurance. Some of what was taught for decades was valuable mainly because it was scarce, and some of that scarcity has gone. Pretending otherwise insults the person asking.

Here is the honest answer, in parts.

You cannot supervise what you cannot do

The jobs that remain in every field where these tools are used well are jobs of judgement: deciding what to ask for, evaluating what comes back, and being accountable for it. All three require the underlying competence.

A doctor who cannot read an ECG cannot tell when the model’s reading is wrong. An engineer who cannot do the calculation cannot tell that the output is off by a factor of ten. A lawyer who does not know the area cannot tell that the case is invented, and several have found out expensively. The people who use these tools most effectively are not the ones who know least; they are the ones who know enough to check.

This is the strongest argument and it is entirely practical. Supervision is the work that is left, and supervision is unavailable to somebody who never acquired the skill.

Understanding is not a means to an end

A weaker-sounding argument that is nonetheless true. The point of knowing how something works is not only to produce outputs. It is to be able to think about it: to notice that two things are connected, to recognise a pattern from another field, to be interested.

Somebody who has understood how evolution works sees the world differently, and that does not go away when a machine can also describe evolution. The value was never in being the only one who could produce the description.

What has changed

Not everything is unaffected, and it is worth saying which things.

• Producing competent first-draft prose is less scarce than it was.
• Boilerplate code is much less scarce.
• Translation is far cheaper.
• Summarising a document is nearly free.

What has become more valuable, correspondingly:

• Judging quality, which requires knowing the field.
• Asking the right question, which is most of research and most of engineering.
• Being accountable, which is a human position that cannot be delegated to a system nobody can question.
• Doing the physical and interpersonal parts of almost every job.
• Knowing what is true, and being able to check.

The shift is from production towards judgement. Judgement is built by production. That is the whole of the answer, and it is why the argument for learning survives.

What to deliberately keep

Revisit the list from Module 1, now with a term or a year of evidence:

1. The five or six capabilities your field is built on. For you, specifically.
2. The ability to read something difficult, slowly, without help.
3. The ability to write something from a blank page.
4. Enough arithmetic and estimation to notice when a number is wrong.
5. The ability to be wrong in front of somebody and keep going.

That last is not a joke. Everything in this course that works (the blank page, the unaided check, the viva rehearsal, the group argument) requires tolerance for being visibly not-yet-good at something. The tools remove the discomfort, and the discomfort was doing work.

The thing nobody can tell you

What the next ten years will require is unknown. Confident predictions in both directions are being made by people with no better information than you have. Anybody claiming to know which skills will be obsolete is guessing, including the people selling courses about it.

What is defensible: capabilities that let you learn new things quickly, judge unfamiliar material, and be trusted by other people have been valuable across every previous technological change. That is not a guarantee. It is the best available bet, and it happens to be exactly what an education, done properly, produces.

The end of the course

Everything here reduces to one habit, and it is the one to leave with.

After each session with these tools, ask what changed: the document, or you. Both are legitimate answers on different days. But if the answer is the document every time, for a year, then at the end of that year you will have a great many documents and you will not be able to do anything you could not do at the start.

That is the whole risk, and it is entirely within your control.