For decades, the concept of “Artificial Intelligence” (AI) has been discussed and the impact it would have on our lives. Cinema and literature have fantasized about the possibilities and dilemmas that arise from HAL 9000 to Skynet. However, the reality was always far from fiction.
In recent years there have been a series of conditions and technological changes that have led to the resurgence of AI and its application in new areas, we refer to Big Data, Cloud, IoT, Mobility as a real new industrial revolution.
Companies are positioning themselves in this new scenario and, without a doubt, Google is one of the references in Artificial Intelligence since its origins. In this article, we are going to talk about TensorFlow, the framework released by Google to develop intelligent algorithms that are at the heart of many of its products.
TensorFlow is a computer numerical software, created by Google, oriented to Deep Learning problems. Deep Learning is a specific area of Machine Learning that is taking great relevance within the world of Artificial Intelligence and that is behind some of the most surprising technological developments of recent years.
The origin of TensorFlow is in years of Google experience in the field of Artificial Intelligence.TensorFlow was born from the work of Google Brain, a group of researchers and engineers from Google dedicated to research in the area of AI, who developed in 2011 DistBelief, the closed predecessor of TensorFlow.
In February 2017, Google released version 1.0 which incorporates many improvements. Some of the most notable are performance improvements that allow you to accelerate up to 58 times the execution times of some algorithms and take advantage of running on GPUs. In addition, integrations with other ecosystem libraries such as Keras have been improved and facilitated.
The impact that Google has achieved by releasing this technology is spectacular and the community has responded. As of May 2017 there are more than 11,000 code repositories where TensorFlow is referenced.
Applications of TensorFlow
Google has been using this technology for several years applying it to many of its products and services such as Gmail, where it is used in the Smart Reply component to generate automatic replies, or in Google Translation where it is used to make millions of translations all the days among many languages.
Another good example is here, where Google shows us in a very visual way how TensorFlow works and we can configure and see in real time how our network of neurons learns, without the need to program a single line of code, it is worth trying.
One of the things that makes TensorFlow more interesting is that Google decided to release it as free software under Apache 2 license at the end of 2015. Since then, the impact of TensorFlow in the community has been enormous becoming the cornerstone of a multitude of new Innovative products that rely on this Google technology at no additional cost. From Silicon Valley startups to large corporations are using TensorFlow in their projects. Some examples are Uber, Airbnb, Twitter, Airbus or IBM.
The applications of TensorFlow are unimaginable. In addition to technology, the advantages of this software also revert to other fields such as medicine or the arts. For example, it is being used in the detection of diabetes through retinopathies, also in the world of artistic creation for musical composition including in the detection and combination of artistic styles in painting.
Future of TensorFlow
Google executives believe that TensorFlow can be very useful for health, insurance and automotive companies. Since releasing the code, several companies use the software or have relied on it to develop applications. The insurance giant AXA already uses the software to predict (with 78% accuracy) the chances of an insured causing a serious car accident. Another popular application derived from the use of TensorFlow data ishe DeepDream image recognition engine. Thousands of startups and researchers University students bet on TensorFlow as a base to develop their own artificial intelligence systems.
In addition to altruism, Google released TensorFlow development technology to extend it in different devices. We can also say that it responds to a long-term strategy to be present in more devices and platforms. One advantage of Google’s policy is that its competitors, Amazon and Microsoft,were forced to provide artificial intelligence tools with open source. Today, offering free access tools for artificial intelligence is a company policy, especially in the technology sector.
Google has been a pioneer since its birth in the world of Artificial Intelligence, promoting research and development in this field. TensorFlow development has undoubtedly taken a step more in innovation and opening the knowledge to many companies, universities, engineers and scientists based on open technologies such as TensorFlow will achieve fascinating achievements in the coming years.
In summary, Tensorflow is an incredible tool that offers us a very powerful framework, but that greatly simplifies the internal complexity involved in the management of deep learning algorithms.