In 1967, they introduced the world to floppy disks. In 1969, they helped land the first two humans on the moon. In 1970, they came out with the hard disk. There were a host of other inventions to IBM Research Labs’ credit — the ATM, the portable computer, the magnetic strip card, the technology behind laser eye surgery and the smartphone. The world’s first corporate laboratory, set up by Thomas J Watson Sr in 1945, has an undeniably illustrious history.
With more than 3,000 researchers and 12 centres spread across six continents, it remains the world’s largest industrial-research organisation.
In 1998, it set up its India wing — IBM Research India — in the IIT-Delhi campus. It was the country’s first industrial science lab; the second centre opened in Bengaluru in 2005. The operations here have grown exponentially — from something as basic as troubleshooting services, the lab is now driving a lot of research work globally. In 2018, IBM researchers from India secured over 800 patents across technologies, including cloud computing, artificial intelligence (AI), quantum computing and automation.
“IBM has two clear strategies — one is the hybrid cloud strategy where we enable our clients to move to the cloud. The second is cognitive enterprise, where research redefines how business is done today,” says Gargi B Dasgupta, director, IBM Research India and CTO, IBM India and South Asia. IBM Research India is focused on applying AI and blockchain technology across agriculture, retail and enterprise automation.
IBM is collaborating with IIT-Delhi and IIT-Bombay to make AI systems that will do more than recognise the face of a President or the image of a flower. The new AI systems will be able to reason and learn, much like a person. AI solutions will be trained to comprehend complex questions using natural language techniques and derive new insights using domain knowledge to help companies with decision-making. “These solutions can service a wider set of use cases across industries, and with lesser data,” says Dasgupta (See: Betting big on AI). AI-derived businesses are already growing at a brisk pace globally, and in India. In the long run, Dasgupta says, they are working to create AI that is bias-free and transparent.
Farmer gains
One sector which urgently needs help with decision making is agriculture. Take for example the Fall Armyworm attack, which has been spreading ruin at an alarming speed through the southern, western and eastern states of India since last June. An early warning would definitely have helped, and it is here that IBM could provide an answer. “IBM Research India is combining multiple global satellite-based information sources to compute actionable insights such as crop health stress alert, water stress, pest or disease risk forecast, to help individual farmers improve productivity at a sub-acre level,” says Dasgupta.
Combining this with data from The Weather Company (an IBM subsidiary) and Watson, IBM Research launched the Watson Decision Platform for Agriculture. It gives predictive insights on how to manage scarce groundwater-based irrigation, optimise the timing of harvest and the amount of fertilisers and pesticides to be used. The insights are delivered to the phones of farmers and agronomists.
IBM Research India is also working with agri-companies and government institutions to improve farmer productivity. For instance, Pune-based agricultural start-up AgroStar tied up with The Weather Company to provide critical insights on crop disease risks to over a million farmers in Maharashtra, Gujarat, and Rajasthan. It has also tied up with the Niti Aayog to develop a crop-yield prediction model to provide real-time advice to farmers in 10 backward districts of Assam, Bihar, Jharkhand, Madhya Pradesh, Maharashtra, Rajasthan and Uttar Pradesh in the first phase. This is to introduce farmers to climate-aware farming techniques, crop-monitoring and early warning on pest outbreaks based on AI innovation. The plan is to implement this in over 100 districts.
Tomato is a hot potato. Its prices fluctuate frequently and farmers are either dumping them on the roadside or selling them at a huge profit, particularly in Karnataka. IBM’s team in India has developed an advanced price-forecasting system, using the Watson Decision Platform in collaboration with the Karnataka Agricultural Price Commission, to help vegetable growers in the state. “Sometimes the prices are so low that it makes little sense to transport the harvest to the mandis,” says Shantanu Godbole, senior manager, industries research, IBM Research India. Using satellite imagery and weather data, IBM’s system will assess the acreage and crop health to predict the production pattern and market price trends on a fortnightly basis. While estimating prices, it also considers the rates in the major markets of neighbouring states. The price-forecasting mechanism is the first-of-its-kind in the country. It is initially being launched for the major tomato-growing districts of Kolar, Chikkaballapura and Belagavi and two key maize-producing districts, Davanagere and Haveri.
Credit, or the lack of it, is another phantom that haunts farmers. The research lab is working on a system that will generate credit scores and help financial institutions assess a farmer’s credit-worthiness. It uses basic farm details to calculate how much yield and profit a farmer would have made in the past few crop seasons.
IBM Research India is also driving global research. Every year, IBM publishes a predictive list of five innovations that will change the world in five years and this year the focus is on farm-to-fork ideas. What is noteworthy is that the origin of three of those five innovations is the research labs in India.
The researchers at IBM are looking to ‘clean up’ the food delivery process. They are working on solutions that detect bad bacteria in food, using the next generation sequencing system, and spot counterfeit products using smartphones and AI systems. They are also tackling plastic use in food packaging with VolCat. This process recycles the packaging by placing it in a pressure reactor to produce a powder that can be used to create high-grade plastic instead of the low-grade variety that comes from the cleaning-melting-remoulding process. IBM Research India is also pioneering a lot of work behind creating digital twins or ‘virtual model’ of farms. This will help participants across the supply chain better understand how the food moves, and where the wastage occurs.
Style check
The green and the grime of open fields may seem a universe away from the glamour of fashion, but clothes retailers grapple with high unpredictability too.
IBM’s Godbole says, “Our retail journey started about five years ago. We wanted to see if we can develop an AI-based system that can understand text and pictures and develop a deep understanding of the fashion domain,” says Godbole. They also engaged with clients to gain an understanding of real-world problems. “Colour in clothes is a complex topic and to develop a set of machine-learning and AI techniques which understand colour is non-trivial.”
IBM’s target clients in the retail business are not e-commerce companies but retailers with a network of physical shops. “E-commerce companies can track customer preferences through cookies and are sitting on large amounts of data. They know their customer history and what they like so they can make personalised offers. For physical stores, if the regular salesman at the stores quits his job, all the master data about the customer is lost,” says Godbole. IBM found that the bigger problems their clients faced were in the back-end and how to manufacture the right quantity so as to avoid dead inventory. “Our clients wanted sustainable supply chains. It was a classic problem around demand forecasting, stock allocation and inventory management,” says Dasgupta.
But retailers are sitting on a treasure trove of data – that which is generated within their enterprise, with their vendors and on social media. Unfortunately, a lot of it is unstructured, for example freeform text and images, and offers little value. IBM Research developed an AI solution that can not only help retailers decide on what kind of apparel to procure but also help designers create the perfect product. The AI utilises concepts of visual search (where people click photos offline and search for it online), recommendations based on natural-language search and the latest fashion trends to analyse and predict customer preferences. Business in retail fashion is often based on a ‘gut feeling’, on what will work and how much will sell. Therefore, retailers are increasingly dealing with the crisis of unsold inventory. According to IBM, the cost of dead inventory to fashion retailers in the US alone is estimated at $50 billion. And in India, with consumer behaviour differing from one suburb to another even in the same city, keeping a fast-moving inventory is a bigger challenge.
For instance, IBM is now helping the Indian arm of the Danish retailer Bestseller on its journey to becoming a fully AI-driven enterprise. Initially, the management shared details about its products and Watson came back with analysis on how they will perform in stores. Today, the AI platform has a larger role. The insights from IBM drive sales and Bestseller plans to use AI to determine the right assortment for each store, predict the next product to incorporate into its mix, and improve the efficiency of its supply chain leading to a reduction in unsold inventory.
Earlier, IBM worked with designer duo Falguni and Shane Peacock to create a line with the theme ‘the future of Bollywood fashion’. Using Watson’s visual recognition Application Program Interface (API), nearly 600,000 publicly available fashion images across the world from 2006 to 2017 were analysed. Watson examined 5,000 images from social media sites and 3,000 fashion-related images in the form of Bollywood movie posters from 1970-2010. Watson’s analysis of more than 100,000 prints and patterns helped the designers create unique patterns and made the line one of their most successful.
“In fashion, you don’t look only at historical data to do the forecasting; you have to look out for what the fashion trends are for the year and they can change constantly,” says Godbole. That dynamism combined with hyper-local demand sensing is what Watson made available at the point of sale. “This kind of granular analysis at a stock keeping unit level can be a gold mine for retailers,” he adds.
Building blocks
Blockchain is seen as the next peer-to-peer technology that could revolutionalise the economy. Needless to say, IBM’s lab has been tinkering with it for years now. “Now it has become a business unit for IBM,” says Dasgupta. The company believes that the framework it has developed can be a game-changer.
Business is all about transactions, or delivering goods and services for money, and keeping a record of these dealings. In traditional business, members maintain individual ledgers for transactions, leading to errors or even fraud. In blockchain, with a single shared ledger, members can see all details of a transaction end-to-end, tracking assets and reducing costs for all parties on the network. But, people can be wary of sharing sensitive information. Therefore, a trusted moderator can be a big asset.
“Participants can be reluctant to share information that could help the network operator derive powerful insights and save significant cost or improve overall efficiency for all parties. In IBM’s framework, such network insights can be provided in a trusted and verifiable manner without revealing the private information,” says Vinayaka Pandit, senior manager, blockchain, IBM Research India.
For instance, industry platforms such as TradeLens (developed for shipping and logistics) and IBM Food Trust (developed for food supply chain) contain a treasure trove of insights on delay patterns in the logistics network and spoilage patterns in the food supply chain.
This data can result in considerable time and cost savings for both logistics providers and clients. “It will go a long way in mitigating privacy concerns of data owners and lead to a wider acceptance of blockchain platforms,” says Pandit.
IBM is also working with clients to make sure they are blockchain-ready (See: Block gain). The tech company has had more than 500 such clients across industries including education, food safety, insurance, luxury goods, supply chain management and trade finance. One such collaboration was with the Mahindra Group, to create a common platform for Mahindra Finance’s supplier-to-manufacturer transactions.
Invoice discounting and the process of bundling and selling invoices at a discount is a major source of working capital for many suppliers. This solution allows more suppliers to have greater access to credit. Traditionally, trade financing can be cumbersome with each party maintaining separate ledgers, and human errors can lead to payment delays. This application allows all parties to update their part of the transaction and view them in real time. The cloud-based application was one of the first blockchain-enabled projects outside of the traditional banking industry.
Whether they are making supply chains more efficient and transparent or improving farm productivity, they are on track to redefine how business is done. The second largest contributor to patents filed by IBM globally, IBM Research India shows no sign of slowing down. For them, the world is their lab.
