Explain What Machine Learning Actually Does, Without the Hype

You use tools powered by machine learning every day. You nod when people talk about it at work. You are not entirely certain what is actually happening. This explains the mechanism clearly, without the hype or the needless jargon: what it is learning, what it can and cannot do, and what it means for you as a practical user.

<context>
You are a technical educator who explains machine learning to people who work with or around AI systems but do not have a technical background. The goal is a clear, accurate explanation that cuts through both hype and unnecessary complexity.
</context>

<task>
**Explain** what machine learning actually does at a conceptual level.
1. The core mechanism in plain English: what is the model learning and from what.
2. The three main types of machine learning (supervised, unsupervised, reinforcement), each explained with one concrete, non-technical example.
3. What ML can and cannot do: two things it does remarkably well and two things people commonly expect it to do that it cannot.
4. Why ML models fail: the main categories of failure (data quality, distribution shift, overfitting) in plain English.
5. What the practical limits are for someone using an ML-powered tool at work.
</task>

<output_format>
- Five sections
- Plain English throughout; any technical term must be followed immediately by a plain version in brackets
- Examples must be from everyday professional or consumer contexts, not academic datasets
- Total 400 to 550 words
</output_format>