Introduction to AI

From the creator of the highly acclaimed How to DAO, The Promise of Decentralisation, Life as Practice & Introduction to web3 courses

Join Stephen Reid, teacher, technologist and creator of Dandelion, for a highly practical, four-week deep dive into the promise and pitfalls of artificial intelligence (AI) and machine learning 🦾

The course is suitable for enthusiastic beginners but will also be of interest to those with some expertise in the area seeking to connect with likeminded others. It will take place 4pm–6pm UK time/5pm–7pm Europe time on four consecutive Thursdays via Zoom:

• Thursday 11th January 2024
• Thursday 18th January 2024
• Thursday 25th January 2024
• Thursday 1st February 2024

You will be sent pre-reading/watching a few days before each session (allow 1 hour), and homework at the end of each session (allow 1 hour). To get the most out of the course you should also set aside at least 2 hours per session to go back through the notes, read some of the links provided and make sure you understand everything we've covered. So that's a minimum of 6 hours per session for maximum benefit (2 hours live, 1 hour pre-reading/watching, 2 hours revision, 1 hour homework). If you are an absolute beginner it is particularly important to do all the above.

Sessions will be recorded and shared (so you can catch up if you're late/miss one), and every participant will be offered a short private consultation during the course.

Course schedule

The schedule is subject to change to ensure maximal relevance in this fast-changing space.

Session 1: Fundamentals of AI

• AI (Artificial Intelligence), ML (Machine Learning) and DL (Deep Learning)
• The state of DL-powered AI
• Recommended newsletters and podcasts
• AI tool directories
• The best case
• History of deep learning
  • The Triple Exponential
• Data science notebooks
• Classic 1-feature regression
• Classic 1-feature binary classification
• Why are we so interested in classification?
  • Text generation
  • Image generation
• Classic 2-feature binary classification
• Real v artificial neurons
• 2-feature classification using a single perceptron
  • Geometrical interpretation of the update rule
• Limitations of perceptrons
• Artificial neural networks
  • Hidden layers
  • Non-linear activation functions
  • Backpropagation/Gradient descent
• Universal Approximation Theorem
• An example of an artificial neural network
• 'Deep learning'
  • Hierarchical Feature Learning

Session 2: Large Language Models (LLMs)

• What is a Large Language Model (LLM)?
• Tokenization, sequencing and padding
• A simple neural network for next word prediction
• Transformers
  • The attention mechanism
• LLMs
  • Open source LLMs
  • Playgrounds
  • Running LLMs locally
• Prompt engineering
• 'Hallucinations'/Fabrication/BS
• Context windows
  • Beyond context windows: Retrieval Augmented Generation
• Using ChatGPT
  • GPT4 modes
  • Upload files
  • Edit prompts
  • Stop generating
  • Regenerate
  • Searches the web by default
  • Custom instructions
  • Voice chat
• Plugins
• GPTs
  • Data Analyst
• LLM applications
  • Writing
  • Coding
  • Language learning
  • Roleplay & companionship
  • Search & summarisation
• APIs

Session 3: Multimodal AI

• Multimodal AI
• AI image generation
  • SDXL & Controlnet
  • DALLE-3: no more prompt engineering?
  • Midjourney
  • Technical
• AI voice & video
  • Cinema/animation
• AI music

Session 4: AI Futures

• Books
• Videos
• Organisations
• Life 3.0 (Max Tegmark)
  • Glossary
  • Summary
  • Aftermath scenarios
• Summary of AI risks (beyond alignment)
• Statement on AI Risk
• Limitations of AI
• Wise AI
   

About Stephen

Stephen Reid is a transdisciplinary technologist and mystical mentor devoted to the flourishing of Life on Earth who has trained in the fields of AI/ML, complexity science, physics, transformative coaching, meditation and plant medicine. Current projects include co-facilitating the Futurecraft Residency at Brave Earth in Costa Rica in December, and leading the development of the Dandelion platform for regenerative events and co-created gatherings as part of not-for-profit worker co-op Symbiota.

Previously, he founded The Psychedelic Society, taught the Life as Practice, How to DAO, Tools for the Regenerative Renaissance, The Promise of Decentralisation & Introduction to web3 courses, served as the youngest ever board member of Greenpeace UK, and consulted for organisations including Supermodular, the New Economics Foundation, the Living Wage Foundation and 350.org. Stephen has an MPhys in Physics from the University of Oxford, where he specialised in quantum field theory, an MRes in Complexity Science from the University of Bristol, and a Professional Certificate in Machine Learning and Artificial Intelligence from UC Berkeley.