Skip to content
AI-grafen
Sign in
☰ Menu

The goal G Frontier Lab

Frontier Lab — an independent research project

Read and reproduce a paper, run ablations, inspect the model mechanistically, and write a report someone else can reproduce.

Knowledge nodes
87
From zero
about 99 h
Labs
6
See what you already know — no account

The diagnostic removes what you already know, so your path is usually much shorter.

What you can do afterwards

Labs along the way

You write the code. Tests you cannot see decide whether it holds up.

Lab: build and validate a benchmarkGa sandbox · about 90 minLab: find the circuit — activation patching and ablation in a small transformerGa sandbox · about 120 minProject 6: Reproduce a paper — Dropout (Srivastava et al. 2014)Ga sandbox · about 240 minLab: ablation — which component does the work?Ga sandbox · about 90 minLab: a neural network in pure NumPy — forward, backprop, gradient checkDa sandbox · about 75 minLab: dot product, norm and cosine similarityDin the browser · about 40 min

The whole path

Everything the goal builds on, grouped by level and in the order it builds on itself. Show on the map

AExplorer2 knowledge nodes
  1. Patterns and categories
  2. Sequences and precise instructions
BInvestigator6 knowledge nodes
  1. Data in everyday life
  2. Algorithmic thinking
  3. Rule-based systems and machine learning
  4. Source criticism and responsibility in AI use
  5. Training data, features and labels
  6. Classification: how a model sorts information
CBuilder9 knowledge nodes
  1. Functions and coordinate systems
  2. Programming logic — variables, conditions, loops
  3. Python — the basics
  4. Python — lists, loops and dictionaries
  5. Statistics — mean, median and spread
  6. Probability — the basics
  7. Linear regression: fitting a straight line to data
  8. Neural networks — the intuition
  9. Prompting — steering a language model
DAI developer30 knowledge nodes
  1. Discrete mathematics: graphs and relations
  2. Derivatives and optimisation
  3. Time complexity and big-O notation
  4. Data structures: lists, stacks, queues, hash tables
  5. Python — functions, scope and exceptions
  6. Python — modules, packages and virtual environments
  7. Git — version control
  8. Correlation and causation
  9. Hypotheses and falsifiability
  10. Probability distributions
  11. Loss functions
  12. The normal distribution and standardisation
  13. Samples and uncertainty
  14. Testing with pytest
  15. Tokenisation
  16. Gradient descent
  17. Vectors
  18. Matrices and matrix multiplication
  19. Linear regression with several features
  20. Neural networks — the forward pass with matrices
  21. Backpropagation
  22. Embeddings — words as vectors
  23. Attention
  24. Overfitting and generalisation
  25. Partial derivatives and the gradient
  26. PyTorch — tensors and autograd
  27. Retrieval — finding the right text
  28. Training, validation and test
  29. Train a neural network in PyTorch
  30. Transformers — the architecture
EUniversity18 knowledge nodes
  1. Graph algorithms: BFS, DFS, topological order
  2. Confidence intervals
  3. Hypothesis testing and p-values
  4. Annotation and labelling of data
  5. Reinforcement learning — the basics
  6. Convolutional networks (CNNs)
  7. Sequence models before transformers
  8. Multi-head attention in detail
  9. Language models — training and generation
  10. Model evaluation
  11. Benchmarks: what they measure and miss
  12. Fine-tuning language models
  13. RAG — retrieval-augmented generation
  14. Scientific method in AI
  15. Baselines and controls
  16. Reproducibility
  17. Experiment tracking
  18. The lab journal and the experiment log
FAI engineering11 knowledge nodes
  1. Contributing to open source
  2. Human evaluation and annotator agreement
  3. Graph neural networks (GNNs)
  4. Saliency and attribution
  5. Benchmark contamination
  6. Evals for language models and agents
  7. RLHF and preference learning
  8. AI in the natural sciences: from proteins to the climate
  9. Activations and linear probes
  10. Reading and analysing research papers
  11. Writing a technical report
GFrontier Lab11 knowledge nodes
  1. Build your own benchmark
  2. Alignment — the problems and the methods
  3. State space models and Mamba
  4. Evaluating dangerous capabilities
  5. Mechanistic interpretability — the basics
  6. Ablation studies
  7. Planning a research project of your own
  8. Circuits, ablation and activation patching
  9. Reproducing a paper
  10. Project G: an independent research project
  11. Reviewing other people's work