Home›Concepts
Concepts
Concepts
How AI works, in plain language: from the artificial neuron to the Transformer, pre-training and reasoning models.
Total 32 · Clear filters
Backpropagation and gradient descentBackpropagation is the procedure that lets a multilayer neural network learn by sending output errors backward to adjust its weights, using gradient descent. A 1986 Nature paper made it widely known, and it is standard in most neural networks today.TokenizationTokenization splits text into the tokens a language model actually reads; most modern systems do it with byte pair encoding, a 1994 compression method adapted for translation in 2015.DropoutA regularisation technique for neural networks: during training it randomly zeroes the inputs or outputs of some neurons so that units cannot co-adapt, which reduces overfitting.Batch normalizationA technique that re-centres each layer's inputs to zero and rescales them to unit variance over a mini-batch, making deep networks faster and more stable to train.Residual connectionA residual connection adds a block's input to its output, so the block learns only the change it needs to make. This makes very deep networks trainable, and it is a standard part of today's transformers.Knowledge distillationKnowledge distillation trains a small model to reproduce the outputs of a larger model or an ensemble, so that much of the predictive performance is kept at lower compute and deployment cost.Positional encodingA component of sequence models, especially transformers, that encodes each token's position as a vector, giving order-blind attention a sense of sequence; its forms range from the original transformer's sinusoids to RoPE and ALiBi.TransformerThe Transformer is a neural network architecture introduced in 2017 that relies on self-attention instead of recurrence or convolution. Most of today's large language models are built on it.Pre-training and fine-tuningPre-training trains a large model on huge amounts of unlabeled text, and fine-tuning then adapts it to a specific task with a small labeled dataset. This two-stage recipe is the basis of the term foundation model.Model parallelismModel parallelism splits one model's parameters and computation across multiple accelerators, so networks too large for a single device can still be trained. NVIDIA's 2019 Megatron-LM paper showed an intra-layer version built in native PyTorch.Scaling lawsScaling laws are empirical trends showing that a language model's test loss falls smoothly, as a power law, when parameters, data and training compute grow. They help plan large training runs and explain the race for compute.In-context learningIn-context learning is a large language model's ability to perform a new task from instructions and a few examples placed in the prompt, with no weight updates. Named and studied in the GPT-3 paper, it moved adaptation from fine-tuning to prompt writing.State space model (Mamba)A state space model represents a sequence as a hidden state that evolves with the input. S4 brought the idea into deep learning in 2021, and Mamba made its parameters input-dependent, enabling linear-time modeling of long sequences.Instruction tuningInstruction tuning is a supervised fine-tuning stage that trains a large language model to follow user instructions; the FLAN paper showed it substantially improves zero-shot performance on unseen tasks.Constitutional AIConstitutional AI is a training method introduced by Anthropic in December 2022, in which a model's behaviour is shaped by a written constitution of principles rather than by human labels of harmful outputs.Direct preference optimizationDirect preference optimization (DPO) is a method for fine-tuning language models on human preference data with a supervised-learning-style objective, removing the separate reward model and the reinforcement-learning loop of standard RLHF.Test-time computeTest-time compute is the computation a model spends while answering a prompt, beyond what training already fixed into its weights. A 2024 study by Snell et al. showed how to spend it adaptively, and how it can stand in for a bigger model.Vision-Language-Action model (VLA)A vision-language-action model (VLA) is a robot policy that takes in camera images and a language instruction and outputs actions directly, usually by adapting a pretrained vision-language model to control.Activation functionA function applied to each node's weighted sum, giving neural networks their non-linearity. Sigmoid, tanh and ReLU shaped early deep learning; GELU now dominates transformers.AI agentAn AI agent is a system that perceives its environment and acts autonomously toward goals. The term covers everything from a thermostat to LLM-driven software that directs its own tool use.