Here’s a great video showing the basic concepts behind machine learning. It’s done with Scratch, a platform intended for children, but that also makes it accessible for non-technical adults.
Notice how the data set is one of the most important parts of the process - and also how simple tools can be built to create a training system.
In the real world, that training data comes from either a) hiring people to do the training, or b) giving away a free service of some kind to then obtain the data. In the former category, a popular solution is Amazon Mechanical Turk (the name is a semi-obscure joke). You can hire people for far less than the prevailing industrialized minimum wage to perform classification tasks. The other option, giving away free services to get data, has been the basis for Google’s offerings for over fifteen years. Here’s a public data set from Google you can explore - and remember, this is the free stuff they are giving away!