Welcome:
I'm Rose Luckin, Professor at University College London's Institute of Education. I've spent decades researching AI in education, watching promises made and broken, hype cycles come and go. On this course I explain what is really happening with AI in our schools, workplaces, and lives – what this means for you, and what you can do about it.
We'll start with the basics: what AI actually is, stripped of marketing speak. Then we'll tackle the questions that matter: Can AI tutors replace teachers? Are we preparing students for jobs that won't exist? Why are we teaching kids to pass tests that robots can ace?
Each week, honest analysis from someone who's been studying this technology since before it was fashionable. No breathless predictions, no doom-mongering. Just clear thinking about where we're heading.
If you're an educator, parent, or someone who wants to understand AI beyond the hype, this channel is for you. Please subscribe and spread the word.
Rose Luckin, Professor of Learner Centred Design, UCL, and Founder/CEO of Educate Ventures Research
Course Content:
Understanding AI, Through the Process of Baking
If you can make yorkshire puddings, you can understand AI! In the next 15 minutes, Rose Luckin, Professor of Artificial Intelligence, will show you how through the creation of the humble Yorkshire Pudding.
Structure, Appearance, and Context in Large Language Models
If you can bake without a recipe, you can understand AI. In the short follow-up to Episode 1's Yorkshire Pudding, Professor Rose Luckin examines structure, appearance, and context in Large Language Models (LLMS), and compares them to the Great British Bake-Off's Fondant Fancies shown on UK TV.
Neural Networks and AI
In episode 3, Professor Rose Luckin creates a beautiful Victoria Sandwich to illustrate the layers to a 'neural network', a computing system that learns to recognise patterns and make decisions by processing large amounts of data, and is inspired by the structure of the human brain.
Understanding AI's Bias
In episode 4 of the course on baking and AI, Professor Rose Luckin creates a a variety of gingerbread people and objects and explains how bias can be baked into the mould, leading to biased output.
If we've only got one cookie-cutter mould, will our AI system recognise our gingerbread people in all their different varieties?
Machine Learning and Data
In episode 5, Professor Rose Luckin makes use of a sourdough starter to create two beautiful brown sourdough loaves in the oven (and dutch oven!).
The fermentation of the sourdough mixture mirrors the iterative nature of feeding a machine learning AI system data to learn, discarding what doesn't work and refining the process over time.
Why AI Needs Feedback to Learn
In episode 6, Professor Rose Luckin bakes a beautiful jam roly-poly and custard, using the creation of the roly-poly to illustrate the light touch needed with training data for an AI system. Too heavy-handed, and the algorithm operates only one way. We need to leave space for learning, we need to go with the general principles of the training data, not try for memorisation, which could mean the AI only outputting exactly what we had very specifically asked it to memorise.
Precision in AI Development
For our 7th episode, Professor Rose Luckin visits Switzerland to explore the world of chocolate. Not only do we need to be precise when we build our AI systems, but we need to ensure that the context into which we introduce them is appropriate and correct and that they are kept in the right state. Much like chocolate melting, seemingly invisible deviations can mean our AI system ends up looking quite different.
Understanding AI's Scale
In our 8th episode, Professor Rose Luckin goes through the intricate steps required for puff pastry, illustrating with layers of dough and butter, kneading and resting, the complex processing that is required for an AI system to get up and running, and why data centres are built to power it.
Is AI Perfect?
In our 9th episode: is AI perfect? Professor Rose Luckin explores a very fussy meringue! The process through which we train our AI systems, creates something quite rigid once we fixed those weights in the neural network - but that can mean that like our pavlova and the meringue its based on, those AI systems can become brittle and easily break! Will Rose's pavlova succeed?
Hallucinations in AI
Professor Rose Luckin looks at chocolate mousse as an analogy to the training data on which generative AI systems are trained. Chocolate mousse is made using a core of chocolate and then some chocolate with cream for chocolate ganache. You whip up the mixture and inflate the thing until it's twice the size of the original ingredients. And for an answer generated by a generative AI system in response to a prompt, where do original facts begin, and the hallucinations that seem real but aren't, end?
The Right Settings for AI
Just as with choux pastry, with machine learning we have to find exactly the right settings, the right conditions and arrangements: too long in the oven and the profiterole gets burned, too little time and the pastry hasn't expanded enough. What can this tell us about how we develop our AI systems?
Human Judgement and Critical Thinking
In the 12th and final episode, Professor Rose Luckin concludes the Great British Bake Off journey with the tarte tatin, examining how human judgement, something tacit and irreplaceable, is so key to many of our endeavours. There is no algorithm for the exact amount of time it needs in the oven, for exactly how thick the caramel needs to be for the tart, our efforts are largely based on our experience and our judgement - our unique and amazing human intelligence.
Reflections on What We've Learned
Time to go back over what we learned in previous episodes!
Rose revisits the yorkshire puddings, victoria sandwich, gingerbread cookies and more, asking what we learning from each baking analogy demonstrated in each episode.