Constitutional AI: Harmlessness from AI Feedback
Yuntao Bai et al. · arXiv preprint
arXiv:2212.08073
In short
Rather than human labels for harmful outputs, the model critiques and revises its own answers against a short list of written principles, then learns from AI-judged preferences (RL from AI feedback). The result is an assistant that stays harmless while explaining its objections instead of being evasive.
Why it matters
It showed that alignment can scale with AI feedback steered by explicit principles, and shaped how Claude is trained.
Read first
The 3 Field Guide ideas this paper leans on.
Starting from scratch? The full route 24 ideas · basics first
- Machine Learning ✓ understood
Building systems that learn patterns from data instead of following hand-written rules, getting better at a task as they see more examples.
- Reinforcement Learning ✓ understood
Learning through interaction with an environment, receiving rewards or penalties to learn optimal behavior policies.
- Dataset ✓ understood
A collection of data examples used for training, validating, or testing machine learning models.
- Training ✓ understood
The process of fitting a model to data by repeatedly measuring how wrong its outputs are and adjusting its parameters to reduce that error.
- Unsupervised Learning ✓ understood
Learning from unlabeled data to discover hidden patterns, structures, or relationships without explicit target outputs.
- Self-Supervised Learning ✓ understood
Training on unlabeled data by making the data supply its own answers: hide or alter part of an input and have the model predict or match it.
- Pre-training ✓ understood
Training a model on a large dataset (often self-supervised) before fine-tuning on specific tasks, enabling transfer learning.
- Fine-Tuning ✓ understood
The process of further training a pre-trained model on a specific dataset to adapt it for a particular task or domain.
- Natural Language Processing ✓ understood
The field of AI that lets computers read, interpret, translate and generate human language, from spam filters and search to chatbots.
- Token ✓ understood
The unit of text a language model reads and writes: a word, part of a word, or a character, drawn from the model's fixed vocabulary.
- Tokenization ✓ understood
Splitting text into tokens, usually subword pieces, and mapping each to an integer ID so a language model can process it.
- Language Modeling ✓ understood
Assigning probabilities to sequences of text, in practice by predicting each next token from the ones before it; the objective behind large language models.
- Feature ✓ understood
A single measurable property of an example, such as a house's floor area or how many links an email contains, used as an input to a model.
- Neural Network ✓ understood
A model built from layers of simple units, each taking a weighted sum of its inputs and applying a nonlinearity; training adjusts the weights so the whole stack maps inputs to useful outputs.
- Deep Learning ✓ understood
A subset of machine learning that uses neural networks with multiple layers (deep neural networks) to learn hierarchical representations of data.
- Representation Learning ✓ understood
Letting a model learn its own internal description of the data, such as embeddings or layered visual features, instead of relying on features designed by hand.
- Embedding ✓ understood
A list of numbers (a vector) that represents a word, sentence, image or other item, learned so that similar items end up close together.
- Attention Mechanism ✓ understood
A technique that lets a neural network weigh every part of its input when producing each output, focusing on the parts most relevant at that step.
- Transformer ✓ understood
A neural network architecture, introduced in 2017, built from stacked self-attention and feed-forward layers; the basis of nearly every modern large language model.
- Large Language Model ✓ understood
A neural network, almost always a transformer, trained on vast amounts of text to predict the next token, which lets it write, answer, summarize and follow instructions.
- RLHF · read first ✓ understood
Reinforcement Learning from Human Feedback - training models using human preferences to align behavior with human values.
- AI Safety ✓ understood
Research and practices aimed at ensuring AI systems are safe, reliable, and beneficial, especially as capabilities increase.
- AI Alignment · read first ✓ understood
Ensuring AI systems behave in accordance with human values and intentions, a central challenge in AI safety.
- Constitutional AI · read first ✓ understood
Training AI systems using principles and rules rather than only human feedback, developed by Anthropic for Claude.