Click any card to flip it. Front: a worry you've probably heard. Back: what the actual research says.
Tap a card to see what's on the back.
It won't take your job. Someone who uses it well might. The pattern is the same as email, Excel, and Google: the tool reshuffles which skills matter. Entry-level writing, research, and analysis roles are under pressure now. The people who learn to use AI well get paid more. The people who don't end up competing against the ones who do.
One query uses roughly the energy of a 100W lightbulb for a few seconds to a few minutes. All data centers worldwide combined — search, streaming, email, AI — are about 2% of global electricity. Between now and 2030, electric cars, new factories, and air conditioning will each add more new demand than AI. Real but not dominant.
As of early 2025, 34% of US adults had already used ChatGPT — roughly double the 2023 figure. Among adults under 30, that number is 58%. The top uses: search, writing help, learning, work tasks. You do not need to be a "tech person" to use it any more than you needed to be one to use Google.
It isn't thinking. It's doing one thing very well: calculating the most probable next word based on patterns in its training. It sounds confident even when it's wrong. It doesn't know you, doesn't care, and isn't your friend. Treat it like a very fast, very confident intern — useful, checkable, not wise.
It's a pattern-matcher. Very fancy autocomplete. It has no goals, no plans, no wants. The real governance concerns are about the humans who build and deploy it — who pays for it, what data they train it on, where it gets used, and who's accountable when it's wrong. Worth watching. Not sci-fi.
AI will invent facts, citations, court cases, and statistics — all with the same confidence it uses for correct answers. It has no internal compass for truth. Any answer that matters should be verifiable. If you can't verify it, treat it as a starting point, not a finishing point. See the Three Questions.
The skipping-the-struggle part is real and worth a parent's attention. The tutoring-after-trying part is genuinely powerful. Banning AI for kids is unrealistic; teaching them to use it well is achievable. The test: "did you try first?" If yes, AI becomes a tutor. If no, they're skipping the learning.
Deepfakes are real, they're getting better fast, and they're being used in scams, elections, and harassment. The UK government projected 8 million deepfakes would be shared online in 2025, up from 500,000 in 2023 — a roughly 16x jump in two years. Deepfake-powered fraud rose 700% globally between Q1 2024 and Q1 2025. The handle: if a video makes you feel strong emotion (rage, fear, disgust), wait 30 seconds before sharing. The people hardest to scam aren't smarter — they're slower.
The basic version at chatgpt.com is free and works well for almost everything on this site. Faster and smarter models are behind a $20/month paid plan. Start with the free version. Upgrade only if you're hitting real limits.
All of them work the same way and everything on this site applies. Pick whichever one is free to you or already available in the tools you use. Claude (claude.ai) is known for clear, careful answers. Gemini (gemini.google.com) integrates with Google. Copilot lives inside Microsoft 365 if your employer has it. They're all fine. The three questions work on all of them.
Yes, and the rule is simple: don't paste anything into AI you wouldn't email to a stranger. That includes full names, social security numbers, account numbers, medical records, login credentials. AI works fine with "[NAME]" and "[ADDRESS]" — redact before you paste. Paid business accounts typically have stronger privacy; free accounts may use your conversations to improve the model.
Use it to prepare, not to decide. AI is great at explaining terminology, drafting questions for your appointment, and helping you understand what you're reading. It is not a doctor or a lawyer. The decision, and the final verification, always belongs to a professional.
For a lot of uses, it already has. For fact-checking, breaking news, and anything time-sensitive, regular search is still better — AI models have a training cutoff and don't see what happened yesterday. Use both. Different tools for different jobs.
Especially for you. The people who get the most out of AI are often not the most technical — they're the ones with real problems they want help with. If you can type a question into a text box, you can use ChatGPT. Ask one family member or friend to show you how to open it the first time. That's the whole onboarding.
Pick something you've been avoiding. A difficult email, a bill you don't understand, a document you've been putting off. Use the Three Questions. That's it. If it goes well, you'll know the feeling. If it goes badly — you just used your calibration skill. That's the point.
Email me. It's free, I'll travel within reasonable distance of Northern Virginia / DC, and I'll do remote sessions for anywhere else. robert@completeideas.com.
I run this as a free 60-minute live workshop for libraries, community centers, civic groups, senior centers, and nonprofits. You bring the room; I bring everything else.