NLP

Natural Language Processing

4 posts

Why GPT Can't Count: Visualizing the BPE Tokenization Trap

Interactive visualization of Byte Pair Encoding (BPE) tokenization. Discover why GPT thinks 9.11 > 9.9 and watch the tokenizer chop your text in real-time.

December 5, 20248 min read

NLPAI

The Telephone Game: Why RNNs Have Short-Term Memory

Interactive visualization of RNN vs Transformer architecture. See why RNNs forget and how Transformers solve the vanishing gradient problem.

December 5, 20249 min read

NLPAI

The Cocktail Party Effect: Visualizing Self-Attention

Interactive visualization of self-attention in transformers. See how LLMs decide which words matter using Query, Key, Value.

December 5, 20248 min read

NLPAI

The Lonely Vector: Why Good and Great Used to be Strangers

Interactive visualization of Word2Vec and word embeddings. See why good and great were strangers in one-hot encoding but neighbors in vector space.

December 5, 20247 min read