An Eye on the Future
Published on : Monday 03-04-2023
AI is not limited to a specific industry or domain, for that matter. Harish Rijhwani on why we need to keep ‘An eye’ (AI).
The Second World War was at its peak, and there wasn't any end in sight. On one side, the Germans sent messages encrypted by the Enigma machine; with 159 quintillion (one followed by 18 zeros) combinations (settings of the device were changed daily). However, on the other hand, Alan Turing and his team had built a machine named Bombe (and later Tunny), which was used to break the unbreakable code. If you have seen the movie ‘The Imitation Game’, you will notice Turing's team was able to crack the code because of a few things. One, they found a German code book from a captured U-Boat; second, they intercepted a message with encrypted weather information sent daily at a particular time. They also found that every communication started and ended with the same ten characters. There are no guesses for what those words were and ultimately using this information and Baye's theorem Alan Turing and his team were able to crack the Enigma Code.
The above is a story from the 1940s, one of the earliest examples of Data Science and Machine Learning. Let's move a little further to the 1960s, where we talk about Human computers leveraged by NASA. How about I reference another movie? Any guesses? If you haven't guessed, it is ‘Hidden Figures’, which showcases Katherine Johnson (one of the main characters), a mathematician often referred to as a ‘Computer’. At the movie's end, we see an IBM computer installed at NASA. It indicated that people purely hired as Mathematicians/Human Calculators would not be needed. But, the movie also showcased Dorothy Vaughan picking up a book to learn – Fortran – one of the earliest programming languages. What am I trying to explain? The scene, in its essence, indicated a transition from one job role to another, from doing things manually to learning how to do the same (and more) quickly using a computer.
Historically there are similar examples of how technology has brought about a revolution (but not without a business need) from the advent of electricity, light bulb, television, telephone, and computers, to mobile phones and smartwatches. Artificial Intelligence (AI) is on the same path of revolutionising many fields of work. But, if you are from a non-technical background, you might wonder what this AI means. Let me give you a simple example, which you might follow daily; I am referring to Cooking. Let's assume we want to bake a Chocolate Cake, and we are doing so for the first time. Mostly, we won't get it right the first time; for example, we might get the quantity of certain ingredients wrong. But, the next time we bake a cake, we will try not to make the same mistake (though we might make another), and after multiple attempts, we will reach our desired cake quality. In the last few lines, I explained a classic example of human behavior, which we mathematically call a ‘Neural Network’ (a Machine Learning Algorithm).
The biggest challenge for a computer is that they converse in the binary language (1s and 0s), which is why they understand numbers and code much better than free text. But with technological growth, we have seen applications of AI and Machine Learning in various domains/industries. So let us look at a few examples to understand what the future holds for us.
Conversational AI
If you are a movie buff like me, then the most relatable example for me to give is that of Jarvis, Iron Man's close friend and computer. But, of course, a real-life example is that of Alexa/Siri, where one asks a question and gets a response you would get from a human. By the way, Siri is also referenced by Rajesh Koothrappali in the Big Bang Theory. All these examples seem new, but do you know conversational AI was first conceptualised and implemented by Joseph Weizenbaum when he created ELIZA in 1966? Yes, much before we had desktops/laptops/smartphones, ELIZA was created in a very primitive manner. But today's tools are more sophisticated. A simple application of such a solution can be seen when we interact with customer service, which can be for a Bank/Insurance/Dell/Amazon, or even GoDaddy. The initial interaction of most of these websites is a chatbot that will answer your questions; if unanswered, you can connect with Customer Service. In addition, organisations like Clinithink have used Clinical Text Mining (Physician & Nurse Notes) to identify the right set of patients for Clinical Trials faster than manually. Though ChatGPT is revolutionising how students write class essays, I am sure there will be other and better examples of conversational AI.
Predictive analytics
One of the core functions of Artificial Intelligence is trying to predict the future. In other words, answering the question, ‘What can Happen?’A common challenge faced by the IT industry at present is related to people management. Organisations are constantly trying to answer two main questions,
i. ‘Will Employee X resign and leave the company?’ or
ii. ‘Will Employee Y join our organisation or join our competitor?’
Can AI answer this question? It definitely can, but it all depends on the data (Quality and Quantity) we feed the mathematical model; else it will be ‘Garbage in, Garbage out’. A more proponent question to ask from a corporate (let's assume a hospital/patient) perspective could be as below:
i. ‘Will my claim get processed?’
ii. ‘What is the probability of my claim getting denied or put on hold?’
iii. ‘What is the length of Stay of Patient X having Pneumonia (specific condition)?
iv. Predicting the probability of 14 possible conditions of a patient from the Chest X-Ray (Search for Chexnet, a model developed by Stanford).
With the advent of the Covid-19 pandemic, pharmaceutical companies leveraged AI successfully for drug discovery. Similar procedures can be followed to discover drugs for other common diseases which can have a positive impact.
Let us talk about one more example you might relate to easily, that of Hershey's, the chocolate brand. Hershey's has a chocolate called Twizzlers, and each packet of this candy should be two ounces. The challenge here was that if we heat the ingredients (licorice) beyond a specific temperature, the quantity reduces, requiring more raw material to be added. To solve this problem, Hershey's leveraged IoT Sensors and Predictive Analytics (Reference). Twenty-two sensors placed on the cooking vat captured the temperature data every second of the cooking process, generating 60 million data points. Based on this data, Hershey's team built a model to predict the candy's weight before the candy was made. The benefit for them was that they saved $500K in raw material costs (sugar and flour) for specific quantities/conditions.
Conclusion
AI is not limited to a specific industry or domain, for that matter. For example, the movie Moneyball showcases the use of Data and Analytics in Baseball; we have seen the same being leveraged in Cricket, Tennis, and other Sports. While there may be more applications or AI in each industry, I have spoken about a limited few. We live in a ‘World of Data’ and need to learn to use the data to improve ourselves constantly; there are no two ways about it. We just need to keep ‘An Eye’ (AI) on the future.
Harish Rijhwani is the Author of the book 9 to 5 Cubicle Tales. He has twenty years of corporate experience and ten years as a visiting faculty teaching in various management institutions. Views are personal.
