AI training and inference, in nine videos: watch guides for Jia-Bin Huang's playlist

Start here (no maths needed)

Nine short films that explain how giant AI models are trained, and how they answer millions of people quickly.

Think of it like this

Building a huge AI is like building a skyscraper. You need lots of workers sharing the job (many computers working together), smart tools (the "optimiser" that decides how to improve each step), ways to adjust the building later without rebuilding it (fine-tuning), and a good front desk to serve visitors fast (serving). Each video explains one of these jobs with animations.

How it works

1. Pick a video. They're 7 to 42 minutes long.
2. Open its watch guide here first, and read the Start-here box.
3. Watch with the guide open. The chapter table tells you what to look for.
4. Answer the self-check questions afterwards.

Why robots care

Robot brains are trained and run with exactly these tools. Fine-tuning (video 3) is how a general robot model gets adapted to your robot. Shrinking big models (video 5) and fast serving (videos 4 and 9) decide whether the robot reacts in time.

Try it

Start with the shortest video, number 7 on optimisers (about 7 minutes). Then tick it off below.

Words to know

Training

Teaching the model from data.

Inference / serving

Using the trained model to answer people.

GPU

The chip that does the maths.

Optimiser

The rule for how to nudge the model to get better each step.