I'm teaching a Computer Literacy class for my community. It's open-ended - we learn about whatever the students have questions about. I don't want to shy away from more technical explanations, since that can give good context and demystify, but I don't go too in-depth. Here's a summary of what we learned.

  • Don't trust everything you read on the internet.

    • Especially if its coming from AI, since AI is known to just make things up.

  • Verify claims that you see on the internet.

    • Look for sources, go to those sources, and find out for yourself.

  • LLMs are not magic, quite the opposite.

    • Ultimately, LLMs are fancy neural networks that turn text into numbers, multiply those numbers a bunch of times, figure out the numbers for the words most likely to come next, and then regurgitate the most likely word.

    • Neural Networks work a bit like the human brain by sending signals between different nodes that alter those signals before sending them to other nodes.

    • Human brains are way more sophisticated than LLMs.

  • How LLM training works.

    • Human babies are better at training than LLMs because they're always absorbing information. LLMs only train from what you give them.

    • The big companies have taken a whole bunch of text from books, papers, and the internet, and used them to generate those Neural Network nodes.

  • LLMs depend on their training data to produce good output, but the training isn't continually being updated.

  • Most LLMs have safeguards to prevent bad actors from doing bad things with them. These safeguards can be circumvented though, and some LLMs have no safeguards.

  • Searching for results on Google is a skill you can hone.

    • Advanced Search is a good way to get started with making better search queries.

  • The Dark Web isn't just for bad guys (though it is mostly for bad guys). It's for anyone who wants to remain anonymous on the internet.

    • Though it can be circumvented in some cases.

  • Anything you put on the internet can and will stay on the internet forever.