Module 1: What Is AI, Really?
AI Glossary and Key Concepts
The term "artificial intelligence" has been confusing people since John McCarthy coined it in 1956. It's simultaneously the most important technology of our era and the most poorly defined. So let's fix that.
The Definition That Actually Helps
Forget dictionary definitions. Here's what AI means in practice in 2025:
AI is software that learns patterns from data and applies those patterns to new situations.
That's it. Everything else — neural networks, deep learning, transformers, large language models — is implementation detail. The core idea is: instead of programming explicit rules ("if email contains 'Nigerian prince,' mark as spam"), you show the system thousands of examples and it figures out the rules itself.
This is fundamentally different from traditional software. When you write a spreadsheet formula, you're telling the computer exactly what to do. When you train an AI, you're showing it what you want and hoping it figures out how.
The Three Things People Get Wrong
Wrong #1: "AI is one technology." AI is an umbrella term covering dozens of different technologies — machine learning, computer vision, natural language processing, robotics, expert systems, and more. Saying "AI" is like saying "vehicle" — it could mean a bicycle or a spacecraft.
Wrong #2: "AI understands things." Current AI doesn't understand anything. ChatGPT doesn't know what words mean. It predicts which word is statistically most likely to come next. The result often looks like understanding — but it's pattern matching at extraordinary scale. This distinction matters because it explains both AI's impressive capabilities and its bizarre failures.
Wrong #3: "AI is new." The field is nearly 70 years old. What's new is that it works well enough to be useful at scale, thanks to three things that converged around 2020: vastly more data (the internet), vastly more computing power (GPUs), and a breakthrough architecture (transformers). The ideas are old; the execution is new.
Complete the practical definition of AI:
The Taxonomy You Actually Need
Narrow AI (what exists): AI that does one thing well. Every AI system you've ever used is narrow AI — ChatGPT processes language, Midjourney generates images, Tesla's Autopilot drives cars (sort of). None can do what the others do.
General AI (what doesn't exist yet): AI that can do any intellectual task a human can. This is what people imagine when they hear "AI." It doesn't exist. Whether it will, and when, is genuinely debated by experts (more in the AGI course).
The stuff in between: AI systems are getting broader. GPT-4 can write, code, analyse images, and reason (somewhat). It's not general intelligence, but it's not purely narrow either. We're in an awkward middle period where the old taxonomy doesn't quite fit.
Match each common misconception about AI with the reality:
Where AI Actually Is Today
The gap between AI in marketing materials and AI in reality is enormous. Here's the honest assessment:
Genuinely impressive: Language generation, image generation, code writing, pattern recognition in large datasets, translation, transcription, game playing.
Decent but overhyped: Autonomous driving, medical diagnosis, scientific discovery, creative work (it assists, it doesn't create).
Still terrible: Common sense reasoning, understanding cause and effect, handling novel situations, anything requiring genuine understanding of the physical world.
The honest summary: AI is a very powerful pattern-matching engine being marketed as an intelligence. It's transformative for specific tasks and useless for others. Anyone who tells you otherwise is selling something.
---
The AI Basics Explainer
Explain to me what AI actually is using three different analogies: 1. An analogy a 10-year-old would understand 2. An analogy that would make sense to a business executive 3. An analogy that captures the technical reality for someone who wants to understand properly For each, also explain what the analogy gets WRONG — where does it break down?
AI Myth-Busting
I keep hearing these claims about AI. For each, tell me whether it's true, false, or more complicated than either: 1. "AI can think" 2. "AI will be smarter than humans within 5 years" 3. "AI is just statistics" 4. "AI is biased because programmers are biased" 5. "AI will take all our jobs" 6. "AI can be creative" For each, give me a one-paragraph explanation I could use in conversation.
AI in My Daily Life
Walk me through a typical day and identify every interaction with AI, including ones I probably don't notice: My day: Wake up, check phone (iPhone), commute (Google Maps), work (Microsoft Office, Slack, email), lunch (order via Deliveroo), afternoon meetings (Zoom), gym (Peloton), evening (Netflix, Instagram). For each AI interaction: what type of AI is it, what data does it use, and what would my experience be like without it?
1. Ask ChatGPT (or Claude) three questions:
- A factual question you know the answer to (test accuracy)
- A reasoning question with a tricky logical twist (test understanding)
- A question about something that happened last week (test knowledge limits)
2. Evaluate: where was the AI impressive? Where did it fail? Did it admit uncertainty or confidently give wrong answers?
3. Write one sentence: based on this test, what IS AI good at, and what ISN'T it?
This exercise calibrates your expectations better than any definition ever could.
---
- 1AI is software that learns patterns from data and applies them to new situations — everything else is implementation detail
- 2AI doesn't understand anything — it predicts patterns at extraordinary scale, which often mimics understanding
- 3All current AI is "narrow" — good at specific tasks, incapable of general intelligence
- 4The field is 70 years old; what's new is enough data, compute, and architectural breakthroughs to make it work at scale
- 5AI is a powerful pattern-matching engine marketed as intelligence — transformative for specific tasks, useless for others