Learning AI is a great thing for students in elementary, middle, or high school. However, what platform should they learn on? What platforms and tools will enable them to build innovations that bring their imaginations to life. Which platforms can they continue to use as they get older? What technologies will help them get internships when they are in high school?
The most important things to factor in when choosing a learning platform are
Can my child get started with the coding knowledge they have? An ideal platform enables the child to learn with the platform - i.e. they can build AIs even if their coding knowledge is limited but be able to do more powerful things as their coding knowledge grows
Does the platform “fall off a cliff”? Any platform that is limited creates a “fall off a cliff” moment for a child. This is a problem for most toy tools including those that rely exclusively on kid-specific programming languages like Scratch. They are fun to use at the beginning, but once the child grows and advances, the tool is no longer effective, and using it did not help the child learn any real-world skills on professional tools.
Can they build anything they like? One of the great strengths of AI is that kids can solve real-world problems and create innovations that they can use for science fairs, competitions, and entrepreneurship. This usually requires that the platforms that they use accommodate real datasets (such as temperature data for climate change projects, MRI data for disease detection, text data for chatbots, and more). Real-world datasets also tend to be large, so the platform needs to accommodate the scale of these datasets.
Real-world platforms
Here are some common real-world platforms that professionals use to build AIs
SciKit Learn (also called sklearn). This is the most popular python package for AI - heavily focused on Machine Learning
Pandas and Numpy. These are the most popular professional packages used for processing datasets
TensorFlow. Originally created by Google - this is one of the most popular open source packages for Deep Learning and Natural Language Processing
AWS Sagemaker. This is Amazon Web Services’ flagship AI package - used on the AWS cloud by professionals
Google Cloud AI Platform. This is Google’s flagship AI package - used on the Google cloud by professionals
Any one of these would be valuable for kids to learn, but most are challenging to start with.
What we do at AIClub
AIClub has created Navigator as a transitional platform to help kids start with minimal or no coding knowledge and seamlessly transition to one or more of the professional tools listed above. The figure below shows how Navigator layers above the professional tools.
We have found that our platform provides easy entry and progressive learning.
Young kids (grades 3-6) can build AIs in as little as 10 minutes, using a simple web-based interface. In the background, however, they are using one of the packages above - the right one is automatically picked for them depending on whether the AI that they are building uses Machine Learning or Deep Learning and how much data they are using.
As kids get older, they learn Python and the core Python functions used in Data Science (such as Numpy and Pandas). When they understand some python, they can see the code that Navigator has created for them using Navigator’s code generator function. They can then take this code, run it anywhere, and also modify it to create new AIs.
Navigator also generates code in SciKit Learn (sklearn) and TensorFlow. This is a great way for middle school and high school kids to get started with these tools. They can download this code directly and run on their local computer, or open in Google Colab to run it in the cloud. They can use it as a starter to write their own code on both platforms.
As kids get yet older, our classes teach them how to build large scalable AIs in AWS or Google Colab. They also learn how to leverage GPUs for large datasets. They can take the code created by Navigator as a starting point or write their own code. Our repository of resources shows them how to handle datasets of different types, and our expert mentors help with specific advanced projects using real-world datasets. We also maintain a repository of over 400 curated and student-friendly datasets.
Our automated data processing enables kids to bring in any dataset they like and Navigator will generate code to process it. Young kids can just use Navigator automation. Older kids can take this code and change it, learning the code along the way.
We combine our platform with other tools like a variety of software development environments (repl.it, local editors like Pycharm) and online AI notebook tools like Google Colab. Students can transition seamlessly between Navigator and these environments.
We have proven this approach with over 3000 students to date:
Young kids can start early, build, customize and explore AIs with no coding
There is no “falling off a cliff”. As kids learn to code, the transition to the real world platforms is seamless. They learn sci-kit learn (sklearn), Tensorflow, AWS, and more. The skills they learn will hold up in internships, in college, and even in jobs.
They can build whatever they like. Our students have built over 500 custom projects, over 7000 AI apps, and built over 55000 AI models, all customized to their ideas.
Comentarios