AI in plain words

Why you should care: you have already been using AI for years without noticing — the spam filter that guards your inbox, the map that reroutes you around traffic, the keyboard that finishes your sentences. What changed recently is not that AI appeared. What changed is that it learned to talk. And a tool you can talk to is a tool you can work with — this course is about learning exactly that.

What “AI” actually means today

Forget the movie robots. The AI you keep hearing about — ChatGPT, Claude, Gemini — is something much more specific and, honestly, more useful:

A system that read an enormous amount of human writing and learned to predict what comes next. Ask it something, and it writes the most plausible continuation — which, it turns out, is usually a pretty good answer.

That’s the whole trick. Not a conscious mind, not a search engine: a very well-read pattern machine. Everything else in this course — chats, prompts, agents, tools — is different ways of pointing that machine at your problem.

Here is that trick, running in front of you. The model has read the beginning of a sentence, and every word below is a candidate for what comes next:

Press “sample” a few times at a low temperature: you’ll get the same word almost every time. Drag the slider up and the unlikely ones start showing up. That is the whole engine — one guess after another — and it is also why the same question can get a different answer tomorrow.

Sixty years in six sentences

You don’t need the history to use AI, but a rough timeline makes the current moment feel less like magic:

• 1950 — Alan Turing asks “can machines think?” and proposes a test: a judge holds a written conversation with a hidden human and a hidden machine, and if the judge cannot tell which is which, the machine passes.
• 1956 — the term “artificial intelligence” is coined at a summer workshop. Decades of optimism, hype, and disappointment (the “AI winters”) follow.
• 1997 — Deep Blue beats the world chess champion. Impressive, but it won by searching through millions of positions and scoring them with rules its programmers had tuned by hand — it never learned from experience.
• 2012 — neural networks suddenly get good at seeing (recognizing images). The key ingredient turned out to be not cleverer math but more data and more computing power.
• 2017 — researchers publish the architecture behind modern AI (the “T” in GPT stands for its name, Transformer). Machines get genuinely good at language.
• 2022 — ChatGPT opens to the public and, by outside estimates, reaches something like 100 million users within a couple of months — one of the fastest adoptions of a consumer product seen at the time. AI stops being a lab topic and becomes a household tool.

The period that started in 2022 is the one we are still in, and nobody — not the researchers, not the companies — fully knows where it goes. That uncertainty is exactly why learning to use the tool now pays off.

What it is genuinely good at

• Explaining. Anything, at any level of depth, in any tone you ask for. “Explain like I’m five” actually works.
• Writing and rewriting. Emails, summaries, plans, translations — drafts in seconds, and you keep editorial control.
• Working with code. Writing it, explaining it, finding mistakes. This is the superpower this course leans on hardest.
• Thinking with you. Brainstorming, weighing options, playing devil’s advocate when you’re stuck.

What it is honestly bad at

• Knowing things reliably. It predicts plausible text, not verified facts. It can state something wrong with complete confidence. Always keep your skepticism switched on.
• Counting and exact math in its head. Ask it to reason step by step or use tools, and it improves dramatically.
• Knowing today’s news out of the box. Its knowledge has a cutoff date unless the tool can search the web.

One more pitfall is not a property of the model but of how you use it: hand it a whole task blindly and you get a result you can’t judge. (Lesson two of this course is entirely about that.)

What people actually do with it

A few real, unglamorous examples — maybe you’ll recognize yourself:

• draft a difficult message to a landlord, a client, a school teacher
• turn a pile of notes into a summary and a plan
• understand a confusing letter, contract, or medical report
• build a small website without knowing how websites work (that’s your first project later in this course)
• learn anything: the best prompt is often just “teach me this”

Try it right now — five minutes, no course needed

Reading about AI is like reading about swimming. Open any of these in a browser — they are free to start and ask nothing but an email:

• chatgpt.com — the most well-known assistant
• claude.ai — strong at writing and careful reasoning
• gemini.google.com — Google’s assistant

Then ask something you actually care about. Not “hello” — a real question:

• “Explain how a mortgage works, like I’m ten years old.”
• “I have chicken, rice and a lemon. What can I cook tonight?”
• “I want to learn guitar as an adult. Give me a two-week plan.”

Notice how the answer changes when you add a detail (“shorter”, “simpler words”, “as a table”). That little dance — ask, adjust, ask again — is 80% of using AI well, and it is exactly where the next lesson picks up.

If you want to go deeper (optional)

• Wikipedia: History of artificial intelligence — the full timeline, honestly told
• 3Blue1Brown — beautiful visual explanations of how neural networks learn
• Andrej Karpathy’s channel — a former OpenAI/Tesla researcher explaining LLMs for normal humans

None of this is required. The course stands on its own; the links are for the curious.