07-llms

Language Models

• LLM / transformer slides

Markov Chains

• 💻 Weighted Selection p5.js example (with temperature!) along with 🚂 Weighted Selection Algorithm video
• 📚 How to Generate Text Hugging Face blog post (covers temperature, top_p, and top_k)
• 📕 Markov Chains by Victor Powell and Lewis Lehe

Sequential Data and Recurrent Neural Networks

• 📚 The Unreasonable Effectiveness of RNNs
• 🎨 Four Experiments in Handwriting with a Neural Network (Drawing)
• 📖 10 things artificial intelligence did in 2018 by Janelle Shane (Text)
• 📖 Writing with the Machine

Transformers and Large Language Models

among the reasons I use large pre-trained language models sparingly in my computer-generated poetry practice is that being able to know whose voices I'm speaking with is... actually important, as is being understanding how the output came to have its shape - @aparrish, full thread

• 📚 Watch an A.I. Learn to Write by Reading Nothing but ____ by Aatish Bhatia
• 📚 Attention is All You Need - Original "Transformer" paper from 2017, also Neural Machine Translation by Jointly Learning to Align and Translate -- Attention paper from 2014
• 📚 What Are Transformer Models and How Do They Work?
• 🎥 How large language models work, a visual intro to transformers by 3Blue1Brown
• 🎥 Intro to Large Language Models by Andrej Karpathy and Intro to LLMs slides
• 📖 Language Models Can Only Write Ransom Notes by Allison Parrish

LLM Training

• 🦙 LLaMA: Open and Efficient Foundation Language Models
• 🦙 The Llama 3 Herd of Models
• 🦜 On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
• 🔍 The Foundation Model Transparency Index
• 📖 Generative AI’s Illusory Case for Fair Use by Jacqueline Charlesworth

Datasets for LLMs

• 🔢 Common Crawl
• 🔢 The Pile

Climate Impact

• 🌏 The Internet’s Next Great Power Suck
• ⚡️ Carbon Emissions and Large Neural Network Training

Code Examples and Implementations

Markov chains

• 🚂 Markov Chain Coding Challenge video + code example
• 💻 Markov Chain p5.js code examples

Replicate

Examples will be shared over email due to use of ITP proxy server.

• 🎨 Single Prompt + Reply to Llama hosted on Replicate
• 💬 ChatBot Conversations Llama hosted on Replicate. This follows the specification in the Llama 3 Model Card.

Ollama

These examples require working with p5.js locally on your computer and outside of the web editor. Some resources for doing so can be found in my workflow video series.

• 🦙 Ollama: Run LLMs locally
• Ollama chat completion API
• 💻 In class Ollama test
• 💻 Ollama with JavaScript

Transformers.js

• Chatbot with Conversation History
• Text Completion

Assignment

• Read Language models can only write ransom notes by Allison Parrish and review the The Foundation Model Transparency Index.
• Experiment with prompting a language model in some way other than a provided interface (e.g., ChatGPT) and document the results in a blog post. You can use any of the code examples above and/or try a variety of LLMs locally with ollama. Reflect on one or more of the following questions:
  • How does the concept of LLMs as “ransom notes” influence your perception of using these models creatively?
  • How does the hidden origin of the text that LLMs generate affect your sense of authorship or originality in your creative work?
  • How does the metaphor of "collage" used to describe LLMs align with or differ from your creative process?
  • How would you compare working with an LLM to other forms of text generation, such as using a Markov chain?
• Document your experminents in a blog post and add a link to the Assignment Wiki page. Remember to include visual documentation (screenshots, GIFs, etc)