A couple of decades ago if you wanted to book a train ticket you had to go through several processes: get out of the home, avail of some local transport to reach the station, collect a form, fill it up, stand in a queue, walk up to the counter and collect the ticket from a human being. In all these steps real human-beings were involved.
Fast forward to the present: we skip all these steps; instead, we simply log-in to the computer, access the Railways’ server, fill-up the online form, enter our bank or credit/debit card details and in a few seconds, like some magic act, we obtain our travel ticket. Here, we have completely removed human beings from the equation. Instead of human interaction, the task is accomplished via the conversation among multiple servers across the virtual world.
Of course, as all of us know, railway ticket booking is just one of the instances of this trend. Almost all such transactions — utility bill payments, bank transactions etc – are now done online.
As Brain Arthur puts it, this transformation of the physical economy (with real people and their tasks) is quietly creating a second economy, a digital one. Business processes that once took place among human beings are now executed electronically. They are taking place in an unseen domain that is formed by the closely linked collection of digital equipment such as servers, switches, routers and so on.
The real significance of this ‘always on’, global, digital economy is that its grip over the physical world activities is gaining strength day by day. The advances made in artificial intelligence further accelerate this process.
Thanks to the new developments in AI/machine learning techniques, many new tasks are getting into the digital economy’s fold. The Amazon Go, the cashier free shop of Amazon is a good example. The very moment you enter the store and scan the QR code, the virtual economy swings into action. When you pick up an item and take it off the shelf, a sensor immediately sends the message to a server and the system identifies the item and it is added to your virtual cart. And in case you return the item it will be removed from the cart. Once the customer leaves the store (after shopping) the cost of the items in the cart is calculated and the total amount is charged to the customer’s Amazon account. This is just the beginning of a new world order, which is going to be dominated by the virtual world.
As pointed out earlier, the rapid developments in the machine learning technology dramatically enhance the sphere of influence of the virtual world in our life. The virtual world is becoming more and more intelligent and many activities that currently require some kind of expertise/intelligence-tasks that are hitherto undertaken by experts- are also being absorbed by the virtual world.
Some people claim that they can read the mind of a person by looking at her face (‘the face is the mirror of the mind’ is a well-accepted quote). Of course, this is a very subjective issue and may not be true always. However, thanks to the advances in (machine learning enabled) facial recognition technology, now emotion detection models are available in the digital world. So, to know the real feeling of a person, simply feed the facial image of her and instantly you will get a clue as to what is going under the hood of her mind!
Another industry being disrupted by the digital economy is health care. A plethora of research projects is in progress to explore different means to identify signs of diseases with machine learning. Medical doctors use some changes in your skin, eyes, etc. to diagnose certain diseases. For example, with the help of a retina photograph, an expert doctor can identify a patient’s health parameters such as age, blood pressure and cholesterol levels. And this data will give help the doctor come to a conclusion about the patient’s health. Now efforts are on to automate this process with the aid of machine learning technology.
A machine learning model that can predict the risk of heart diseases by scanning the rear interior wall of the patient’s eye is being tried out. According to the Google Research Blog, “Using deep learning algorithms trained on data from 284,335 patients“, by studying retinal imags their team could predict cardiovascular risk factors with very high accuracy.
A machine learning algorithm that can detect diabetic retinopathy is another such example. The system automatically detects the disease symptoms by simply scanning the retinal images.
We don’t need to elaborate further on this trend. In the future, we may find automated medical shops with automatic diagnostic facilities. So, you enter a medical shop, the health care system installed on its premises will scan your facial image (or your whole body perhaps), analyse it and alert you if you have some disease. Further, the system could (automatically) suggest you some medicine and show the place where it is kept.
You go there, simply pick up the medicine and leave the shop- a doctor free/pharmacist free/cashier free health care centre. No, this is not wishful thinking, something that can happen anytime soon. Perhaps you may wonder how can one leave health diagnosis to sensors installed in medical shops. This author also shares this concern. Leaving health diagnosis entirely to an algorithm does not seem to be an acceptable trend. Of course, it can be used as an adjunct diagnostic aid.
The other side of the story
The digital economy senses all digitised tasks in the physical world and the digital footprints we make, turn into valuable data. Machine learning models that use this steady stream of data continue to learn and evolve. The more data a model gets, the more intelligent it becomes.
When you make a purchase or watch a YouTube video or write a blog post, a software powered by a machine learning algorithm lurking somewhere in the wilderness of Internet senses it, analyse the information and further enriches its understanding. But wait a minute before basking in the glory of this new automated world powered by algorithms. The power of an algorithm or a machine learning model is fueled entirely by the data fed to it.
This means one can easily trick the model by feeding it with manipulated data. The data picked up by a machine learning model may not necessarily be coming from a real person. A bot can write blog posts, twitter messages and all that; an evil mind can send doctored images of a health care model – such possibilities are endless.