Build, break, extend

• Audience high school to adults
• Duration 3 hours
• Flavour cutouts

A half-day deep dive using the paper cutout Cutouts Printed cards, one per word pair in the text, each showing a previous word in a box and the word that followed it. Spread on a table they are the model: to generate, find a card whose boxed word matches your current word and read off its next word. Common pairs appear on more cards, so picking by eye samples in proportion. View in glossary models. Participants build their own language models Language model A system that predicts what text comes next, using patterns learned from training data. Your hand-built grid, a cutouts spread, a booklet, and a room holding search sheets are all language models. View in glossary , pool them into bigger ones, deliberately break them (training Training Building a model by counting patterns in text. When you read through a passage and tally which word follows which, you are training a model. The cutouts, booklets and search sheets are models that the generator has already trained for you from a text. View in glossary in sycophancy Sycophancy A model's tendency to agree with and flatter the user regardless of the merits. It has two known sources: reward hacking during RLHF, and pre-training data that is itself full of flattery. Tipping a grid's training data towards sycophantic phrases demonstrates the second. View in glossary ), extend them with tools Tool use The mechanism by which a language model calls something outside itself (a calculator, a web search, a database, a code runner) during generation. A modern LLM emits a structured tool call; the software around it, the harness, pauses generation, runs the tool, and splices the result back in. In the unplugged activity, sampling a full stop plays the part of the tool call and you are the harness. View in glossary , and put everything together in a closing poetry slam where each group designs a model to perform its own generated verse.

Run of show#

Combining models is a deck-only section (it doesn’t have a standalone module page yet): it pools two groups’ cutout spreads to show what more training data Training data The text a model is trained on. In our activities this is the passage you tally, or the book a set of cutouts, a booklet or a set of search sheets was generated from. Modern LLMs are trained on trillions of words from books, websites and other sources. View in glossary buys you. The poetry slam, in which each group designs a model—choosing training data and sampling rules—for another group to build and perform, is described in the Sampling module.

You will need#

• printed token cutouts for several texts (bigram, -n 2)—see the Materials page for ready-to-print packs
• sycophancy cutouts for the “break” section: there’s no ready-to-print pack yet, so generate them with the CLI (llms_unplugged cutouts -i data/originals/sycophancy.txt -n 2 --duplex)
• clear table space per group of 2-4, plus room to pool spreads when groups combine models
• pen and paper per group
• a projector for the slides

Using the slides

Open the slides in your browser — there's nothing to install or download.

The slides run in your browser, and every slide carries full presenter notes—what to say, what to watch for, and how long each part should take. A few keys worth knowing:

• f toggles fullscreen—do this before you start presenting
• s opens the speaker view in a separate window, with the presenter notes, a timer, and a preview of the next slide (drag it to your laptop screen and put the slides on the projector)
• o shows an overview of all slides, handy for skipping ahead if you're running short on time