The program receives a result, and based on that, must find rules for how it should arrive at the correct answer. Example – An image recognition AI program is shown many pictures of cats and dogs and learns which pictures represent what. The AI then has to create rules for how it should be able to recognize other images that depict cats or dogs. Here the program does not get a conclusion. The AI is presented with a large amount of data and must find patterns and similarities on its own. For example, in a large number of images.
The example with the dog pictures is called supervised learning. A human monitors the system and presents a conclusion on what are pictures of dogs and what are not. Another type of learning method is called unsupervised learning. Then you ignore the first step and let the system completely on its own try to sort millions of images and categorize them based on patterns it can identify.
An important method for today’s AI systems is also the one called deep learning. It is a very complicated variant of machine learning based on so-called artificial neural networks. It is a collective name for several different self-learning algorithms that try to mimic how biological neural networks work, for example a brain.
Advanced type of machine learning powered by an artificial neural network. It is a computer network inspired by biological neural networks such as the human brain. Deep learning can be both unsupervised and supervised. Today’s advanced AI systems have usually been created through a combination of these three learning methods. You may have initially used supervised learning, but then let the system develop itself through unsupervised learning and deep learning.
Writing and drawing are just two well-known examples of what AI tools can do for us already today. But there are many more areas of use where AI can make work easier. There are apps and websites that can create music, cut a video together, enhance a podcast recording, fill a blog with content, make a Power Point presentation and much more.
These different tools are usually called generative AI. It is thus AI that can be used to generate, i.e. create something new, such as text, sound and image based on the data the tool has learned from. Says that generative AI is not really anything new, but that the tools have now become so capable that they can generate things that are of high quality and feel very human.