Decision Trees: Machine Learning

In class we learned about decision trees, calculating EMV, risk analysis. As a business student, I learned about decision trees and EMV in the context of supply chain management. This was a sparse introduction that did not even consider the implementation of decision trees in a greater capacity such as machine learning and artificial intelligence. From Hollywood to real world implementations like IBM’s Watson – AI/Machine Learning is becoming more and more relevant. Northeastern offers AI courses (CS4100/5100) and some universities even have AI degrees.  Dynamic, infinite decision trees are utilized in AI as the backbone of machine learning. Facebook was known for implementing machine learning with it’s advertising practices (now they are in legal trouble for selling information). Forbes recently wrote an article about the prevalence of artificial intelligence in our everyday lives:

https://www.forbes.com/sites/forbestechcouncil/2018/03/07/the-impact-of-artificial-intelligence-in-the-everyday-lives-of-consumers/#6a8b3a2b6f31

The article emphasizes the massive amount of data that is mined about you from daily activities. Your social media, your interactions with various smart/personal assistants such as Amazon Echo and Google Home, and your credit card use are all part of AI systems that are constantly using this data to to improve. To do this, they are using complex webs of decision trees that are constantly adjusted based on data. A perfect example is the game 20 questions – a series of questions are asked that lead to successive nodes on the decision tree and ultimately provides an answer when it reaches the end node of the tree. This is further developed when the tree provides an answer and asks the user whether it is correct. This can be seen with the online guesser – http://en.akinator.com/. Someone preconceives a celebrity (virtually any celebrity/sports/icon) and the website asks a series of questions (sometimes less than 5-6) and it will accurately guess the person you conceived and provided answers about. It is extremely accurate. Sometimes it takes more than 5-6 questions but it is very rarely incorrect. When the ‘genie’ provides the answer it asks whether it was the right person and if it was incorrect it continues to ask additional questions until it takes another guess. The backend of this system is most likely a massive decision tree and neural network. Each question moves farther down the tree and at a specific certainty level an answer is provided. Based on the responses of people these certainty levels are constantly adjusted, AKA machine learning. The system also takes data from news and search engines to add more celebs and their information. For example, a recent celebrity is the young yodeling boy who’s video went viral last week. In just 9 questions ranging from “Is this character real?”, “Is the person blonde?”, “Is the character older than 20?” the ‘genie’ guessed properly. This type of AI engine is simple and fun but is actually complex. Concepts and research taken from engines like this are being used for self-driving and medical diagnoses (IBM Watson).

The field of AI is growing at an exponential rate and the further I delve into concepts like machine learning the more I realize how interested I am.

By | 2018-04-18T03:16:39+00:00 April 17th, 2018|0 Comments