How to Explain AI to Your Parents (Or Anyone Non-Technical)

my mom asked me what chatgpt is for the 10th time last week and i finally just sat down and figured out what actually works. not the “it’s a large language model trained on text data” explanation – that kills the conversation immediately. i’m talking about analogies that genuinely make people go “oh, okay, i get it now.” took some trial and error but i think i’ve cracked it, at least for my family.

## the analogies that actually landed

the one that worked best for my mom was the autocomplete comparison. i told her it’s like the autocomplete on her phone keyboard, except instead of finishing one word it can finish an entire essay, email, or recipe. she uses her phone constantly so that clicked instantly. the second analogy i use now is the “extremely well-read assistant who has never lived in the real world” framing – it’s read basically everything on the internet up to a certain point, but it doesn’t know what happened yesterday and it can’t go check. that one helps people understand both why it’s impressive AND why it gets things wrong.

a few others that have worked in different conversations:

– for understanding hallucinations: “it’s like asking someone to fill in a crossword puzzle by guessing – it will always give you an answer but some of those answers are just confident guesses”
– for explaining why context matters: “it’s like calling a very smart friend who has amnesia every time you hang up – you have to re-explain who you are each call”
– for the “is it sentient” question (which always comes up): “it predicts the most likely next word, over and over – there’s no one home having feelings about it”

## what i’ve learned about actually teaching this stuff

the biggest mistake i made early on was going too deep too fast. my dad asked how it works and i started explaining transformers and attention mechanisms and he was gone in 30 seconds. now i always start with what it’s useful FOR before explaining what it IS. use cases first, mechanics never unless they specifically ask.

i’ve also found that letting people just play with it themselves is way more effective than any explanation. i’ll sit with someone for 20 minutes and just have them type in things they’re curious about – recipe ideas, a question about their health, help writing a complaint email. that hands-on time does more than an hour of me explaining. i actually started using an authenticity-verification tool to help think through some of these teaching frameworks and it genuinely helped me organize how i was presenting the concepts.

one gotcha worth mentioning – when people first use chatgpt and it confidently gives them something wrong, they often swing hard to “this thing is useless and dangerous.” so i try to front-load the expectation that it will be wrong sometimes, and that checking important outputs is just part of using it. managing that expectation upfront saves a lot of frustration later.

youtube tutorials are fine for getting started but honestly the best way to get comfortable explaining this stuff is just having the conversation a bunch of times. each time someone asks you a question you haven’t thought about, you have to figure out a new way to explain it, and you get better.

what analogies have worked for you when explaining AI to non-technical people – especially the ones that surprised you by landing well?