Artificial Intelligence, Machine Learning and Deep Learning and how they Interact with the World of Today

Artificial Intelligence (AI) has come a long way from the impressive imaginations of fiction authors or the computer, Deep Blue, that beat Chess Grandmaster Kasparov in 1997. AI is something most of us utilise on a daily basis, even if we don’t know it. AI is behind a multitude of digital interfaces we interact with, from Netflix suggestions to Facebook’s facial recognition. At the other end of the spectrum we have Watson’s Health Imagery which seeks to diagnose patients accurately and efficiently. This implementation of Machine Learning, a form of AI, seeks to revolutionise the speed at which treatment can be given. The potential scope of AI ranges from simply improving the quality of our lives, to helping save them.

Google recently spent $400 million on Deep Mind a British company that was using Deep Learning to teach computers through repetitive failures and adjustments called reinforcement training. Reinforcement training allows computers to self-educate in similar to the way a baby puts objects in its mouth, to learn textures and flavours, just on a much larger and rapid scale. This purchase reflects the value the tech giants place on this AI with potentially limitless uses.

 

What is Artificial Intelligence?

Artificial intelligence aims to replicate human thought in computers. Unfortunately, creating artificial intelligence is not quite as simple as a blonde dying their hair brown. While currently there are no androids dreaming of electric sheep, there has been impressive progress in Machine Learning, a subfield of AI that is far removed from the panic educing robots of the sci-fi world. Machine learning (ML) is the kind of AI we see utilised by businesses and has produced impressive results. PwC estimates China alone will see a 26.1% growth in GDP by 2030 owing to AI, the vast majority of which will be ML, and huge portions of that ML will be Deep Learning (DL).

Deep Learning utilises artificial neural networks, these networks use algorithms to quickly analyse data and produce reliable results, while minimising human input. There are many different arenas within AI traditionally broken down into narrow and strong AI. Strong AI would be self-aware and able to create original thought – this is what Siri and Alexa pretend to function as. Siri and Alexa are still forms of narrow AI alongside machine learning, they just work to create the illusion of theory of mind. Narrow AI such as Siri and Alexia, are still reactive machines and have many business applications in terms of chat box and customer services. Machine Learning, while still a form of narrow AI opens up a near limitless number of prospects in the ever more digital business world.

What Are Neural Networks?

Under the umbrella of AI, we have Machine Learning (ML) and at the cutting edge of ML is Deep Learning (DL). ML requires the time consuming, manual entry of features which then allow the machine to learn and identify various subjects. This means the list of features cannot be hugely extensive without significant time investment. For example, to teach a machine to tell the difference between a tennis ball and a hockey ball, you would have to manually input an algorithm that identifies features such as weight and colour. Deep learning uses artificial neural networks (ANN). What these ANN’s do is function structurally like our own neurological networks. They use mathematical algorithms to skip the manual step and allow the machine to identify its features independently; these are called ‘hidden pathways’ which are used to create results. Once you have input an image these hidden pathways can be highly numerous, allowing the ANN to go very deep and produce continuously more accurate results, hence – ‘’Deep’’ learning. While the idea for ANNs is not new the raw potential uses for them has only recently been unlocked thanks partly to enormously powerful Graphics Processing Unit (GPU) originally created for intense computer gaming. 

Deep Learning – the forefront of the cutting edge

Deep Learning creates the opportunity to rapidly harness data with minimal human input, allowing more time be spent on utilising the information that ANNs can produce. For example, Watson Health Imagery is going to reduce diagnosis times dramatically allowing specialist consultants to start treating patients sooner. This pace of action is vital in treating things such as cancer. The ANNs are capable of an enormous amount of interactions to create multiple accurate predictions, meaning they are certainly at the spear-point of business innovation when it comes to improving work efficiency and creating lean production. Time saving is perhaps the most desirable benefit AI offers people in the workplace. A 2016 study in India shows that 87% of workers wanted the contribution of AI to result in time saving, opposed to 49% who wanted help getting things done.

What does Artificial Intelligence offer the businesses of the future

Looking at the types of AI being deployed by companies in 2017 we can see that Robotic Process Automation (RPA), which is essentially clerical software alongside Statistical Machine Learning are at the forefront. Then language processing such as Google Translate, closely behind. These forms of AI have been frequently used by companies in recent years and continue to be used to reduce customer contact times and speed up production. Deep Learning (DL) sits at 34%, which is incredibly impressive for how new to businesses DL is. While interest in DL is new, it has been very significant. The total value of deals surrounding mergers and acquisitions of DL shot up from 4.1 billion US dollars to 16.9 billion. What AI has to offer companies has still yet to be fully uncovered, however the big players are making sure they are able to fully capitalise on this new cognitive revolution.