// Stage 9 · Safety, Evaluation & Frontier
AI for Science & Domain Applications
Graph neural networks, AlphaFold, AlphaGeometry and automated research — linked with AIMap’s “AI × Disciplines” section.
about 10 hours4 algorithms3 paperslesson text in progress
01Goals
- Understand GNN message passing
- Know the methods behind milestones such as AlphaFold
- Find suitable AI tools and datasets for your own field
Do these first: Transformers & Attention · Unsupervised Learning
02Algorithms to master
Each card has the original paper, prerequisites and resources.
03Must-read papers
04Courses & videos
- Google DeepMind Science ↗Official docsEN
- dmol.pub ↗BooksEN
- pytorch-geometric.readthedocs.io ↗Official docsEN
- Distill：A Gentle Introduction to Graph Neural Networks ↗ExplainersEN
05Hands-on: project, code & exercises
Free, official or author-published pages. Most run in the browser or on a free Colab GPU.
PROJECT
Pick a public dataset from your own field, reproduce a baseline model and write a one-page report on what AI can and cannot do there.
- 1
- 2
06Self-test
Answer in your head, then open to check.
What does each GNN layer do?
Aggregate neighbours’ information to update node representations.
AlphaFold 2’s main inputs?
The amino-acid sequence and multiple sequence alignment (plus templates).
What does the language model do in AlphaGeometry?
Propose auxiliary constructions.
On these routes: Domain Scientists