What a hot topic, AI and Data Science! We at EdTech Finland often get intriguing questions related to educational tools, solutions, digital and hybrid learning. The Edtech field is wide and deep, however, as we are very well networked with education experts, I knew who to contact with this technology education / STEAM related one. As we all know, there is no such thing as just a quick question when it is aimed at passionate experts!
The contributors to this discussion full of concrete tips on where to start are:
- Kaisu Pallaskallio, CEO, Partner, Code School Finland.
- Majella Clarke, Senior Data and AI Strategist at DAIN Studios. AI Artist.
- Otso Havanto, New media artist and maker. CEO, Mehackit. Consultant, Design & Technology, Columbia Road.
- Sanna Reponen, Technology education and e-learning professional, Product Owner, Elements of AI, University of Helsinki.
- Tiago Martins Pinto, Multimedia Artist, Mad House Helsinki. Mehackit community member.
Heini Karppinen, Chair, Edtech Finland association: Hi, I have a quick question. I got an inquiry regarding tools for teaching the basics around AI and data science in K-12 in Finland. I know you follow the developments of this closely. What would you recommend?
Otso: Excel 😅
Heini: This cannot exactly be recommended as a Finnish innovation though. Anything else? 😊
Sanna: Well, if you ask me the one I’m involved with, the Elements of AI, is pretty cool for learners 15+ years. The Elements of AI course continues with a course called Building AI. There we already go deeper in programming.
Otso: Probably The Elements of AI would be really good. There are also plenty of easy-to-use libraries out there, but they require basic programming skills. Probably not exactly what they are looking for, but I use Google Colaboratory environment with Python for my master’s thesis. I train my models with Colab as I don’t have a powerful enough computer. I use Python and the TensorFlow library. But it gets pretty deep.
Kaisu: We at Code School Finland provide a 7-lesson online AI course for kids as young as 5 years old together with their parents. My AI Robot offers a non-computer approach to understanding elements of AI – and although it is a course for children, it can also help parents to understand AI too!
Majella: When it comes to kids, young learners in particular (and all learners of all ages), I think that the learning message needs to be memorable and the learning process needs to be engaging and enjoyable. So with respect to the question, I like to use the tools from Experiments with Google to demonstrate and talk about the different aspects of AI and data.
Kaisu: Experiments with Google is also one of our top picks for services that allow experimenting with AI! It offers a collection of various simple AI experiments for anyone to start exploring machine learning, through pictures, drawings, language, music, and more. For example the online game Quick, Draw! uses neural networks to recognise doodle drawings. It’s a fun introduction to machine learning that suits all ages.
Majella: Exactly, there are several things I like about demonstrating with this toolkit. To start, you can apply the content to all ages, but simplify or vary the complexity of the message depending on the level of knowledge of the learner; Also, there are so many choices with the different types of AI to be able to demonstrate – from text to speech, computer vision and more; Another advantage of the Google experiments toolkit is the different applications of AI – from art and music, to health and the environment. Seeing Music is in my top 3 favourites with young kids.
Sanna: At Code School Finland you have AI content in the curriculum.
Kaisu: Code School Finland specializes in providing an AI curriculum for educators. Many countries’ decision makers, schools and teachers want to teach artificial intelligence to children without understanding that the AI doesn’t stand on its own. The AI content in our curriculum focuses mostly on what AI is in different contexts, the potential in AI technology and the process of machine learning.
Kaisu: Yes! The Teachable Machine provides simple, step-by-step instructions for creating machine learning models. With loads of project examples and tutorials, it’s a great starting point for experimenting with AI.
Otso: That’s good! There’s also Wekinator which is similar. With Teachable Machines you can export a TensorFlow model and use it for example with p5js. Wekinator is more a standalone desktop application and is often used for sending out OSC-data (Open Sound Control). The OSC-data can then be used to control for example a software synthesizer.
Majella: When it comes to teaching the basics of data and AI to young kids, the favourites in my top 3 also include Word Synth and Spectogram and Oscillator. They demonstrate some level of data that show the concepts of how e.g. computer vision, text to speech and sonification/visualization can be applied to produce new sound/music/song making concepts. It allows young people to learn and experiment through creativity.
Otso: There are some ready made online AI applications like Artbreeder for trying out GAN models (Generative Adversarial Networks). And a lot of interesting examples, such as This Does Not Exist. This is actually a very current topic as I’m just fine tuning a GPT-2 model with Google Colab for my thesis at Aalto Media Lab! It’s a poetry AI, trained on discussions at Untitled Festival last fall.
Sanna: Cool! You can also do some machine learning stuff with Scratch. I understand that the Code School Finland package I mentioned is related to this too.
Kaisu: Machine Learning for kids provides an extension to Scratch, the well-known visual programming environment. It provides commands and functions for creating your own machine learning projects with Scratch.
Otso: Maybe I could make a little teachable machine and a p5.js tutorial for Mehackit creative technology curriculum? I was about to do some AI tutorials with p5 anyway for a few courses I’m teaching.
Tiago: Scratch and p5.js would be my bets too! Though I loved the Excel suggestion – at the end of it all, it will be your swiss-army knife! 😀
Sanna: p5.js ❤️ I have tried to make a linear regression presentation with quality data of Portuguese wines using p5.js and TensorFlow 😃. I ran out of skills a bit, though I watched Daniel Shiffman’s supercharged Coding Train videos.
Tiago: I think the peeps haven’t talked about this one – Arduino. Indeed it raises the complexity of teaching and it demands having the components with you (so not much logistic-fit to these pandemic times…). Nevertheless, Arduino allows you, when combined with p5.js to grasp concepts of machine learning in a tangible way. At least, in my perspective, I think that it is important for someone that is learning programming concepts at an early stage, to be able to have a more concrete real-world like bridge to it. The Mehackit course Electronics and Programming Basics is a great way to start learning Arduino for the later one, build the bridge, I’m writing about.
Heini: Such great tips, thank you so much! I knew I could trust you!
Are there some other great resources that you love? Or do you have an intriguing question for us to answer? Share your thoughts in the comments section!
Got interested? Here are some opportunities for you:
- Code School Finland helps schools and teachers to include AI education in their school curriculum. Visit their website or contact Kaisu (email@example.com) to hear more about AI teaching materials and teacher professional development programs.
- Mehackit workshops, events and creative partnerships are great for individuals, companies and institutions such as museums. Mehackit’s goal for all partnerships is an impact on as diverse audiences as possible.