Automatic learning is the foundation of what we now know as Machine Learning. It all starts from the theory that computers can learn to perform specific tasks without being programmed, through the use of algorithms that let them learn and recognise data.
Let us now look at how machine learning works specifically and what practical applications it has.
What does Machine Learning consist of?
Machine Learning (ML) is a branch of artificial intelligence that lets computers learn and improve without being explicitly programmed. It lets computers learn from existing data and use this information to make decisions or carry out actions in the future.
Machine Learning is based on a wide range of algorithms that let computers learn from data. These algorithms can be divided into two main categories:
- Supervised learning: the computer is given a set of input and output data and then uses this data to learn to generate similar outputs for new inputs.
- Unsupervised learning: in this type of learning, the computer is given only a set of input data and uses this data to identify patterns and relationships within the data.
The basis of learning in Machine Learning: repetition
The most important aspect of machine learning is repetition: the more the machine is exposed to data, the more chance it has of learning, studying and assimilating it on its own, strengthening its capacity for understanding and the accuracy of its answer. Computers therefore learn from the history of their processing, producing results until they are able to make reliable, repeatable decisions.
Automatic learning has many everyday applications: a classic application we all know is the voice recognition that smartphones are equipped with, which lets us trigger commands with our own voice. Home automation applications work in the same way.
Another example is the way we keep coming across products or services online that we have previously searched for: companies can therefore create advertising tied to the interests of the user, which they know through the searches the user has carried out.
dotenv’s challenges with Machine Learning: an example
Using the latest technologies and shaping them to the customer’s needs is what we at dotenv are committed to doing. The real challenge is building customised software solutions, analysing the specific use case, the needs and any critical issues.
A real example is that of a customer who needs to receive a message from its users and then publish it on social media. The complexity lies in the fact that the photos or videos must not contain sensitive material (nudity, weapons, etc.), must be in line and consistent as a reply to the message sent, and the audio of any video must not contain sensitive words.
The first challenge was to imagine the whole application process and to write the use cases; you start by listening to the need, the real requirement, and then turn it into a flow chart. From there you move on to splitting the timeline for creating the application into steps, to arrive, finally, at creating the tasks that each of us programmers carries out to get to the finished product.
It is therefore possible to upload an image or a short video of a few seconds for the system to analyse, and it will respond by reporting what it sees in the uploaded files. Through the demo, the machine is trained by the uploads and learns day after day to categorise objects better and better. What is more, thanks to this constant improvement, it can also pick out, by means of coloured areas, some of the objects directly in the photo or video (in the latter case showing them in real time).
The applications can be many, thanks to the machine’s ability to learn and return results at a computational speed that makes it possible to have them in real time. For us at dotenv, this field is highly challenging, and every day we focus on strengthening our skills, also learning from the case studies submitted to us.