Discover Machine Learning
Recent developments in AI make it possible for the computer to understand images, text and real-world data - exposing it to a broader range of problems it's able to solve.
Different than the regular software
A traditional computer program is expressed as a set of rules, just like a cooking recipe.
The software architect collects requirements, then the engineers design efficient algorithms (rules) to develop a system fulfilling these requirements.
Most business software you use was created this way.
ML learns by examples, not by rules - just like us, humans
Not all problems can be solved this way. Some programs, like recognizing a person on an image, would have an impossibly complex and difficult set of rules for a human to write.
With Machine Learning, a set of examples (like photos of a person with a person boundary marked) is provided along with a program source code.
A model instead of source code
During the training phase, the training algorithm employs statistical reasoning to look for commonalities between the examples and builds a generalized idea of its task.
This way a model is computed.
After training, the model can be used on new data (e.g. spotting new humans on pictures).