In many of the classes that we teach, the first AI that we have students build is called "Guess My Mood". This is a simple example of Natural Language Processing (the type of AI that understands languages), and we have found that it is a great way to get kids of any age to appreciate how AI works.
What is the Guess my Mood AI activity?
It is a very simple AI that predicts whether you are Happy or Sad based on your mood. The dataset (crowdsourced from high school kids) has sentences that convey happiness or sadness. For example, see the below:
What is this AI exercise? What AI do they build?
The kid uses this dataset to train their AI. They then ask the AI questions and see how well it has learned the concepts of Happy and Sad. They can then "fix" the AI, adding their own examples to help it better understand Happy and Sad, or even teach it a new emotion - like Hungry. The more they interact with their AI, the smarter it becomes, and the more the AI learns to "think" like the kid doing the AI training. For example, if the kid hates chocolate, the AI can be taught that too!
The whole exercise takes less than 30 minutes.
Why is this a great exercise?
This exercise, while being very simple, intuitive and engaging, has nonetheless some very important concepts that kids can learn:
- The problem looks simple but it is not. Since the user can enter in any sentence, there is an infinite combination of possible ways to express happiness and sadness - and they do not need to include the words happy and sad! Kids can appreciate that it would be impossible to write a set of rules that factor in every combination of every word. They learn that AI is great for these kinds of problems because AI can learn patterns. Any problem where there is a pattern but it is hard to discover, is a good candidate for an AI. Any problem where the pattern is obvious (like looking for the word "sad"), does not need an AI.
- Every kid has their own opinion of moods and emotions. For example - some young children will enjoy teaching the AI that their little brother is the source of all sadness! While having fun, they will learn a second very important concept - that AI learns anything that it is taught, and that it is the responsibility of humans to teach AIs the right things. This is one of the core principles of AI Ethics.
- The first thing most kids notice is that the AI only knows Happy and Sad - because that is what it has been taught. This is a third important concept - AI knows only what has been taught. For example, an AI that has only been given examples of Happy and Sad will never identify Angry - no matter what sentence you give it. Kids learn that they cannot expect an AI to answer questions well if it has not been given enough information. This is also one of the core principles of creating safe and unbiased AIs that treat all people fairly.
How can I do this exercise myself with my students?
The full guide to building the Mood AI is here.
Comments