Written by Alfonso Ibañez, Data Science manager at LUCA consulting analytics
In the era of Big Data a day does not pass without us reading some news about Artificial Intelligence, Machine Learning or Deep Learning, never really knowing what they refer to. The “experts” in the sector mix and exchange the terms with all naturalness, which only contributes to their hype. The simple fact of mentioning them catches the attention of investors and convinces them that these techniques have an almost magical power.
Machine Learning is a scientific discipline coming from Artificial Intelligence, that studies how systems can be programmed to learn and improve with experience without human intervention. To address this problem, new paradigms emerge daily that allow the discovery of knowledge based on specific data deriving from solid statistical and computational principles.
One of the approaches that is receiving more interest from the scientific community refers to the Neural Networks. These networks are inspired by the animal nervous system constituted by a system of cells that collaborate with each other to produce a response to an external stimulus.
As the topology of these systems becomes more complicated, we approach what is known as Deep Learning, a new marketing concept conceived to refer to complex neural networks. The idea under this new paradigm is that with a large number of neurons and many levels of interconnection between them, predictions can be improved in complex data sets.
The value of Deep Learning in companies
The use of Deep Learning in businesses is booming. More and more companies recognize the value of these techniques, since they allow to us work more efficiently and provide a real advantage over competition. Its emergence in the business world was favoured by the confluence of three key factors: algorithms, computing and data.
On the one hand, the algorithms are constantly growing with the continuous improvement of existing techniques and the appearance of new ones. On the other hand, the evolution of computing capacity together with the cheapening of computer equipment, have allowed the analysis of gigabytes, terabytes or even petabytes for parallel information at high speed, thus allowing us to learn and make decisions in a much more efficient and agile manner than it was possible only a few years ago.
The last factor is the access to large amounts of data with which to learn. Such data can come from multiple sources such as traditional business systems, social networks, mobile devices, the Internet of things or smart cities, among other things.
Thanks to the presence of Deep Learning in events, meetings and the press, a large part of society is fascinated by the potential of these techniques and believes that these statistical models can be the perfect solution for any complex situation. However, the reality is not as glamorous as a journalist can make it look, since it is the Data Scientists (the sexiest profession of the 21st century) who perform the “magic”. If the knowledge of the domain in question, the ability to deal with multiple data and intelligence when deciding which algorithms to use is limited, then the capacity of “learning” by the machines will also be limited.
The world as we know it is changing thanks to the potential of Deep Learning techniques, and surely with the passage of time it will be present in all aspects of our lives. According to Bill Gates we always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. We will have to wait until then to really know if Deep Learning is a reality in our daily lives or a simple hype.