I tried one of the suggested tools in the course, a Cognimates project.

The process is quite complex and I found the description lacking in that it made several assumptions about the user.

Using terms like API and URL even model, makes significant assumptions about students’ or teachers’ knowledge.

The process didn’t improve my understanding of AI and wouldn’t unless I had it set up for me so I could see the outcome of the process and ask “How does it do that?”

So, I think the activity is out of sequence for a learner.

I’ve used Seeing AI on a phone to introduce the process of understanding supervised learning by a Machine and that has been really successful as a learning hook.

Once hooked, I asked the students to take an ordinary object such as a chair or table/desk and look at what it is that makes a chair a chair. How do they know what it is? How did they learn its name?

The look at a similar, but not the same, item; or look at the same item from a different perspective. What is is that portrays chair-ness? or table-ness?

Now, can I set up a decision tree to work out it an image is that of a chair or a table?

Could I do the same with a breed of dog?

I can then introduce the idea of a model and ask does the computer learn the same way we do? What is common to our learning process? Repetition? Considering things from another’s point of view?

One of the challenges of teaching this topic in the classroom is the attempt by many websites to diagram a neural network. These diagrams usually have no meaning unless you understand what a neural network is in the first place. A bit counter-productive I think.