Skip to content

Releases: ContextLab/llm-course

Winter 2026 — First Complete Offering

18 Mar 14:20
@jeremymanning jeremymanning
v1.0-winter2026
a5b23e0

Choose a tag to compare

View all tags
Winter 2026 — First Complete Offering Latest
Latest

Models of Language and Communication — Winter 2026

First complete offering of PSYC 51.17: Models of Language and Communication at Dartmouth College (January 5 – March 9, 2026).

Course Overview

A 10-week undergraduate course covering the full arc from early chatbots to frontier AI, taught through 27 lectures, 15 interactive web demos, 5 problem sets, and a final research project.

What's Included

📊 27 Lectures (Marp slides, HTML + PDF)

  • Week 1 — Introduction & String Manipulation: consciousness, ELIZA, pattern matching, rules-based chatbots
  • Week 2 — Computational Linguistics: data cleaning, tokenization (BPE, WordPiece, SentencePiece), text classification
  • Week 3 — Vibe Coding & Classical Embeddings: POS tagging, sentiment analysis, AI coding tools, LSA, LDA
  • Week 4 — Word Embeddings & Modern Methods: Word2Vec, GloVe, FastText, contextual embeddings (ELMo, USE, BERT), dimensionality reduction, cognitive models
  • Week 5 — Transformer Architecture: attention mechanisms, training (scaling laws, RLHF), RAG
  • Week 6 — Encoder Models: BERT deep dive, BERT variants (RoBERTa, ALBERT, DistilBERT, ELECTRA, ModernBERT, Gemma Encoder), language & thought
  • Week 7 — Diffusion Models: text diffusion (MDLM, LLaDA), image diffusion (latent diffusion, CLIP, DiT), video/audio generation, ethics
  • Week 8 — No classes (project work)
  • Week 9 — The Frontier: chain-of-thought & reasoning models, MoE, agents & MCP, alignment faking, interpretability, copyright, existential risk
  • Week 10 — Final project presentations

🎮 15 Interactive Web Demos

Browser-based demos covering ELIZA, tokenization, embeddings, attention, transformers, GPT playground, RAG, topic modeling, sentiment analysis, POS tagging, word analogies, semantic search, BERT MLM, embeddings comparison, and chatbot evolution.

📓 Companion Notebooks

Google Colab notebooks for hands-on exercises: ELIZA implementation, text classification, dimensionality reduction, BERT exploration, BERT variants, video/audio diffusion, chain-of-thought reasoning, MoE routing, and agent building with smolagents.

📝 6 Assignments

  1. Building the ELIZA Chatbot — Pattern matching and the ELIZA effect
  2. SPAM Classifier — Multi-method comparison (ML, neural, ensemble)
  3. Wikipedia Embeddings — 10+ embedding methods with UMAP visualization
  4. Customer Service Chatbot — RAG-based semantic search system
  5. Build GPT — Implement a GPT model from scratch (optional/extra credit)
  6. Final Project — Open-ended team research capstone

📚 100+ Paper References

Non-redundant bibliography spanning from Chomsky (1956) and Weizenbaum (1966) through DeepSeek-R1, circuit tracing, and the 2026 International AI Safety Report.

Links

Assets 2
Loading