Artificial Intelligence, a Primer for Novices (like me)

0 179

I have become obsessed with artificial intelligence over the past few months (also known as AI). It is of particular interest because living in Silicon Valley provides access to many bleeding edge experiences in technological innovation. Self-driving cars are on our roads, robotic restaurants serve us our food, and even barista robots make our coffee. This new technologically enhanced world is upon us. So rather than fight the inevitable, I find it more productive to seek to understand and to consider the impact of these new technologies in my life and society. In an interview last summer with Jeremiah Owyang, industry analyst, CEO of Crowd Companies and emerging trend spotter, he shared a great piece of advice:

  it is always important to get in front of the parade rather than stay behind it

In this case that “parade” marching forward is AI and all things both artificial and augmented. So in an attempt to deepen my own understanding of the topic, I have prepared a quick primer on what I’ve learned on world of AI.

The World of AI

Artificial intelligence is really not anything new, in fact it has been something scientists, computer engineers and philosophers have been exploring since the 1940’s. The Alan Turing test, which was developed in 1950 was designed to compare the level of intelligence of a computer to that of a human. Turing posited that one day a human evaluator would not be able to distinguish conversations between a human and a machine from one between two humans.  When the responses of a computer are deemed to be indistinguishable to the responses of a human, that means it has passed the Turing test.

That day is finally upon us. We have had robot victories over humans in games such as Go, Jeopardy, Chess, and computers have passed many other tests as well, including voice-operated customer support.

The acceleration of machine learning is attributable to
many factors: one being Moore’s law which predicted that chip performance in computers would double every one to two years. That prediction has mostly held true and also impacted every aspect of technological innovation. We see that today in the smart phones most of us use daily. These magical devices hold millions of times more power than the combined computing power of the computers that NASA used in 1969, the year the first man set foot on the moon.

Technology is obviously a key driver that has gotten us to where we are today. We can now process enormous data sets and store them at reasonable prices in the cloud. Data which would have taken weeks to process previously, is now processed by algorithms to develop invaluable insights in an instant. We have made enormous advances in this field just in the past few years that are already enriching our daily lives. So let’s take a moment to examine each of the different facets of machine learning and how they’ve contributed to how we live today.

What is Machine Learning

Let’s start with a definition of machine learning. This is software that improves over time through an algorithm (or computer program) that identifies patterns in data and then predicts similar patterns in new data, learning how to make better decisions in real time without human intervention to deliver better outcomes more consistently.

Some examples of machine learning include:

  • Amazon’s personalized recommendations which take your data and learn your preferences over time, serving recommendations that make sense specifically for you.
  • Facial recognition in the form of Facebook’s auto-tagging suggestions when you put up a picture with friends, Facebook’s algorithm has learned who those people are from previous tags.
  • Fraud detection is designed to identify out of norm patterns in your regular finance activity and raise an alarm when something seems out of whack.
  • Credit scoring which looks at a range of different customer profiles and takes their existing credit scores to use that data for assessing scores of similar profiles.
  • Sentiment analysis which is able to determine whether phrases are deemed positive, negative or neutral based on the words used and natural language processing. This is often applied to social media mentions to determine overall brand or campaign sentiment.

In terms of the machine learning algorithms – there are 2 main types, supervised and unsupervised learning.

Supervised learning is when you are feeding the computer a set of data and identifying their classification, such as uploading millions of images of horses. From these millions of images we have identified as horses, the computer “learns” how to recognize a horse so that when the algorithm is given another image of a horse, it can identify it correctly. More importantly, if a new image is validated or invalidated as being a horse by the person uploading it, the algorithm learns, and gets smarter. With enough data, a computer can be trained to recognize people from horses or cars. It is what visual search is based on and why it’s so easy to find those cute kitten pictures when you need some respite from all the political conversations.

Unsupervised learning is when images are fed into the computer but there is no guidance given as to what the images represent. On its own, the computer groups the images in the way it sees fit. An example of this is where the computer is fed an assortment of animal pictures, including horses, dogs, fishes, and birds. From these various images, it will create its own groupings based on color, size or other key characteristics. This is called clustering, a term you hear a lot when it comes to unsupervised learning. The advantage unsupervised learning has is it can take massive amounts of data inputs and organize them much faster than a human.  An example of unsupervised learning you experience daily is Google search results  whether you type car or auto, you will still get similar results due to algorithms’ clustering of data as then verified by user clicks that improve the accuracy of the results.

Often, unsupervised learning works in tandem with supervised learning. Also known as “Deep Learning” which provides a foundation the AI can build upon. It basically takes layers of machine learning patterns to create a “deeper” understanding to apply to things like complex feature recognition. One of the more recent and fascinating breakthroughs in deep learning came in the form of Google Translate which was given the exercise of translating English to Korean and vice versa, English to Japanese and vice versa. Google Translate then created its own “language” of pattern and meaning to translate Japanese to Korean. The Google Translate app even includes the ability for you to point your smartphone at text in another language while traveling and receive real time translation into your chosen language.

This type of Deep Learning is achieved through neural networks which are a subset of machine learning and mimic the functions of the human brain! This gives the AI the ability to apply human-like tendencies, going beyond pattern recognition and into emotional intelligence. Recently, AI was purported to beat humans playing poker using intuition! At the bleeding edge of this type of AI are companies focused on bringing emotional intelligence to machines. Soul Machines, spun out of the University of Auckland Laboratory for Animate Technologies, is a company developing an AI which has a mandate to take emotional intelligence and combine it with logic and cognitive reasoning. The embodiment of their technology can be seen in BabyX and is expanding into life-like digital avatars that can read human emotions and react accordingly.


All of this is really heady and fascinating stuff, and will be impacting us all more and more every day before we even know it. Many futurists believe that computers will eventually take care of all of our mundane, repetitive tasks and free humans up to create art and other higher pursuits. Others warn that even highly specialized and advanced professions such as medicine, law, accounting and engineering will be at risk. Why will we need a doctor, when a robot can read our complex physiologies more quickly and prescribe, manufacture and distribute the exact blends of medications we need more accurately?

One thing I know for sure is that we all need to get ahead of this change in our societal dynamics and figure out if we are in a job that can be easily automated. Everyone needs to educate themselves on how to leverage AI to augment their work and personal lives, rather than one day being replaced by artificial intelligence.

For a deeper conversation on AI, please check out this podcast where I interview Steve Ardire, a guru in the space and AI startup advisor.

Special thanks to my geeky editing crew who helped make sure this article was on point, Chris Heuer, Stu Berman and Bryan Eisenberg!