Deep Learning and the Rise of AI

by David Vivancos

We live a changing world, and the things we take for granted now, a few weeks later can be torn apart, and there is no doubt that technology is playing a key role in this. (Not to dismiss many other factors)

Artificial Intelligence (AI) in particular and how to outsource any task to a computer in general, is something that has fascinated me for more than 20 years, being involved since 1995 in its research and development.

The last 3/4 years will be remembered as the rebirth of AI, the beginning of the golden years of AI, and the path to its full fledge version coined “Artificial General Intelligence” (AGI).

For example, the “hand”-written text in the image of this article has been created by a computer using a Recurrent Neural Network built by Alex Graves (now at Deepmind Technologies part of Google), something unthinkable not so long ago. (The background is a representation from MindBigData.com Algorithms)

Deep Learning is a term coined by G. E. Hinton in 2006 (now at Google), to refer to a set of computational algorithms to create models or representations of data and learning form them, enabling the understanding of underlying characteristics or features (the term is now used (or misused) in a wider sense) .

The truth is that most of the theoretical basis, was developed some decades ago, discarded at that time due to the long processing time (Centuries) needed to prove its validity, and enabled now thanks to the great processing power of current CPUs, GPUs (Graphics Processing Units) and lately ASICs (Application-Specific Integrated Circuits) too.

Together with the Algorithms and the Computing power, the other key pillar is the Data (good data) now available at large in the Net, and being built every second directly by all of us & the increasing number of Internet of Things & Quantified Self sensors. For example at MindBigData.com as today, we have +330 Million Data Points (from +1 Million Brain Signals).

The possibilities in every field are to be defined, but now in conjunction with other technologies, is enabling for example: real time & accurate decoding of voice (even language translation), or making sense without human interaction of what is inside a picture/video or the classification of unstructured text, to mention a few.

In the following years the term “smart” will surely mean something different.

The number of research institutions and companies involved in this field is growing exponentially, if you want to learn more about the wider landscape of Machine Intelligence, my recommendation is a great article by Shivon Zilis from Bloomberg Beta.

My final view on it, is that there is transcendental shift beyond this technology: “We have been creating tools since the beginning of time and soon the tools will become the next us.”