Artificial Intelligence : Chronicles of a #SalesDummy Series

Obscurum per obscurius or the "unknown by means of the more unknown", is exactly what we attempt to set right with this series.

Chronicles of a Sales Dummy is a series of blogs intended to break the onslaught of complex technology and solution jargons that relentlessly overwhelm sales teams. The series will take us thru the several (mis)adventures of Mr Ignotious, our sales dummy, as we try to rescue him from impending doom.

In the first article of this series, we will attempt to demystify 'Artificial Intelligence' and save Mr Ignotious from the AI bombs.

Ignotious, "What is Artificial Intelligence?"

Artificial Intelligence (AI) is the discipline of computer science that is focused on mimicking human like capabilities when it comes to thinking and decision making in machines. Thru AI, machines can today perform one or a combination of the below;

a. Learn from examples and experiences

b. Understand and respond back in natural languages

c. Recognize objects and patterns

d. Make decision and solve problems

Artificial Intelligence in many ways is the convergence of computer science, mathematics, statistics, robotics, linguistics, morality and philosophy, all into one super domain.

Fun fact for Sales Dummies to impose on unsuspecting prospects

Did you know that AI and Robots are not a new fad but rather a really old concept? Well, peek into the Greek Mythology of Talos or of the Pandora, or Indian Mythology which features Flying Chariots, Giant Robots and other automaton or of the android warriors guarding Buddha's relics, and then tell me if it isnt all references to this whole AI fad.

Ignotious, "Ok. So this is a zillion years old!"

Ignoring mythology and the attempt by philosophers describing parallels between human thinking and systems, a key moment in history dates back to 1956, a conference at Dartmouth College, New Hampshire, where the term "artificial intelligence" was coined.

Though it generated an initial buzz, due to lack of funding and more importantly due to the what we now know as the lack of computational power of machines including the availability of 'Big Data', the field entered into what is famously called the 'AI winter' (1974-1980).

There was some uptake in the 80s followed by a lackluster interest most of the next two decades. But then the signs began to appear. In 1995, ALICE the chatbot came into limelight winning several prizes for 'The most human computer'. In 1997, IBM's Deep Blue becomes the first computer to beat a chess champion, victim being the Russian grandmaster Garry Kasparov. In 2002, an autonomous vacuum cleaner called Roomba appears. 2005, an autonomous robotic car wins the DARPA. In 2011, IBM's Watson wins the quiz show Jeopardy. And since then, there has been no stopping the onslaught of AI.

The key factor that was limiting the advance of AI was not a challenge anymore. Moore’s Law, which estimates that the memory and speed of computers doubles every year, had finally reached a point where the raw compute power needed for AI to work was now available to us.

Fun fact for Sales Dummies to impose on unsuspecting prospects

AI is core to any digital conversation. And when you want to sound dope on digital, pop in these three laws to any discussion - Moore's Law (discussed above), Butter's Law (Communication speed doubles every 9 months) and Kryder's law (Storage capacity double every 13 months). Tell them without these laws, digital would just be a mountain goat dreaming of a world without war where everyone eats doughnuts for breakfast.

Ignotious, "Are you telling me Terminator is real?"

Maybe I should have clarified this much earlier but here we go. So when it comes to AI, we can classify AI into;

• Artificial Narrow Intelligence - Also knowns as Narrow or Weak AI, is that which is focused on a single narrow task. Ex. Systems diagnosing critical diseases, the Alexa in your home, the Google Lens, predictive asset maintenance applications etc
• Artificial General Intelligence - This is an emerging field where AI can be as capable as a human being.
• Artificial Super Intelligence - Stuff of the movies where AI will be able to perform better than humans when it comes to not only decision making but also arts, emotions, relationships etc.

So coming back to Terminator, thank goodness that all the AI we know of in the world today, is classified under the Artificial Narrow Intelligence. So for now, 'astalavista baby' to Terminators roaming around in our neighborhoods.

Fun fact for Sales Dummies to impose on unsuspecting prospects

If you ever get cornered by a prospect when trying to sell AI, just ask them what is their view on Roboethics. Don't worry if you have no clue about what that means. Engineers, programmers, futurists, philosophers, and AI researchers across the world are still loggerheads on this topic. So with no right answer yet, just keep nodding your head to whatever the person opposite to you says. You can always end by saying, "Your take on Roboethics. Whew... that was enlightening. I cant wait to hear what you think about Transhumanism.....". You can thank me later over a beer for this tip, when you close that deal.

Ignotious, "AI, ML. Tomayto, Tomahto ehhh"

Buckle your seatbelt Dorothy, 'cause Kansas is going bye-bye!"

To summarize the above figure, Artificial intelligence is the broader domain with Machine Learning and Deep Learning being the practices that brings solutions to life.

Machine learning is about teaching machines to learn. So instead of instructions, ML uses the philosophy of learning from examples. The three prominent ways of helping machines learn are;

• Supervised Learning - We feed the algorithm with labeled data and continue to give it several different data sets so that we reinforce learning and hence improve the outcomes. Ex. we can label pictures of cars as vehicle and people as human. We run the algorithm with this data set, with the outcome being the machine being able to identify a picture of a Ferrari as vehicle and a photo of you as human. We continually feed more and more such labeled dats to improve the accuracy to ensure that a photo of our nearest cousins, the chimpanzee or a bonobo is not identified as human.
• Unsupervised Learning - In this case, we let the learning happen with the machine finding commonalities and patterns in the input data on its own. Ex. if we feed a set of shapes as inputs, the machine will leverage certain algorithms to categorize say the input based on its characteristics. So it might put all triangles into category A, circles into category B and squares into category C.
• Reinforcement Learning - Now this is interesting. Here we use the approach of rewarding positive behavior and punishing negative behavior. Over time, this can helps the algorithm to determine the optimal behavior for a particular situation. 

Now Deep Learning as you might have assumed by now is a bit more complex to conjure. Deep learning mimics the human brain and helps in areas like detecting objects, recognizing speech, translating languages, and making decisions etc. Deep learning AI learns without human supervision, drawing from data that is both unstructured and unlabeled. Some examples of this is fraud detection, adding color to B&W pictures, autonomous vehicles, image and voice recognition etc.

Ignotious, "My head is still reeling but what was that about algorithms?"

Well, lets try and keep this at 50000 ft level. AI uses algorithms to help machines learn. Broadly these algorithms can be classified as Regression algorithms, Classification algorithms and Clustering algorithms.

This is where it hovers down to a 10000 ft level. Classification algorithms are used to divide input data into different classes and then predict the class for a given input. Regression algorithms can predict the output values based on input data points fed in the learning system. Clustering algorithms helps segregating and organizing the data into groups based on similarities within members of the group

Regression and Classification algorithms are part of supervised learning while clustering is used in unsupervised learning.

And if you can handle more, some examples of these algorithms below;

• Classification - Random forest, Decision tree, K Nearest neighbor
• Regression - Linear regression, Lasso regression
• Clustering - K-Means clustering, Fuzzy C-means clustering etc.

Now Python is....

Ignotious, "P..P...Pyth....Python!!!"

Oh, I should have known better esp. with your fear of reptiles. Let me clarify.

Now since this whole this is computer voodoo, to make things actually work, programmers use different languages. Python is the most famous of them all when it comes to machine learning. Then there are others like R, LISP, Julia, Java and Javascript.

Got it? Also, lets discuss some real world examples of the use of AI to reinforce this topic;

• Banking applications to detect credit card fraud, offer customer service etc
• In social media to detect faces and do auto-tagging
• In e-commerce sites to provide buying recommendations or movie recommendations in Netflix
• Several NLP based Chatbots applications
• Computer vision applications in self driving cars, in security cameras analyzing real time information, in retail to do smart inventory management by analyzing shelves and do placement strategies etc
• In healthcare to detect risk of cardiac arrest and other critical diseases
• Several Predictive analytics applications in Asset heavy enterprises etc

There are several more such examples but guess that's a lot for a sales dummy to digest. Guess this is enough food for thought for now.

Ignotious was then never afraid of the AI bombs. The last we heard, he was seen debating Sophia, the social humanoid robot, on the ethics of AI. Sophia seemed flabbergasted but we hope that her AI brain will survive the ordeal.

Next Episode - Click Here