Like several generation before us we are at the edge of another technological revolution but unlike our predecessor this time the innovation has the potential to be independent from their creators via Machine Learning (ML). This “new” technology is Artificial Intelligence (AI) and whether we are prepared for it or not it is here to stay.
Machine Learning is the most significant technology of today’s world as it is the key piece that gives machines the ability to improve themselves without humans’ involvement. This is important as humans in general are able to accomplish several things that we can’t explain like recognize objects and faces, identify depth to distinguish between a painting or picture and a panoramic scene. With ML this is possible and machines can teach other machines how to accomplish the same results.
Although AI is being utilized worldwide today; the impact of AI will be more significant in the next 10 years as several industry start taking advantage of this technology including manufacturing, education, finance, insurance, retailing, healthcare, law, transportations and basically every industry. The biggest obstacle for AI today is leadership, implementation and our imagination.
Machine Learning
The simple way to understand Machine Learning is to think of it as a new or different way to creating software. In this new method a computer is able to learn from examples instead of following instructions written to achieve an expected result. This is important because as stated before we know more than we can teach or tell. This fact was described by Michael Polanyi, philosopher and polymath, in 1964 and it was given the name of Polanyi’s Paradox. In today’s AI the Polanyi’s Paradox is losing its validity as Machines are learning by examples and solving their own problems based on organized feedback.
Supervised Learning System
There are different categories of Machine Learning. After all there are different companies working in Machine Learning like IBM with Watson or Intel working into personalized healthcare, the cloud and financial market predictions. Also, Google and their AutoML language, Amazon with personalized suggestions, Boston Dynamics changing what robots can do and Tesla, Apple, Uber and more working on autonomous car to mention some. All these companies rely on AI as the core of their technology. However, most of the recent successes have been using the supervised learning system where is presented with examples of the right response to any given problem. This technique requires mapping from a set of inputs, X, to a set of outputs, Y. For example, the inputs could be faces of people and the correct responses could be labels for the type of hairs like blond hair, brown hair, black hair and bald. The supervised learning system is shown below in figure 1.
Figure 1. Supervise Learning Systems
In order for the supervised learning systems to work requires from thousands to millions of examples labeled with the right response. With this information the system is allowed to work on its own. If the training worked the system will predict responses with a high degree of accuracy. The algorithm behind the supervised learning system is called deep learning. Neural networks are utilized in deep learning, which allow supervised learning to get better prediction as more data is presented for the training.
As with everything we create there are risks. For example, machines can start communicating among themselves and create a langue of their own that will not allow us to know what they work on. In fact we are experiencing an inverse Polanyi’s Paradox where machines know more than what they can teach or tell us. For instance, machines can rejects or accept applications but they cannot explain why they arrived to that decision. Also, if the system makes mistakes it is difficult to identify where the error occurs. However, not all changes are progress but without change there is no progress. We need to find solution to our current issues to ensure we continue to grow our society.
https://futureoflife.org/background/benefits-risks-of-artificial-intelligence/
https://futurism.com/google-artificial-intelligence-built-ai/
https://www.ibm.com/watson/products-services/
https://www.bostondynamics.com/robots
https://hbr.org/cover-story/2017/07/the-business-of-artificial-intelligence