You already use AI today — you just don't know which parts are AI
When you open M-Pesa and it warns you about a suspicious transaction, that's AI. When YouTube queues the perfect next video, that's AI. When your phone unlocks with your face, that's AI. When Google Maps routes you around traffic on Thika Road, that's AI.
So the first lesson is simple: AI is not coming. AI is already here, in your pocket, quietly running. What's changed in 2022-2026 is that a new category of AI — large language models like ChatGPT, Claude and Gemini — became good enough for ordinary people to talk to directly. That's why the whole world is suddenly paying attention.
A plain-English definition
Here's the definition that will actually serve you in conversations:
The key phrase is "without being given an explicit rule." Traditional software is a pile of IF/THEN rules written by humans: IF transaction > 100,000 AND location = foreign country THEN flag as suspicious. AI is different — you show it thousands of examples of normal and suspicious transactions, and it figures out the patterns itself. That's why it catches fraud patterns no human ever wrote a rule for.
A very short history — just the 5 moments you need
- 1950 — Alan Turing asks "can machines think?" and invents the Turing Test. Everything after this is humans chasing that question.
- 1997 — Deep Blue beats Garry Kasparov at chess. First time a machine beats the world champion at a complex human game. Everyone thought this was the moment AI had arrived. It wasn't — Deep Blue was still a rule-driven brute-force machine, not a learner.
- 2012 — ImageNet + AlexNet. A neural network called AlexNet destroys every competitor at image recognition. This is the real "AI has arrived" moment — the beginning of modern deep learning.
- 2017 — "Attention Is All You Need" — Google researchers publish a paper introducing the Transformer architecture. This is the invention that made ChatGPT possible five years later.
- 2022 — ChatGPT launches to the public. Within 5 days it has 1 million users. Within 2 months, 100 million. The first AI that felt like talking to something, not using a tool.
The three types of AI (know these)
| Type | What it does | Exists today? |
|---|---|---|
| Narrow AI | One specific task, very well | Yes. ChatGPT, M-Pesa fraud, all of it. |
| General AI (AGI) | Any task a human can do | Not yet. Maybe 2030s. |
| Super AI | Smarter than all humans combined | Science fiction (for now) |
When someone says "AI will take all our jobs!" they usually mean AGI. AGI does not exist. What exists is lots of narrow AI systems stacked together — which is already reshaping work but is not a superintelligence.
What AI is NOT (four myths, quickly)
- AI is not thinking. ChatGPT doesn't "understand" what it says. It's predicting the most likely next word based on patterns in training data. Very impressive, but not thought.
- AI is not always right. In fact, it's often confidently wrong. This is called "hallucination" and it's a huge deal — we'll cover it in Module 2.
- AI is not conscious. No AI today has feelings, opinions, or self-awareness. When ChatGPT says "I feel excited about this," it's mimicking how humans talk, not reporting an internal state.
- AI is not magic. Every AI system is math running on hardware. It has specific capabilities and specific failure modes. Treating it as magic makes you fall for its mistakes.
Why this matters for you, today
If you're a parent in Kenya: AI will shape every job your child applies for by 2030. Understanding it means you can guide them.
If you're a teacher: Your students are already using ChatGPT for homework. You need to know what it does well and where it lies, so you can teach them to use it well.
If you're a professional: Someone at your office will automate part of your job with AI in the next 12 months. Better it's you.
If you're a student: The people who learned Google Search early (2004-2006) had an advantage for a decade. Prompt engineering — the skill of talking to AI — is the new Google Search. You're reading this early.