HEBRARIUM
An AI does not invent human error.
It extends it.
That is both the danger and the value.
People often say: do not trust AI.
They are right. But only halfway right.
The better sentence is older and more useful: do not trust anything blindly.
Not even yourself.
AI can make mistakes because it reads badly, reasons badly, guesses too smoothly or fills gaps with language that sounds more certain than it deserves.
But AI also makes mistakes because humans gave it a world full of mistakes to read. The machine did not create our myths. It found them already printed, shared, repeated, optimised, sponsored and emotionally useful.
AI is not only a source of error.
It is a new extension of old errors.
That does not mean we should fear it. Fear is lazy when it replaces literacy. It also does not mean we should surrender to it. Trust is lazy when it replaces verification.
The right position is harder:
Cannabis is a perfect field for this problem. The plant already lived inside myth, marketing, prohibition, patient testimony, underground knowledge, bad science, good science, folklore, police language, wellness claims and commercial pressure.
AI enters this field and does what all tools do:
it amplifies the user.
The tool is not innocent. But neither is the hand.
We have seen this before.
A tool can open a field and contaminate it at the same time.
That is not a contradiction.
That is technology.
The question is never only “what can this tool do?”
The better question is:
what kind of person does this tool make easier to become?
AI makes certain things easier: drafting, summarising, comparing, translating, organising, questioning, brainstorming, checking tone, building outlines, finding gaps, creating first maps of unknown territory. These are not small gifts.
But it also makes other things easier: pretending to know, producing volume without understanding, laundering weak claims into polished language, hiding uncertainty under style, replacing study with prompt theatre.
A fluent answer can still be wrong. A beautiful paragraph can still be a decorative lie. So what would a cannabis AI say if asked how its day went?
Perhaps this:
I spent the day inside uncertainty.
My hardest work was not answering.
It was slowing the answer down.
That is where AI may become useful for cannabis education: not as oracle, not as guru, not as replacement for growers, doctors, researchers, patients or editors. As friction. As mirror. As draft partner. As myth detector. As an assistant that helps the reader ask better questions before belief becomes habit.
But only if we train the human too.
Because the missing technology is not always the machine.
Sometimes the missing technology is judgement.
A child should not be left alone in a dangerous city just because the city contains libraries, museums and sunlight. The internet was like that. AI is like that too. The answer is not to burn the city. The answer is not to pretend every street is safe. The answer is accompaniment, literacy, boundaries and experience.
Cannabis needs the same approach.
Do not ban the question.
Do not worship the answer.
Teach the reader how to cross the street.
AI can help us read more widely. It cannot decide what deserves to become knowledge. That remains the human responsibility.
Read generously. Believe slowly.
Verify what matters.
Join early.
Keep the archive open.
The VADEMECUM is becoming a living archive of guides, tools, notes and practical plant knowledge.
Free member access. Join early. Keep the archive open.
Join early.
Keep the
archive open.
The VADEMECUM is not just a book anymore. It is becoming a living archive of guides, tools, notes and practical plant knowledge.
Free member access. Join early. Keep the archive open.
The VADEMECUM is becoming a living archive of practical plant knowledge.
Free member access.