Artificial Intelligence and Big Data

Intelligence can be roughly defined as learning, understanding, and using the knowledge learned to achieve goals. So artificial intelligence can be initially, simply defined as intelligence exhibited by machines.

Artificial Intelligence (AI) is a branch of computer science that builds programs and machines that can creatively solve problems and are able to perform complex tasks. They do so by automatically combing and analyzing large amounts of observational data applying algorithms designed to find patterns.

AI and Machine Learning

Besides the proven capability of processing enormous data sets, AI is characterized by making possible for machines to learn from experience and adjust to new inputs. Machine Learning is achieved by applying iteratively complex mathematical calculations addressed to recognize patterns. AI adapts through progressive learning algorithms and finds structure and regularities in data so that the algorithms acquire a skill: The algorithm becomes a classifier or a predictor.

To do so iteration on large sets of data is essential, because when new data is provided the AI System applies what has already learned from previous computations, iterates the computation on new data and produces more reliable, results and decisions each time.

IA System learns only from the data provided. That means that any inaccuracies and obsolescence in the data set will be reflected in the results, therefore any additional layers of prediction or analysis need to be added separately.

As a result, most of today’s AI systems are trained to be limited to do a clearly defined task. These AI Systems are highly specialized: The AI System that detects for example insurance fraud, cannot identify identities on pictures of faces, even more, the system that can detect insurance fraud cannot detect tax fraud.

A.I. EVOLUTION

AI concept has been surrounded by hype since the concept was created along with the first neuronal mathematical models where created in late 1950s. Despite significant scientific advances, it was only in the 1990s that all the hype started to be justified. In the 1990s, Machine Learning became really functional and was quickly refined, leading to the massive adoption and use by industry and services by the end of the 90s.

Machine Learning and Integration

The improvement of AI accelerated during 2000s-2010s, further developing Machine Learning. However, it is only after 2010 that AI learning capacity has accelerated exponentially and expanded into a broad range of functions and tasks by integrating other new technologies.

Distributed Cloud Computing provides the massive computing power necessary to train algorithms and more complex models, sensors of Internet of Things and BigData generate and supply massive amounts of data that are necessary to train AI algorithms.

A good example of the advancing of AI by integrating new technologies are Autonomous vehicles that take advantage of innovations in sensors, mapping, machine vision, navigation algorithms, satellite technology, and robotics all brought together. Though impressive, AI being developed based on Machine Learning remain limited to solve specific problems and cannot tackle general problems in much the same way that humans do.

Learning Refinement Techniques

The learning is further improved applying innovative Data Training methods for a system to learn the relationships of a set of given inputs to a given output, this is used for example to recognize objects within an image or detect objects and describe their content. There is also the field of Reinforcement Learning to train AI System to learn from trial and error, by giving the AI System virtual rewards and punishments through a scoring system.

Presently there is a justified excitement over the exponential refinement and capability growth on Machine Learning by applying these new learning techniques. Additionally, and more importantly there is also the emerge of a new, more advanced field in AI known as Deep Learning.

A.I. NEW DEEP LEARNING

Deep Learning is the next level of machine learning, it is based on the addition of Neural Networks to the AI System, that simulate interconnected biological neurons and that model the way that neurons interact in a brain. This new neural-network provided Systems apply innovative learning techniques, which have already enormously improved them to classify, recognize, detect and describe the surrounding environment, in short providing AI Systems with something close to the capacity of understanding. Limitations remain, AI still faces many practical challenges, though new techniques are emerging to address them.

Deep Learning Integration Enablement

Besides radical improvements in algorithms, Neural networks and programming learning approaches, it is again the integration with other technical improvements that allow new efficiency and functions to AI Deep Learning Enabled machines.

Big Data plus Internet Data Streaming provide ever increasing amounts data from Internet of Things and its sensors and also from social media textual data from, plus human notes and transcripts, and with more data more neural networks can be build and more neural layers can be trained.

Cloud Computing Capacity Growth of to support deep algorithms training. Deep Learning allows State-of-the-art AI to achieve incredible accuracy which was previously impossible, for example our interactions with Google Search or Alexa are based on Deep Learning, and they become more accurate the more we use them. With Deep Learning, the algorithms are self-learning. BigData has become more important than ever, the more data the better the Deep Learning and the accuracy and complexity of tasks conducted by the AI System. More data, more competitive advantage.