Invisible AI

The most successful AI is the kind you don’t notice. You can’t see it, you don’t know it’s there, but if you removed it, your daily life would change dramatically. Today we’ll tour all the AI applications you’re probably already using.

Your phone is an AI factory

Every time you take a photo, your phone fires up several AI models simultaneously.

Night mode: When you take a photo in low light, the camera doesn’t just crank up the brightness. An AI model analyses the image, separates signal from noise, and reconstructs details the human eye can barely see. That’s why night photos on modern phones look like miracles — they are, computational ones.

Portrait mode (background blur): Your phone has a small, flat camera that physically can’t blur the background like a professional camera. An AI model detects where you are and where the background is, and applies the blur artificially. So well that most people can’t tell the difference.

Facial recognition: When you unlock your phone with your face, an AI model compares your face to the one it has stored. And it’s not a simple photo comparison — it works from different angles, with or without a beard, with glasses, at night.

Predictive keyboard: Every time you type a message, the keyboard predicts the next word. This is a small language model trained on millions of texts. It’s not as powerful as ChatGPT, but the underlying idea is the same: predict the most likely word.

Google Maps / Waze: They don’t just calculate the shortest route — they predict future traffic. They analyse data from millions of phones in real time, combine it with historical patterns, and calculate arrival times with surprising accuracy. This is pure Machine Learning.

Uber / Lyft: The trip price isn’t fixed — an AI model adjusts it dynamically based on demand, traffic, time, and area.

Assisted driving: Cars like Tesla, Mercedes, or BMW use Deep Learning models to process camera images and detect lanes, cars, pedestrians, and signs.

Entertainment

Netflix: 80% of what people watch on Netflix comes from AI recommendations. The model analyses what you’ve watched, for how long, when you stopped, what you skipped, and compares it with millions of other users. Even the thumbnails you see are personalised.

Spotify / YouTube Music: Spotify’s Discover Weekly is a brilliant example of AI. Every Monday you get a list of 30 songs you’ll probably like. The model analyses your listening history, finds users with similar tastes, and offers you what they listen to that you haven’t discovered yet.

YouTube: YouTube’s recommendation algorithm is one of the most powerful (and controversial) AI models in the world. It decides what shows up on your home page and what’s suggested after each video.

TikTok: TikTok’s “For You” is recommendation AI taken to the extreme. It analyses not just what you watch, but how you watch it: how long you view each video, whether you rewatch it, share it, or comment on it.

Communication

Spam filter: One of the oldest and most successful examples of everyday AI. It analyses content, sender, formatting patterns, and links to decide whether an email is legitimate or junk.

Machine translation: Google Translate went from being a joke to being surprisingly useful thanks to Deep Learning. Since 2016 it uses a neural network that translates whole sentences instead of word by word. DeepL raised the bar even further.

Voice assistants: Siri, Alexa, and Google Assistant combine several AI models: one to understand your voice, one to interpret what you want, one to generate the response, and one to convert it to speech.

Shopping and finance

Amazon: The “customers who bought this also bought…” recommendations are collaborative filtering AI. But Amazon goes much further: it uses AI to decide which products to show you first, what price to offer, and how to organise its warehouses.

Banking fraud detection: If your credit card gets blocked when you make an unusual purchase, it’s because an AI model detected an unusual pattern in milliseconds.

Healthcare

Medical imaging diagnosis: Deep Learning models can detect tumours in X-rays and mammograms with accuracy comparable to expert radiologists. They don’t replace doctors — but they help them avoid missing anomalies.

The question you’re left with

After seeing all this, you might wonder: if AI already does so many things for me, why do I need to understand it?

Because there’s a difference between being a passive user of AI that others have designed for you, and being an active user who understands how it works, chooses which tools to use, and even creates their own agents.

That’s the difference we’re building in this course.

Key concepts today

  • Invisible AI: The most successful AI is the kind you don’t notice because it’s integrated into everyday products
  • Recommendation: Systems that predict what you’ll like based on your behaviour and that of others
  • Collaborative filtering: Recommending things based on what similar users do
  • Speech-to-text / Text-to-speech: Converting voice to text and vice versa — both are AI models

Next lesson: We start Week 2 — What is an LLM? The model that understands (and generates) text.