Why Automation is a Necessary Evil for your Enterprises!

One of the focus areas of CXOs today is speedy adoption of new technologies (RPA, AI, Machine Learning, intelligent automation etc.). On one side, they have to Invest in these technologies, and at the same time, they have to focus upon cost-reduction. Adding to the challenge is that few software vendors can truly promise that AI will give them X% benefits within the first quarter of deployment. In many cases, they do not have either the processes or the systems in place to quickly demand  (ROI)returns from their investments in these fields. Hence, comes the element of RISK appetite for the people in the leadership.

Why is Automation Necessary?

Creating Competitive Workforce

There is a lot of mundane, repetitive work in enterprises, which is currently carried out by humans. For e.g. cleaning of disks space, eye-ball monitoring of screens for issues, installing common software in desktops/laptops etc. These activities do not require any intelligence & personally, I feel that doing this work is a very restricted use of the education, which these people have received. In addition, I find that, after interaction with some of my friends in such situation, they mostly talk about switching company because of no career progression.

Therefore, in this case, automating such processes, which are typically L0/L1, is only going to benefit every stakeholder. For e.g. now my “correctly” complaining friends can use their time in learning new skills and controlling their career progression.

With the acquisition of new skills-set, now, they can at least think of rising quickly in the hierarchy (than their earlier case). So now, we have a workforce, which is skilled, solves challenging problems and mentally satisfied (finally they are utilizing their time). Many people will not be able to align themselves with this idea and will find themselves redundant and irrelevant. These people form the basis for my third point.

Improving Efficiency of Processes

I think we all agree on this point. Intelligent software do not sleep, do not question-follow rules, do not take break, give consistent resolution, are faster etc. If we have to compare the performance of machines and humans for the same process, ceteris paribus, the efficiency of intelligent software is better. The soft benefits are of using intelligent software is immense.

Cost Optimization

In my opinion, employees should be given freedom to experiment with ideas & concepts & encouraged to suggest & design new ways to solve problems in order for the enterprises to grow in today’s VUCA world. This necessitates providing continuous learning opportunities to employees by the enterprises. The more the employees are able to up-skills themselves, the better it is for the whole enterprise.

However, the sad truth is that, many organizations are still conservative in their approach towards adopting new things. Employees are skeptical on taking risks and bringing change in their day-to-day approach to tackle problems. They still find the one way, which worked once, to be the golden rule to make that thing work repeatedly. Hence, enterprise stubbornness becomes norm & professional growth starts to stagnate.

This brings me back to my first point. In my opinion, it is better to let go of people who are not adding value to the enterprise, than to bleed by taking care of them.

Additional benefits, from a vendor or a service provider stand point, of letting go of these people is that the operating margins starts increasing. The logic for this is simple. Vendors are still getting paid for the outsourced activities they were carrying (with human employees), but now, instead of humans, they have intelligent software doing it. So, all the salary which was to be paid, gets added to the bottom line of the vendor.

Answering the Ethical question

This is where automation becomes a Necessary Evil. Given the pros, any CXOs, who fall in the innovators group, will be willing to adopt automation (thereby downsizing) to cut costs. However, the cos is that the very people who worked hard for the enterprise to grow, who stood with the enterprise during recession and during other market uncertain situations, may be shown the pink slips. Hence, it is important to give the employees chance to either up-skill or cross-skill so that they continue to add value to the enterprise.

Employees should remember that there are No Free Lunches, but, at the same time, enterprises should remember that, the employees who helped it to reach today’s heights are its Biggest Assets.

Why Human like Chat bot is still a thing of the future?

I would start this article by giving 2 examples of chatbot:

1.     Microsoft Tay:

The “Think About You” or Tay was released on Twitter on Mar 23, 2016. But only after 16 hours of its launch, it was shut down as it began to post misogynist, racist & sexually explicit messages.

2.     The Loebner Prize, is one of the most famous & tested method to test which computer programs are most human-like. It is based upon Turing test. In 2013 & 2016, it was won by Mitsuku developed by Steve Worswick. But based upon my experiment with it (dated Jun 18, 2017), though human-like, it still gives me incorrect responses to publically available information (President of India as Pratibha Patil & President of the United States as Barack Obama)

Both these instances reveal that though significant progress has been made in chatbots field, there is a long way to go for chatbots to become fully operational as humans.

Training Artificial Intelligence

Artificial Intelligence is very good at parsing input text – for e.g. Skype’s translator or Google translator. But they are still very distant to understand the semantics of the language – deciphering the meaning of the sentences. Adding to that, there are other issues like human languages are very complex: meaning is spread across levels, from alphabets, to words, to phrases, to sentences; adding to developing a data model to converse like humans. All languages, in my opinion, are abstract-words & are simple ways to explain emotions which machines can’t experience.

The data used to train the bots defines how the bots are going to perform. The Garbage-In-Garbage-Out principle is always valid when we try to curate a bot.

Challenges

In my opinion there are two barriers which are hindering the development of human like chatbot

1.     AI scientist and programmers are less focused upon developing conversational bots, but rather intelligent systems which can provide end-to-end solutions to people’s problems: digital assistant to anticipating & catering to their owner’s wants

2.     Lack of training data, or creating environment where language simulation can be done. For e.g. you may think that internet is an abundant source & can be used to teach chatbots & improve their maturity. But the challenge is that what stands for what is NOT available in a machine-understandable and digestible format.

Way Forward

Chatbots can be a very potent means to resolve problems where expertise requirement is limited & hence a great source for cost cuttings. Hence, enterprises should work progressively to integrate chatbots into their environment.

There has been a great advancement in Deep Learning (see my article on deep learning here). Using Deep Learning along with Reinforcement Learning (I am soon going to write about it) is going to be the easiest way to make human like chatbots which are effective in every environment.

Serving Customer Experience on Platter: Data Mining

In today’s world, customer’s data is getting generated from multiple sources. Companies are looking for making sense of this data for micro-marketing (for a target group), enhancing seamless customer experience, and reducing attrition of customers or increasing customer retention.

With this current explosion of data, it becomes important to understand what can be achieved by using this data to gain insights & how to dig deeper to provide a personalized experience. Data mining can help you achieve all this.

In simpler terms, data mining is an analytical process used to extract important knowledge from a large mass of data. Some ways via which data mining can help in this field is:

1.Generating a 360 degree profile of customers:

Customer data is generated from multiple sources. Enterprises are unclear as to how to integrate & synthesize data from social networks, mobile applications, in-store POS etc. Understanding this data & integrating it with a Customer Management Platform is the key to generate a 360 degree profile of the customers. Mapping these profiles & behaviors (for e.g. a vegetarian) to specific product recommendations (e.g. food ordering app only shows vegetarian restaurants in recommended places) will eventually lead to stronger brand building and high loyalty amongst customers.

2.Understanding the Sales Life-cycle of customer

The buying patterns & behaviors of different customers are different and hence when you have created a 360 degree profile of customers, it becomes important that you map them to where they fall in the sales life-cycle. For e.g. if you are an IT service provider and you have got a lead on a customer. Now, if the customer is looking for a solution next year, then your approach towards him will be different from a customer who is looking to buy your services in the next few months. Hence, knowing the current position of your customers in the sales life-cycle becomes important.

3.Decoding the path of Least Resistance

Many a times you get a new visitor to your store (whether in brick & mortar or online) & many a times a customer does not know what she is looking for. Data mining can help us understand that. So, if we can deliver it to her with minimum clicks (efforts), then we have won that customer. Hence, Data mining can help us to determine that optimum & convenient position (whether in brick & mortar or online) to place a product which increases the probability of making a sale

4.Determining cross-selling opportunities

This is the classic case of what we call association in data mining. To understand this in a real scenario, let me give an example. Father’s Day 2017 in India shall be celebrated on June 18; to gift my dad, I go to Amazon.in, and there I find that though I am looking for Aviator sunglasses, it tells me under “Frequently bought together”, that a medium case cover pouch is also available. There is a complete TechTarget article on association, you can read it here

The opportunities provided by data mining is immense. From figuring out new opportunities & new markets (remember Ansoff matrix), to increasing customer retention, building genuine loyalty and enhancing the overall customer experience, data mining can help businesses drive growth in innovative & myriad ways.

Why Enterprises Should Not Believe In Data Analytics

Big data algorithms, machine learning & reasoning has become the heart of almost all applications today. These smart applications are solving crucial business problems and helping decision makers in quickly reaching a business critical decision in a matter of minutes. These techniques are defining the norms by also using statistical analysis & predictive modelling.

But, all that glitters is not gold and we have to understand that it’s NOT always the case that all insights that spawns out of such models is CORRECT. Business leaders have to understand the inflexion point where data starts to control them rather than other way round. If they are thinking that insights coming out from machines will be always Right & Correct, then it’s a Mistake!

In this post, I shall be explaining the various issues which comes with using the data analytics as it is

Simpson’s Paradox

The best way to understand this statistical paradox is – the groups have averages that point in one direction whereas the overall averages points in other direction.

Let’s understand this with 2 real world example:

Take #1:

In tennis, if the loser of the match has actually won more games than the winner, then we have an example of “Simpson’s paradox”. For example, though not very possible, if the final score is 0-6, 7-5, 7-5; then the loser has won more points in the game (16) than the winner (14).

A real game example is Isner–Mahut match at the 2010 Wimbledon Championships. If you see the Records section, the last but one point explains it. Mahut won 502 points in the match as compared to Isner’s 478 (difference of 24). But we all know that Isner won the match 6-3, 3-6, 6-7, 7-6, 70-68.

Take #2:

Suppose your enterprise has two business application towers: A & B. Now let us analyze the overall tickets generated from those applications:

If you look at the analytics reports & dashboards created at the Business Applications level, you shall see that the predicted tickets matches with the Actual tickets. So,

Inferences:

1.    the maturity of the model is very high

2.    resulting to say that we can scale it up to new towers.

However, if you deep-dive into the two towers, you can see that this inference is Incorrect. This is one of the most important challenges in the reporting & dash-boarding world. It is easy to think that we are meeting our numbers, when in reality; the case might be completely different.

Idiosyncrasies in the data

In many businesses, important decisions are made based upon the statistical inferences using the historical references and experiences. A major caveat here is that if the sample size in use is small, then few outliers can skew the understandings/inferences a lot.

Many predictive models use historical data to make predictions on the future. Hence, if the past data and its data model upon which it is based relies heavily on past incidents, and then it may not accurately give predictions on the future.

Believing numbers blindly

Too often, we are so driven by numbers that we forget that there are biases, which creep into the system, possibly during the initial requirement validation phases, designing the data model phase etc. Such biases, though very small, constraints the way with which we look upon the end-results (in the form of dashboards & reports). In addition, it is also important to continuously normalize & check the data for inconsistencies and do a ground check verification before any major decisions can be taken. To give an example, it may be the case that business/operations leaders may be seeing a high inflow of tickets, though at a ground level, those tickets always existed in the system; only thing is that they were NEVER being tracked!

Though these challenges exists, the Solution to these challenges are domain expertise, tacit business knowledge, common sense & above all, Critical Thinking, which can help business manager, escape such caveats.

Intelligent Automation in As-A-Service Economy

In a new as-a-service economy, characterized by continuous disruptions and very low entry-barriers, enterprises are becoming more prone to being part of the crowd. To counter such vulnerabilities, integrating Artificial Intelligence into the IT landscape has been one of the highlights of today’s enterprises. Many enterprises across verticals are willing to accept intelligent automation in their day-to-day operations for cost efficiencies, routine task and process automation and portfolio rationalization. But here it becomes very important to set realistic goals from intelligent automation implementation and cognitive technologies. Hence it becomes imperative to understand the layers of intelligent automation which can be addressed and delivered in the IT landscape of any enterprise.

I like to classify automation into three tiers:

Traditional rule based automation which has its scope in automating particular tasks. Some e.g. includes scripts to migrate applications, services scripts, upgrading enterprise applications etc. Tasks & Stack automation too falls under this tier.

Knowledge based Process Automation: Process can be a combination of multiple chains of tasks bundled together. Many times, SOP (Standard Operating procedure) and many L1/L1.5 tickets automation falls under this tier. With Knowledge based Process automation, such candidates are either eliminated or automated to achieve business outcomes.

Cognitive Automation This tier comprises of AI algorithms like natural language processing, semantic data processing, knowledge management, reasons, and expert systems. The vision here should be in integrating many DIY tasks so that instead of creating tickets for those issues, the end-user can themselves resolve the issue using cognitive capabilities. Another use case can be the use of Machine Learning models to monitor the current IT infrastructure and come up with recommendations (actionable intelligence) to proactively reduce infra related incidents.

Performance and Governance Automation, for e.g. business process monitoring and managed services analytics are required at each of the above automation tier. These are also implemented to monitor the bots deployed. [Read the definition of bots here]

The goal of any automation solution, at any point of time, should be to provide end-to-end automation. This requires the right practice and solutions to be proposed and deployed in the customer’s environment. A consulting approach by the service providers can be the key here.

Defining the Business Value

Intelligent automation should be platform agnostic, readily plug-&-play and easy in deployment. The categorization of intelligent automation into the three tiers allows to provide values to customers in terms of reduced cost of operations, marriage of technological levers with business objectives and mitigation of risks. The above separation of intelligent automation into three tiers also helps in easy customization of each component to suite the business needs. The commitments to leverage intelligent automation to help customers move to an IT landscape which is characterized by loosely-coupled, best of breed components and improved flexibility has led to tremendous business improvement in their IT lifecycle.

Measuring Outcomes

Because of intelligent automation being implemented across various divisions of enterprises, different types of performance metrics have entered the business domain. FTE reduction, Decreasing Mean Time to Resolve (MTTR), reducing number of hops, accurate assignment index of tickets, up-time of infrastructure, availability of apps etc. have gained wide acceptance.

Commuting one lever higher

The role of intelligent automation does not stop at just achieving the mentioned automation metrics, and hence, there has to be a feedback and learning mechanism which can, in the future help the end-user to either self-heal the issue or predict and inform the agents/end-user about the possible outage. The predictive analytics engine of the intelligent automation system should be capable of ensuring this. Intelligent automation has to not only predict but also resolve those new typical L1/L1.5 issues by itself in future from then onwards (a concept of self-learning system).

The Moral and Ethical Issues

FTE reduction has been gaining traction in the minds of the agents for some time now. They feel that intelligent automation will take away their job. I, believe this to be untrue. I believe that a human resource should not be wasted in trivial, mundane and repetitive job, rather, the human mind should be leveraged to produce amazing innovative and creative results. In my opinion, by freeing up these agents & re-training them, their skill set can be improved and they can move up the value-chain, thereby achieving more on their professional front.

Making Intelligent Automation work

In today’s world, data is flowing like never before. They originate from myriad of sources, intelligent devices, millions of connected devices to name a few. Hence, it becomes imperative that these data are properly processed and the information derived is progressively and incrementally analyzed.

Intelligent automation is the integration of automation and artificial intelligence. It has become top-of-the mind issues for CXOs around the world. Intelligent automation, going forward, is going to define who becomes the winner and who loses out in the long run. It has already helped enterprises to go beyond the conventional methods of doing things and achieve unprecedented levels of quality and efficiency. The ability to command savings on incremental dollars spend on ITSM issues like service desk operations w.r.t to the current systems has made business leaders across verticals to seriously ponder on integrating intelligent automation into their IT estate. The disruptions in this field is revolutionary, and happens at each level of tasks or process performed by the enterprises.

How can Intelligent Automation Help?

The class of business & IT problems which can be solved using intelligent automation is growing rapidly. With technologies like Natural Language Processing, Pattern recognition, Voice recognition and Machine Learning, the value to the enterprises can be delivered is immense. I have shared my views in my previous post.

What Makes Intelligent Automation Work? What are the Challenges?

Data. The only input to an intelligent automation system is data. Intelligent automation platforms consume data, run algorithms and then come up with some data models, meaningful information and predictions which can be then used to achieve efficiency improvements. The nature of intelligent automation is such that it starts to learn with every new use case which comes it way, thereby, reducing human intervention to a minimum.

There are majorly 3 challenges in making the intelligent automation work

Data is getting generated from multiple sources: images, videos, flat files etc. Then there is data segmentation w.r.t geography, category, time, transactions, performance etc. Also a lot of value can be generated by studying the sentiments of the agents and the end-users of the systems (very potential source of data). Hence, with so disparity in the source of data, it becomes difficult to sanitize the data so that it can be normalized and fed to the intelligent automation system to process, learn and then perform. So business decision as to sourcing of data and what data should be used to initially teach the intelligent system becomes critical.

The second challenge is building and training process of the intelligent platforms themselves. It is a highly math-oriented, algorithm intensive process which requires lots of capabilities in the form of skilled developers and engineers so that they get the analytical model right. Here, it becomes important that these people know the latest mathematical modeling, techniques and statistical methodology so that they build and train the intelligent automation platform which can deliver maximum benefit to the business enterprises.

The last challenge is to get the acceptance of intelligent automation in the IT estate of the enterprises. It becomes meaningless if the predictions of the intelligent automation platforms is neglected at the ground level. It makes little sense if the predictions of intelligent automation, does not result in some action by the stakeholders, does not invoke the innovative thinking to proactively resolve business or IT issues on a day-to-day basis.

Where to Use Artificial Intelligence in Your Enterprise?

By 2020, customers will manage 85% of their relationship with the enterprise without interacting with a human. -   Gartner Predicts

Artificial intelligence is a concept that is causing many ripples in the technology space. Growth in hardware technologies, analytical models and engines, & finally data are the chief reasons creating this hype. In recent times, we have seen many ground breaking news on AI, ranging from self-driving cars to technology major acquiring AI startup, to defense, to healthcare.

Among all these hype, the basic question for the business world is: How can AI help them bring their cost down and performance efficiencies up!

In my view, AI is still in its nascent stage for adoption by enterprises. Though enterprises have a large pool of data, they are skeptic in terms of how benefits can be availed given their scope and size.

I shall try to trace the use-cases corresponding to stages of application & then to maintenance of infrastructure.

Some examples are:

1. Fast-Coders in the Making AI via Machine Learning has shown its capability to understand human language (e.g. Siri). Siri, not only responds to your queries, but also understands the intent behind your query. Envision a scenario that you are using any SDK to write your code. Now the moment you put // or /* … */ (documentation comments) and write the intent/functionality/use-case of that code in plain English, the bot pulls out the relevant code from code repository (SVN/Team Foundation Server) and helps you complete the code. Alternatively, it can refer to those codes and help you to finish a piece of logic faster!

Therefore, in this case, we have a coding-helper bot, trained on the code repository of the enterprise (for more maturity, code available on GitHub can be used), and can suggest code modules/functions which can be used by the developers for faster coding.

2. Automated Testing Automated testing has become an integrated solution as part of many managed services offerings and is a highly competitive field. Almost all major service providers have presence. For more information, you can refer to any of the analyst reports, Everest etc.

3. Bots for Maintenance AI bots may soon replace physical human beings in doing mundane maintenance tasks like swapping server racks. There may be bots, which are monitoring each of your IT estate and predict network and storage failures, storage limits threshold crossing, temperature regulations etc. Maintenance activities are bread & butter of many IT companies and currently many such companies are working to utilize their expertise to build bots for predictive and preventive maintenance.

4. IoT is here to Stay The concept of tools, devices, objects (electronics used in daily life), and infrastructure being connected to each other and working in tandem to create an ecosystem of smarter & responsive devices brings with it unprecedented convolution. The challenge here is going to be on how to make sense out of all the unstructured data, which can help in deriving actionable intelligence. This is where enterprises will have to use the AI algorithms for classification etc. for actionable acumen.

5. Robust Cyber-security We all know about the two attacks (largest as well) on the security breaches in Yahoo network. Similar case came for Apple as well.

AI can be used to divulge in-progress attacks as it can learn the patterns across devices and network, and report any anomalies! Hence, mitigation action can be taken while the breach/intrusion is still in-progress!

Artificial Intelligence, though far from being accepted as an end-to-end solution, has been adopted in the form of various point solutions at application and infrastructure level. The time is not far when enterprises start taking definitive steps to integrate it within their overall strategic framework to achieve business goals.

Analytics Driven Managed Services

In today’s world, the amount and volume of data is rising abruptly and rapidly. We see tremendous and continuous disruptions in data analytics, knowledge management, business intelligence and intelligent automation. As such, new trends like Big Data and Analytics have become very pertinent with the CXOs IT landscape modernization and rationalization agenda. As the insights-driven road-maps and strategies take shape, these will become very important source of competitive differentiation in the market. The current challenge for many enterprises is how to utilize and capitalize big data analytics and derive maximum benefits considering the technological disruptions and very light budget.

Breaking down Analytics Driven Decision Making

Analytics refers to the discovery and communication of relevant insights from data. We understand that data-driven decision making emphasize upon quantitative aspects of data: number crunching and proper data processing to come up with results based upon numbers as the underlying facts.

Now, analytics driven decisions takes data-driven decision making to the next step, into the domain of qualitative analysis. This allows for the integration of quantitative and qualitative data and hence another layer of insights gets added which allows for one more layer of consideration and results in showing various data-points which influence the decision making process.

Consider an example, using analytics driven decision making, CXOs can not only focus on which IT lever is causing the main problem, but they can also get insights on what can be done to prevent and predict it before occurring, so that proactive measures can be taken and the business sees no down-time. By concentrating on analytics driven decision-making, enterprises can focus upon important questions of What and Why. It implies that decision-makers can see an overall view of what is happening and why is it happening along with how it can be prevented. It can have an immediate business impact in terms of accurate measurement of key metrics and costs efficiencies.

How to become Analytics Driven?

The key to launching analytics at a corporate level starts with utilizing the right tools and proper training on these tools. For enterprises, which want to leverage the power of analytics into their business domains, experts and tools, customized to their needs is the starting point. For many enterprises, it is logical to assume that this is not their core competency and hence they need external consulting to find ways to integrate business intelligence, revamp performance management, risk mitigation, compliance and governance mechanisms.

Therefore, the important decisions with the CXOs is that before pondering over capitalizing on Analytics, they must have a strategic road-map regarding governance models at information, technological and project levels. This shall ensure alignment of analytics with the core business objectives. This is also necessary to integrate new technologies into existing IT landscape resulting in maximum value derivation to the enterprises.

Beginner’s Guide to Artificial Intelligence, Machine Learning, Neural Network and Deep Learning (Part 2/2)

This article is in continuation to my previous article. You can read that post here.

Artificial Neural networks (ANN) and Neural Networks (NN), is another approach to teach computers to think, decide and decipher the environment like humans. This approach is synonymous to our understanding of human brain (biology): interconnections among neurons. NN are typically visualized as systematic interconnection of neurons, which exchange data or messages among each other. These connections have weights (numbers) that is updated based upon experience, thereby making NN adaptive to inputs and capable of understanding and learning.

Hence, this approach works on probability: based upon input, it gives recommendations or predictions with a certain confidence level. A feedback mechanism enables learning. Hence, by feedback, it understands if its recommendations/predictions are correct or incorrect, and consequently, updates the approach it undertakes for the future event. For example, it can say with 80% confidence that an image is a cat’s image, 10% confidence that it is a leopard’s image, 6% confidence that it is a cheetah and so on – and then the feedback mechanism of the network architecture tells NN if it is correct or incorrect.

Because of the high computation intensity required to run even the most basic neural networks, it was not commercially feasible and not practical. The advent of GPUs in this field is promising and we hope to see some results in near future. The advantage of pursuing NN is that it retains the advantages of machines over humans like speed, lack of bias and accuracy while trying to mimic human brain.

Once we have a basic understanding of NN, let us now shift our focus to Deep Learning. Deep Learning refers to NN that are many layers deep. Deep Learning is deep because of the structure and architecture are ANNs. When NN was conceptualized, they were just two layers deep and I just mentioned earlier, it was computationally not feasible to build large networks. With GPUs, it is possible to build NN with 10+ layers.

Therefore, in deep learning, layers of neurons are stacked on top of each other. The job of lowest layer is to take inputs in the form of text, images, sound etc. Each neuron, then, stores some info about the data elements they encounter. Now, at the above layer, a more abstract version of the data is transmitted. Hence, the higher the layer, the more abstract you learn.

The best use case of ANN is extraction of features from images without any human intervention. Feed ANN an image and it will compute features like colors distribution to something like if a cat is running or sitting. The only requirement of such computation is training of ANNs, which require massive data.

With Big Data sources like Twitter, Facebook, etc. we have data corpus not available 2 or 3 decades back. Still, the challenge lies in cleaning and processing of data into right format, which can be fed to the machine learning algorithms.

I sincerely want to thank Michael Copeland and Bernard Marr (@bernardmarr) for shaping my thoughts on AI over the months.

{The examples used in the above blog are a bit far-fetched and ahead of the current time. They are provided to draw parallelism from the real word and easy understanding.}

Beginner’s Guide to Artificial Intelligence, Machine Learning, Neural Network and Deep Learning (Part 1/2)

AI, ML, Neural Networks and Deep Learning are some of the buzzwords of today’s world. They are disrupting the way in which traditional business operate. Many service-based organizations are branding themselves as pioneers and leaders in these frontiers. However, before putting money into any of these AI branded assets, it becomes very important to understand the business use-case of these technologies (on which I shall write later). Currently, I shall focus upon facilitating the beginner’s understanding of these buzzwords.

Artificial Intelligence, in the easiest language, is used when machines can take decisions and perform actions (easy or complex) intelligently and smartly, implying, it can mimic human activities, like learning and solving problems.

AI may be classified into two categories: Applied AI & General AI. Applied AI is what is creating the buzzword in today’s world: autonomous cars (Volvo S60 Drive Me), virtual agents (Louise, the virtual agent of eBay), playing strategic games (Go & Chess) against humans etc. Hence, they are specific to a case in point. General AI is what we have seen in movies like Ultron (from Avengers series) & Ava (from Ex Machina) i.e. they have the capability to mimic the human and can perform actions like those that humans do.

An interesting observation is that actions taken by machines, which were once categorized as intelligent, are no longer considered intelligent. For e.g. Optical Character Recognition. Hence, just like human’s approach (metrics to measure) to intelligence (psychology) varies, the metrics as to what actions define artificial intelligence and what not, may require continuous change.

Now, let us try to understand Machine Learning. ML, in the simplest form, is the ability of the machines to parse data, categorize it, learn that categorization and then perform some actions or give some predictions on cases for which it was not trained. So, rather than the traditional IF… ELSE statements, using algorithms like clustering, decision tree, inductive logic & Bayesian networks, the machine is trained using large volumes of data after which it can perform some task for which it was trained for.

I shall take a very novice example to explain it. Suppose the machine is trained on all the past matches of Roger Federer. His opponents, tournaments, practice sessions, performances at all levels, etc. Now based upon this training, if the machine is able to identify Federer’s odds of winning against any opponent.

Two more “learning” keywords used frequently are Supervised learning & Unsupervised learning.

Supervised learning happens when a bot is trained on corpus of data and the output is defined. If the outputs are defined as classes, then it is a classification problem. If the output is continuous, then it is a regression problem. There are many use-cases defined for classification. For e.g.:

1. To classify, if the financial transaction is fraudulent or not

2. To classify the different types of objects in an image (fruits, vegetables)

3. To classify the given texts into different categories (if the tweet is about football, cricket etc.) in Natural Language Processing (NLP).

Unsupervised learning takes place when the bot starts to learn and take decisions from itself (a concept called self-learn).

Let us understand these two concepts from a real world example.

Case 1: Supervised Learning

Vipul is a kid. He sees different kinds of fruits. His father tells him that this particular fruit is an apple, orange etc. Now a new fruit comes in front of Vipul, which he has not seen before. Vipul identifies it as an apple – and not as a mango, papaya etc.

Here, I had a teacher to guide me and help me learn new concepts, so that when a new object came my way to which I had not been trained, I was still able to categorize and identify it.

Case 2: Unsupervised Learning

Vipul is a kid. He went to North Korea, a country about which he had no prior knowledge – no information on their culture, food, tradition, language etc. However, Vipul tries to learn and make sense of his surrounding – what to eat, how to greet people, how to pray etc.

This is unsupervised learning because in this case, though, I had many data around me, I did not know how to derive meaning out of it or rather what to do with it. Here I had no teacher to guide me and I had to figure out a way on my own. Then, after some time, based upon certain learning, I started processing these data into information categories that made sense.

(The rest of the knowledge will be shared in the second part. )