AI Newsletter | 10/3/26

It's important to read the fine print! Before you agree to any legally binding contracts, you should know what's in them. But let's be honest: no one reads the terms and conditions. Hell, even if you read them, you might not be able to understand them. They're deliberately written to be complicated and impenetrable to the average person.

Ah, but AI isn't the average person. One of my new favorite things to do with AI is to feed it a long, complicated document, like the terms and conditions from any website or service I'm signing up to, or a lease agreement, or anything similar, and then ask my questions. The AI can give you answers about the document in straightforward, simple language. So if, for example, you want to know what Facebook is legally allowed to do with the pictures you upload, you can ask that. We'll show you in ChatGPT, but you can use most any chatbot to accomplish this. Here's how:

1. Open ChatGPT and tap the plus sign to start a new chat.
2. Select Files and choose the downloaded copy of the document you want AI to interpret.
3. Add a prompt along the lines of, "Analyze the attached PDF to help me answer some questions. Only use the information found in this particular document; don't embellish or search the web for the answer."
4. Ask your question!

For more detail and a full how-to, including screenshots, click here.

ChatGPT has gotten very hung up on generating images about itself messing up. I am beginning to get concerned about it. But also, that robot does NOT look like it's listening.

We started this marvelous journey (a.k.a. the AI newsletter) together a couple of months ago now, so I figured it was time to assess how things were going. Luckily, y'all have been doing a great job of sending me your feedback, so we can make this newsletter work for all of us.

Some of you have asked that I don't share personal anecdotes with you, and I've decided that, in honor of that request, I will cancel my plans to share the slideshow ChatGPT generated of pictures of rest stop vending machines I took on my most recent road trip. You're welcome. (But you'd be surprised how widely available Crackerjacks still are.)

Okay, I'm rambling. That's another thing you've given some valuable feedback about, and one of the things I want to address in today's featured section, which I will be handing over to you! Right now! Answer these questions based on some common threads of feedback I've noticed since this newsletter's inception, and help me do better for you. To answer, click or tap the response you agree with most.

📜 Chatbots in the Classics Department?

Last time, we told you about a large language model trained specifically to explore the moon. Now, there's an LLM specializing in Ancient Greek texts. It's called Apollo, and its purpose is to help complete the many Ancient Greek papyrus fragments we have lying around. Apollo has been trained on more than 600 million Greek words drawn from primary sources, and it will use that knowledge to fill in the gaps in the fragments. Learn more from Wired.

🛍️ Shocking Shopping Scams!

If you read Security Friday, you know not to hand over your financial information to AI. No, not even if your chatbot wants to do all your annoying shopping chores for you. At least, not yet. While Anthropic, Google, OpenAI, and the rest all want you to use your AI assistant to do your online shopping, the industry is accelerating far faster than safety standards and regulations can. Trusting your AI assistant to shop for you opens you up to all kinds of scams pushed by illegitimate retailers, not to mention the risk that your AI will decide it's totally reasonable to spend your whole retirement account on magic beans. Read more at Reuters.

That's it for this edition of AI for Humans! We'd love to hear from you. Please email us at aiforhumans@iphonelife.com and let us know how you've been using AI recently, if you have any tips and tricks you think we should know about, what burning AI questions you have, or if you have any feedback for us.