The theatrical release of James Camerons sci-fi film Terminator 2, featuring Arnold Schwarzenegger as a cyborg with a computer brain, had a crucial scene deleted. The scene, part of the extended release of the movie, shows young John Connor and his mother opening up the head of the cyborg to switch its computer brain from read only to learning mode. The cyborg (Schwarzenegger) then picks up human values and mannerisms as the movie progresses.

For movie buffs, the deleted scene is worth seeing for special effects and also to catch a glimpse of Linda Hamilton (playing Johns mother Sarah Connor) with her twin sister Leslie playing her image in a mirror. In the theatrical release, where the scene is omitted, the cyborg just tells John that its brain is a neural-net processor, a learning computer, without mentioning any on/off options. That was back in 1991. Today, in 2017, a learning computer is much more of a reality.

While artificial intelligence (AI) and machine learning (ML) concepts have been around since the 1940s and 1950s (See ABC of AI, ML and Deep Learning), the availability of huge amounts of data is making the difference now. A learning computer does not need to travel back in time like in the movie and many are solving real problems in India. For example, in healthcare, ML is helping oncologists sift through huge amounts of cancer cases and suggesting preferred treatment; in education it is predicting who might drop out of school; and in fashion it is forecasting colours that can dominate the next season. Retail, transportation and financial services have adopted ML in different forms. The learning switch is turned on in India. Every large organisation was sitting on data. The cloud is bringing computing power to it and ML is creating actionable intelligence, says Anil Bhansali, MD, Microsoft India (R&D) Pvt Ltd.

Machine vs machine A war of machines scenario seems appropriate to discuss it. Consider this example. In October 2016, K Sandeep Nayak booked three flight tickets for his wife and children to fly to Mangaluru from Mumbai during the Christmas holidays two months later, hoping to get a low fare. He spent Rs 7,500 per ticket. Later, when he decided to join his family for the trip, just a day before the journey on December 25, he could book himself into the same flight at Rs 4,000 only. I wish I could find out if airfare could fall, says Nayak, an executive director with Centrum Broking.

Actually, there is a way.

Today, most airlines follow a sinusoidal graph (S curve) for pricing tickets, often dictated by an algorithm to maximise revenues pushing up prices following buying behaviour.gregator app Ixigo. It can predict whether the price of an air ticket on a particular date is likely to fall. When a customer enters the date of journey, the app predicts, with more than 80 per cent accuracy, how much the airfare may drop for the sector on that date and how the prices could vary over that period. (Ixigo also has a railway app that predicts if a rail ticket on a wait list may get confirmed.)

We have a huge data set created by 4 million active users, 50 million sessions per month, Aloke Bajpai, Ixigo

Ixigos global peer Kayak is one of the pioneers in fare prediction. If airfare prediction seems like a machine-vs-machine scenario, there are more such examples: programmatic advertising algorithms that compete for advertising spots, or algorithmic trading applications that compete to get the best trades in the securities market.

Here is something a little more interesting.

Arya.ai is a Mumbaibased startup, founded by Vinay Kumar and Deekshith Marla, both IIT-Bombay grads. In 2016, Arya.ai was selected by French innovation agency Paris&Co, from 21 global companies, for an international innovation award. Kumar still looks like a college student and moves around Mumbai on his motorbike. One of the current projects that Arya.ai is working on involves creating an ML application for selling securities without letting prices crash. The client, with a mandate to sell a large block of stock or bonds in the market, wants Arya.ai to create an algorithm for selling so that it does not lead to prices of the security dropping.

At the same time, there are ML algorithms as well as human intelligence trying to buy the security at the lowest price possible, says Kumar. Algorithmic trading has been around for a while and brokers with proprietary trading arms often use it to gain a few seconds advantage. Now research is focused on whether an ML layer can be built on top of the algo. Can the machines be allowed to alter the trading algorithm on their own and what will this mean for the securities markets?

Last month, JP Morgan released a report in New York, Big Data and AI Strategies, with the subhead, Machine Learning and Alternative Data Approach to Investing

Written by Marko Kolanovic and Rajesh T Krishnamachari, the report suggests that analysts and market operators need to master ML techniques as usual indicators like company quarterly reports and GDP growth data will soon be predicted early by ML programs. It says that just as machines with ML are able to replace humans for short-term trading decisions, they can also do better than humans in the medium term. Machines have the ability to quickly analyse news feeds and tweets, process earnings statements, scrape websites and trade on these instantaneously. Back in India, here is another scenario. Vertoz is a Mumbai-based programmatic advertising company that works with clients (advertisers) and online media in placing digital advertising, targeting the advertisements and bidding for the best spots.

We need to find which inventory is good for us, says founder Ashish Shah, referring to spots on popular media websites. If we had to do it manually it would be like finding needles in a haystack. Vertozs programs compete with the likes of Google, bidding for top slots in global digital media.

Man Fridays While the buzz on big data analytics came first, the focus on ML has been facilitated by larger players like Google, Intel, Microsoft and Amazon making off-the-shelf modules available in India. But, then, some platforms have been around for decades. Says Shah: Most of our work is based on Java and Python that are 1980s technologies. We have built our layers on top of that.

Ixigos chief technology officer Rajnish Kumar mentions Googles TensorFlow and Amazons AWS Machine Learning as examples of off-the-shelf modules. Microsoft offers its Azure platform for others to create their own ML offerings. A Google spokesperson told ET Magazine that in future it expects to offer non-experts the ability to create and deploy ML modules: At Google, we have applied deep learning models to many applications from image recognition to speech recognition to machine translation. In our approach a controller neural net can propose a child model architecture, which can then be trained and evaluated for quality on a particular task. This is machine to machine learning.

Going forward, we will work on careful analysis and testing of these machine-generated architectures to refine our understanding. If we succeed, we think this can inspire new types of neural nets and make it possible for non-experts to create neural nets tailored to their particular needs, allowing machine learning to have a greater impact on everyone, adds the Google spokesperson. Google offers some simple applications of ML. CESC Ltd, Kolkata-based flagship of the RP-Sanjiv Goenka Group, is using a Google API (application programming interface) which records the reading of the electrical metre when the numbers are read out loud. Instead of keying the reading in or taking a photo of it, the staff can speak into their phone app chaar-shunyo-teen-paanch (4, 0, 3, 5), says Debashis Roy, vice-president (information technology ), CESC Ltd.

Roy says that when the project started, the app showed only 40% accuracy, but it is learning to recognise more and more Bengali dialects as well as Hindi and English. No matter what the dialect of the staff, the reading can be recorded. We will launch it fully when we get to 95% accuracy, says Roy.

Another Google partner is Pune-based Searce, a 12-year-old operation led by founder Hardik Parekh, who finds it convenient to work with Googles APIs as he feels the company almost embodies the open source or democratic spirit. Parekhs ML offering HappierHR tries to automate much of the routine HR operations right from initial interviews of job applicants and induction of new employees to creation of their email ids and leave approvals.

Supervisors also get suggestions to give leave to subordinates on, say, their wedding anniversaries, if there arent any important meetings scheduled for that day,says Parekh. While Google, Amazon and Microsoft offer platforms for others to use, IBM has its own ML suite called Watson, a complete offering at the premium end of the market for end-users. One of the earliest projects IBM took up in India was with Manipal Hospitals in oncology. Manipal was an early adopter: it was globally the second or the third hospital to adopt it, says Prashant Pradhan, chief developer advocate for IBM in India and South Asia.

This is how it works. For a medical board on breast cancer, the Watson program is made a member along with other doctors. Given a specific case, Watson gives its opinion and preferred treatment after going through millions of cases that are loaded on to it. Entire cancer research can run to 50 million pages, and 40,000 papers are added every year.

It is impossible for a doctor to go through all of that. The ratio of cases to oncologists is 16,000:1, adds Pradhan, stressing why ML is a great application to use in cancer treatment. Microsoft, too, has used its ML offerings in Indias healthcare. In Hyderabad, it has helped LV Prasad Eye Institute treat avoidable blindness. A second project it has worked on is helping children who wear glasses.

The work started in India has gone global, and LV Prasad Eye Institute is now part of the Microsoft Intelligent Network for Eyecare, which includes five other eyecare facilities from across the world. Microsoft has also studied 50,000 students in Class X in Chittoor in Andhra Pradesh to predict which ones may drop out. It allows the schools to send them for counselling.

Machine radiologists and bankers There is enough indication that ML bots or apps can often deliver better results than humans. Last month, IT services giant Wipro said it got productivity of 12,000 people out of 1,800 bots (software programs that perform automated tasks). Automated bots are not quite ML, but are an indicator of what may come. Rizwan Koita, serial entrepreneur and founder CEO of Citius Tech, a healthcare-focused tech company, recalls a conversation with his niece two months ago. She had qualified to pursue a course in radiology or anaesthesiology and was seeking my advice. I had to tell her that in a few years a radiologist may not have a job, says Koita. He argues that a radiologists job is to interpret images. Therefore millions of existing images (X-rays, sonograms, scans) and their interpretations can be fed into an ML algorithm; it may be a matter of time before a machine gives better interpretations than a human radiologist.

From healthcare to fashion. Mumbai-based designer couple Shane and Falguni Peacock have been using IBMs Watson for a couple of months now. The system helps the duo go through designs and silhouettes that have been shown at fashion events across the world over the last decade. They are using Watson for a project that uses international designs in Bollywood. Watson predicts colours that may be in vogue six months from now and warns if certain silhouettes have been overused in the last couple of years.

Watson is able to tell us what colour may be in six months from now, says Shane Peacock

Says Shane: Suppose we want to work on a Mughal theme, we can feed images of Mughal-era paintings, architectures and colours into the system, which is able to turn out its unique prints. It also reproduces Mughal prints created by other human designers, just for comparison. The designer couple have one more exciting project for which they are using Watson. A dress that changes hues according to the time of the day or the mood of the person wearing it. We can use two colours, say black and white. The dress can become fully white or fully black or a combination of black and white. An app on the wearers phone can control it. The change can happen on the go, while the dress is worn. You can get into a car in white and come out in black. In financial services, Kumar of Arya.ai points out that the loan approval process is an area where he sees a lot of human effort being bested by machines. In fact, Arya has implemented a program where an ML app sifts through loan applications.

ICICI Lombard and Birla Sun Life Insurance too have created bots as the first interface with customers Not to be left behind, the Indian IT biggies, TCS, Infosys and Wipro, have their own ML and AI offerings (See Machine Learning in India). Google announced in March that it will mentor half a dozen AI startups. A report by Tracxn, a venture capital research platform, noted that there are at least 300 startups in India using ML and AI technologies. An opportunity also presents a threat. Before ML can replace humans in core functions, it will need humans to create applications. Says Bhansali of Microsoft: These are still early days: technologies are on trial and talent is scarce. Ixigo CEO Aloke Bajpai echoes him when he says there are no trained engineers in AI and ML in India, and his team is entirely trained in-house.

There is definitely a shortage of talent for AI technologies. Only 4 per cent of AI professionals in India have worked on core AI technologies such as deep learning and neural networks, says Akhilesh Tuteja, partner at KPMG. Bridging the gap will be key in turning a potential weakness into a strength.

See the original post here:

artificial intelligence: War of the machines: The opportunities in ... - Economic Times