teaching_llm_applications

Scaling

Resources

Large Language Model Scaling: Papers, Interactive Resources, and Further Reading

[!NOTE] Learning Goals

By the end of this topic, you should be able to:


Why Study Scaling?

The modern era of AI has been shaped by a simple observation: larger models trained on more data with more compute tend to perform better in predictable ways. Understanding these relationships has become essential for understanding the development of Large Language Models (LLMs).

Scaling laws help us answer questions such as:


Essential Reading

1. The Philosophical Foundation

Rich Sutton — The Bitter Lesson

Why read it?

This short essay provides the intellectual backdrop for modern AI. Sutton argues that throughout AI history, methods that leverage increasing computation repeatedly outperform approaches built around handcrafted human knowledge.

đź”— https://www.incompleteideas.net/IncIdeas/BitterLesson.html

[!TIP] Discussion Question

Is the success of modern LLMs evidence that Sutton was right? What examples from AI history support or challenge his argument?


2. The Original Scaling Laws

Kaplan et al. (2020) — Scaling Laws for Neural Language Models

Key idea: Language model performance follows remarkably smooth power-law relationships with:

This paper established scaling laws as a central framework for understanding AI progress.

đź”— https://arxiv.org/abs/2001.08361

[!IMPORTANT] One of the most surprising findings is that performance improvements remain highly predictable across many orders of magnitude.


3. Scaling Produces New Behaviours

Brown et al. (2020) — Language Models are Few-Shot Learners

The GPT-3 paper demonstrated that simply scaling a language model can lead to qualitatively new capabilities such as:

đź”— https://arxiv.org/abs/2005.14165

[!NOTE] This paper sparked renewed interest in the possibility that new capabilities may emerge from scale alone.


4. The Chinchilla Revision

Hoffmann et al. (2022) — Training Compute-Optimal Large Language Models

This paper challenged prevailing assumptions about model scaling.

The central argument:

đź”— https://arxiv.org/abs/2203.15556

[!WARNING] Bigger models are not necessarily better models. Training strategy matters.


5. Revisiting Chinchilla

Epoch AI — Chinchilla Scaling: A Replication Attempt

An important follow-up study that revisits and tests the Chinchilla conclusions using a much larger collection of models.

đź”— https://epoch.ai/publications/chinchilla-scaling-a-replication-attempt

[!TIP] Ask yourself:


Interactive Resources

[!NOTE] These resources are ideal for tutorials, lab sessions, and self-study.


AI Safety Course — Chapter 1: Capabilities

A highly accessible introduction to scaling and capability development.

Features:

đź”— https://ai-safety-course.github.io/chapters/chapter-1/


Epoch AI Distributed Training Simulator

An interactive simulator that allows you to explore:

đź”— https://epoch.ai/latest/introducing-the-distributed-training-interactive-simulator

[!TIP] Classroom Activity:

Have students estimate the infrastructure required to train a GPT-4-scale model and compare assumptions.


Data Bottleneck Explorer

Epoch AI — Will We Run Out of Data?

Investigates a major question for the future of scaling:

What happens when we run out of high-quality human-generated text?

đź”— https://epoch.ai/publications/will-we-run-out-of-data-limits-of-llm-scaling-based-on-human-generated-data

[!IMPORTANT] Future AI progress may be constrained by data availability rather than compute.


Hands-On Notebooks

Empirical Scaling Harness

Students can:

đź”— https://github.com/mmcmanus1/empirical-scaling-harness

[!TIP] Recommended mini-project:

Replicate a scaling-law curve using a subset of experiments and compare your fitted exponent with the published result.


Reasoning Scaling Law

A research-oriented notebook exploring scaling behaviour in reasoning tasks.

Students can:

đź”— https://github.com/WANGXinyiLinda/reasoning-scaling-law


Chinchilla’s Wild Implications

An accessible explanation of the practical consequences of compute-optimal scaling.

Includes:

đź”— https://www.alignmentforum.org/posts/6Fpvch8RR29qLEWNH/chinchilla-s-wild-implications


Further Reading

AI Safety, Ethics, and Society Textbook

Section 2.4 — Scaling Laws

A concise pedagogical introduction with exercises and review questions.

đź”— https://www.aisafetybook.com/textbook/scaling-laws


JAX Scaling Book

How To Scale Your Model

A practical guide to large-scale training systems.

Topics include:

đź”— https://jax-ml.github.io/scaling-book/


Suggested Lecture Flow

Part 1 — Why Scaling Matters

  1. The Bitter Lesson
  2. Historical trends in AI

Part 2 — Empirical Scaling Laws

  1. Kaplan et al.
  2. Power-law behaviour
  3. Predictability of performance

Part 3 — Emergent Capabilities

  1. GPT-3
  2. Few-shot learning
  3. In-context learning

Part 4 — Compute-Optimal Training

  1. Chinchilla
  2. Parameters vs. tokens
  3. Replication studies

Part 5 — Limits of Scaling

  1. Compute bottlenecks
  2. Energy constraints
  3. Data limitations

Part 6 — Hands-On Exploration

  1. Scaling-law notebook
  2. Interactive simulator
  3. Student experiments

Reflection Questions

[!QUESTION]

  1. Why do power laws appear so frequently in modern AI?
  2. Is scaling discovering intelligence or merely exploiting statistics more efficiently?
  3. What might eventually limit further scaling?
  4. Does scaling alone lead to reasoning and abstraction?
  5. How do scaling laws relate to François Chollet’s arguments in ARC and the measurement of intelligence?
  6. Are there domains where scaling may fail entirely?

[!SUMMARY]

The central story of modern AI can be viewed as:

The Bitter Lesson → Scaling Laws → GPT-3 → Chinchilla → Data and Compute Limits

Understanding this progression provides a foundation for understanding both the successes and the future challenges of large language models.

🎮 💡Practicals