is ai reliable for learning new skills? not if you can't tell a weed from a seedling

today i taped with a founder who builds food forests in georgia.

if you've never seen one, here's the idea. you plant a central fruit or nut tree. around it you build up to six layers of other plants that support each other. after about three years it manages itself. and it keeps producing food for generations without anyone replanting it.

she looks at hunger differently from almost everyone i've heard talk about it. most people call food insecurity a distribution problem. get the food from where it is to where it isn't. she calls it a production problem.

she said the us used to have three layers of food. people grew some at home. communities grew some together. and then there were markets. covid showed how many of us now depend on only one of those layers.

then she said a line that stopped me.

"we've got all this land all across this country that we are watering and maintaining, you know, cutting the grass and trimming the ornamental shrubbery. we're not growing any food there."

think about how much effort goes into lawns that feed nobody.

she works a lot with schools, and she never charges them. one school's after-school food forest program is full. "their after-school program is at capacity of 150 kids with the 30 kids on the waiting list."

so of course i asked about ai. could a kid learn to grow food from a chatbot?

she was skeptical. and her reason is the most useful thing i've heard about ai and learning in a while.

ai is unreliable when you don't already know enough to filter what it gets wrong.

if you know gardening, ai is a great assistant. it reminds you of spacing, suggests companion plants, catches what you forgot. you can tell when it's off because you've seen the real thing.

if you know nothing, you can't tell. it says something with total confidence and you have no way to check. you pull the seedling and leave the weed.

so is ai reliable for learning new skills? it's reliable for people who already have the base. for everyone else it's a confident stranger.

that's not only true in a garden. it's true in your company.

i watch founders use ai to learn fundraising, sales, finance, law. they ask it how to structure a round and get an answer that sounds great. some of it is right. some of it is wrong in ways that cost real money. and the founders who get burned are the ones who had nothing to check it against.

the fix isn't to avoid ai. it's to build the base first. talk to people who've done it. do a few reps by hand. learn enough to recognise a weed. then let the machine speed you up.

i asked her about robots in industrial agriculture too. she shrugged. "you do you." her view is that the more machines do the farming, the more valuable it becomes to know how to grow food yourself. the knowledge gets rarer, not cheaper.

what i love about her work is that it's patient. a food forest takes about three years before it manages itself. you can't prompt your way to that. you plant, you watch, you learn what thrives and what doesn't. the knowledge comes from time in the dirt, and once you have it, ai becomes a useful helper instead of a confident guess.

long term, i think that's the pattern everywhere. ai raises the floor for people who know nothing and raises the ceiling for people who know a lot. the gap between them gets wider.

i've written about how i build that base before i hand work to the machine in the ai method.

so in the thing you're learning right now... could you spot the weed if ai handed it to you?