To make artificial intelligence, you need :
- a training algorithm
- a powerful computer.
Let the algorithm train itself with the data, then check the result.
If you’re not satisfied, start again!
Once satisfied, you’ve got your first model.
What is training? It’s letting a program find similarities and differences between data.
What is a model? It’s a mathematical formula that generates or predicts data.
Let's picture it!
I’m looking for an image of a cat.
So I do a search for the keyword “cat” in my browser.
A “classic” algorithm will search its database to see if it has an image with the associated “cat” keyword and return it to me. On the other hand, if it has an image of a cat without the associated keyword, then it won’t show it to me in the search results.
An artificial intelligence algorithm will search and analyze certain elements in the image.
What is the shape of his ears?
Does he have a moustache?
The algorithm then compares the image with its database and selects the images it thinks correspond to a cat. The results will then become part of the search results!