# Do you need maths or coding to learn AI? · AI in Bed

https://www.aiinbed.com/blog/do-you-need-math-to-learn-ai · Published 2026-11-18

**Short answer:** Not to understand or use AI: you can learn how it works and use it well with no maths or code. To build AI models, you'll want Python and three areas of maths at roughly late-school to early-university level: linear algebra (vectors and matrices), calculus (derivatives, for gradient descent) and probability and statistics. You can learn them as you go rather than all up front.

### It depends what you want to do

- **Understand and use AI.** No maths or code. You need clear explanations and practice. This is where most people should start, and where our [Beginner level](https://www.aiinbed.com/level/beginner) lives.
- **Build with AI.** Using AI services in apps, or fine-tuning existing models: you need Python and a light touch of maths.
- **Build models from scratch or do research.** You need solid maths, mostly linear algebra, calculus and probability, plus strong programming.

### The maths, and where it shows up

Here's what each area does inside an AI model, which makes it much easier to learn:

- **Linear algebra: vectors and matrices.** AI stores words, images and users as lists of numbers called vectors. Similar things get similar vectors: that's an [embedding](https://www.aiinbed.com/learn/intermediate/embeddings). A neural network layer is mostly multiplying a vector by a matrix.
- **Calculus: derivatives.** Training means repeatedly adjusting a model to make its errors smaller. A derivative says which way to adjust each number, and [gradient descent](https://www.aiinbed.com/learn/intermediate/gradient-descent) follows it downhill. [Backpropagation](https://www.aiinbed.com/learn/intermediate/backpropagation) is the chain rule applied efficiently.
- **Probability: uncertainty.** A chatbot chooses each next word from a probability distribution. Probability also explains [temperature](https://www.aiinbed.com/glossary/temperature) and why models can be confidently wrong.
- **Statistics: measuring.** Judging whether a model is good, avoiding [overfitting](https://www.aiinbed.com/learn/intermediate/overfitting), and understanding [precision](https://www.aiinbed.com/glossary/precision) and [recall](https://www.aiinbed.com/glossary/recall) all rely on statistics.

> You rarely do these calculations by hand. Libraries like NumPy and PyTorch do them. What you need is the intuition: what a gradient means, why vectors that point the same way are similar.

### How much coding you need

Python is the standard language for AI. For most practical work you need:

- Python basics: variables, loops, functions, lists and dictionaries.
- **NumPy** for working with arrays of numbers, and **pandas** for tables of data.
- **scikit-learn** for classic machine learning, and **PyTorch** for deep learning.

That's a few months of steady practice for a beginner. You can try small runnable Python examples right in the browser in lessons like [regression](https://www.aiinbed.com/learn/intermediate/regression).

### Ways in if you're not a maths person

- **Go top-down.** fast.ai's free course starts with working models and adds theory later, which suits many people better than starting with equations.
- **Learn visually.** 3Blue1Brown's videos on neural networks, linear algebra and calculus build intuition with animation.
- **Learn the maths for a reason.** When a lesson mentions gradients, that's the moment to learn derivatives.
- **No-code tools.** Many AI tools let you train simple models or build AI workflows without code, a good way to learn the ideas.

### Free resources for the maths

- [Mathematics for Machine Learning](https://mml-book.github.io/): a free book covering exactly the linear algebra, calculus and probability ML uses.
- [3Blue1Brown](https://www.3blue1brown.com/): the Essence of Linear Algebra and Essence of Calculus video series.
- Google's [Machine Learning Crash Course](https://developers.google.com/machine-learning/crash-course): explains the maths as it's needed.

For the full path from no background to building, see [how to learn AI for free](https://www.aiinbed.com/blog/how-to-learn-ai-for-free).

### Key takeaways

- Understanding and using AI needs no maths or code.
- Building AI needs Python and a working grasp of three areas of maths.
- Libraries do the heavy calculations; you need the intuition.
- Learn the maths alongside code, when you need it.

### Questions

**Can I learn AI without maths?**

Yes, to understand and use AI. You can learn how models learn, how chatbots work and how to use AI tools well with no maths. You'll need some maths only if you want to build or research models.

**What maths do I need for machine learning?**

Mainly linear algebra (vectors and matrices), calculus (derivatives and the chain rule), and probability and statistics. Roughly late-school to first-year university level covers most practical machine learning.

**Is Python required for AI?**

Not to use AI, but Python is the standard language for building it. Nearly all major AI libraries, including PyTorch, TensorFlow, scikit-learn and Hugging Face Transformers, are used mainly from Python.

**Am I too old or too bad at maths to learn AI?**

No. Many people learn AI later in life, and the intuition matters more than calculation speed. Start with the ideas, then learn the maths slowly, alongside small coding projects.

### Sources

- [Mathematics for Machine Learning](https://mml-book.github.io/) by Deisenroth, Faisal and Ong
- [Neural networks](https://www.3blue1brown.com/topics/neural-networks) by 3Blue1Brown
- [Practical Deep Learning for Coders](https://course.fast.ai/) by fast.ai
- [Learn Python](https://www.kaggle.com/learn/python) by Kaggle
