Twenty years ago, Steven Spielberg made a movie named A.I. on our increasing distrust and fear of artificial intelligence. That paranoia has subsided but it lingers. We aren’t as worried about AI taking over the world as much as we are worried about it bungling things up. The new intelligence seems to be ridden with the same biases that we lug around.
Companies that have little option but to adopt AI fret over the consequences and lack of accountability of handing over decision-making to lines of code. Therefore, IT vendors have to do more than just make AI tools—they have to win the trust of client companies. That is one of the goals IBM Research has set for itself.
With companies accelerating their digital journey, IBM’s research team is working on innovations, particularly in AI and hybrid cloud.
In AI, besides promoting natural language processing (NLP) and automation, the research lab has chosen the building of trust as an area of focus.
In terms of deploying AI, it is developing a framework to ensure data quality, fairness, accountability, explainability and even alignment with the values of the client company. “Trust is essential to AI adoption. IBM is helping companies achieve greater trust, transparency, and confidence in business predictions and outcomes by applying the industry’s most comprehensive data and AI governance solutions,” says Gargi Dasgupta, director, IBM Research India. While this is a global effort, the Indian team has been a key contributor to developing toolkits such as AI Explainability 360, AI Fairness 360 and AI FactSheets. IBM India is the second-highest contributor to the MNC’s patents with 930 patents granted to inventors from India.
“(IBM’s) India labs are closely aligned with our strategy, and pioneering work from these labs is integrated into IBM products, solutions and services. For instance, a lot of the core innovation and technology work for the IBM POWER10 processor, which plays a critical role in our hybrid cloud strategy, was designed in India by our IBM Systems Development Lab,” says Sandip Patel, MD, IBM India.
The first two toolkits—AI Explainability 360 and AI Fairness 360—do what their names say. One helps add explainability and the other places guardrails to ensure fairness in AI models which are essentially lines of code that are capable of arriving at decisions without a human’s guiding hand.
The third one works like an ingredient list at the back of a biscuit packet. Once you know the ingredients that have gone into an AI model that you are using, won’t you be more at ease using it? Yes, they thought so, too.
Once trust is built, ease of deployment would come in handy. Therefore, the research group’s other area of focus is automation of an AI model’s lifecycle—from curation of data to deployment. For this, they have developed AutoAI. Dasgupta says that it has reduced the time to train a model “from weeks to days to even hours”.
Besides this end-to-end automation, the research group has also developed tools to grade the quality of data being fed into the system and tools to identify biases during the processing of this data and even mitigate the effect of these biases.
Imagine building the Hulk and giving him the ability to control his anger by reminding him, ‘Hey, maybe your brain is misreading (bias) that action (data) as a slight”. That has to be useful.
Translating Data into Language
Few people like to read reams of dry business data. Can you imagine picking a company’s report on energy consumption over the past six months for reading at a café? It is better to palm that job off to a bot.
One of the areas of focus for the research lab, therefore, is NLP. “We are helping AI understand the language of business by advancing its capability to generalise, reason and recognise the relationships between words in context and their unique nuances which are an important part of human communication,” says Dasgupta. So, the AI tool can now scan information on PDFs, including tables and charts, and make sense of its meaning at different levels—from literal to implied. Its capabilities range from low-level understanding such as sentence splitting and recognising parts of speech to higher-level understanding such as the sentiment behind the text. If you are new to the business world and its jargon, and would like to ask and retrieve data using everyday language, the research lab is refining the tool for that, too. That is, if you are a facility manager who has asked, “What was my energy consumption pattern in the last six months?”, it can respond with easy-to-understand graphs or pie charts.
Changing the mindset of its client companies towards AI and making it easy for them to adopt it is part of the IT giant’s larger strategy. In the meanwhile, it is also working on cool, individual projects. Dasgupta cites these as the most exciting innovations the lab is working on—an AI application to enable intelligent, self-correcting and climate-aware supply chains for manufacturing, retail, fashion and food; the AI FactSheets; and Common Sense AI.
Canada’s Énergie NB Power, which has been generating electricity and delivering it to tens of thousands of homes, businesses and critical operations such as hospitals, has had to keep its operations going during trying conditions. The area it serves sees temperatures dropping even to -30°C. Its team has to be on its toes to keep the power running and, therefore, IBM’s Outage Prediction has proved critical. The analytics-driven solution gives it a 72-hour window to prepare for outages so teams can be stationed to respond as soon as needed.
AI FactSheets, as said earlier, helps organisations place more confidence in the solution by informing them about who built it, who owns it, who validated it, what metrics were captured, whether it falls within a company’s policies and so on. “By using AI FactSheets, they (business owners who have deployed an AI solution) can collect the information that is relevant to their needs in a timely fashion, producing reports and analytics that are tailored to the goals of the enterprise,” says Dasgupta.
More along this explainability line is another one of the research lab’s projects called Common Sense AI. As IBM’s blog on it explains, it helps AI explain how it arrived at an answer to an everyday question that requires simple common sense. For example, again borrowing from the blog, if a child and an AI were asked what they should do when hungry—eat or play—both may pick the right answer. But the child will perhaps be able to explain itself better than the AI. Common Sense AI now equips the AI to reason better.
Dasgupta sees a real-world application of this in primary education. The AI could converse with children and help them understand why the world works the way it does—for instance, explain to them why we need to give way to an ambulance in traffic.
IBM has always been deeply engaged with education. Dasgupta reminds us that the multinational has a long history of collaborating with academia through numerous initiatives. For example, allowing students to access IBM resources free of cost, conducting training sessions and even having a mentoring programme for students. In fact, IBM has recently made available its quantum computing learning resources to students and faculty of India’s leading institutes. “Now, students can write algorithms in Python and see if they work on an actual quantum computer. This kind of hands-on access allows students to learn in a few days what would take several weeks and leads to a more immersive learning experience as students experiment with the quantum systems using a familiar programming language,” she says.
Technology was anyway travelling at a speed close to that of light till the pandemic came along and hit the pedal. “With the acceleration of digital transformation during the pandemic, many organisations in India turned to AI to address exposed vulnerabilities ranging from an inability to absorb spikes in customer service volume to recalibrating broken or uncertain supply chains. They found AI as a boon to enhance customer engagement and maintain operational flexibility,” says Kamal Singhani, country managing partner – global business services, IBM India and South Asia.
In India, with the faster pace of digital adoption and the Digital India vision, Patel sees technology playing a sutradhar (facilitator) with hybrid cloud, AI, security, blockchain and IoT; helping businesses carve a differentiated edge. “We are on a path to growth and see hybrid cloud and AI as a $1-trillion market opportunity,” he says.
He sees opportunities across BFSI, telecom, retail, manufacturing, travel and transportation, global capability centres (GCCs), and FMCG, among other spaces. “We are assisting our clients in their next chapter of transformation as they build agile organisations fuelled by data, guided by AI insights and able to work in any cloud environment,” says Patel.
Hybrid cloud is another engine of growth for IBM. “IBM is all-in on hybrid cloud and AI,” says Singhani, adding, “Our hybrid cloud platform contributed to strong performance in GBS (Global Business Services).” The global revenue for GBS, which includes consulting, application management and global process services, in the quarter was $4.3 billion, up 11.6%, with cloud revenue for the segment growing by 35%.
The attitude toward cloud adoption has changed among companies. “For years, many clients believed that wholesale cloud migration was the best way to achieve digital transformation,” says Singhani. “However, extensive, complex IT estates cannot simply be lifted-and-shifted to the cloud. Enterprises want the freedom to choose from multiple providers to best meet their business and IT needs—whether they’re building banking solutions, running airline reservation systems, or responding to seasonal capacity demands,” he adds.
In this scenario, IBM’s flexible approach to its client’s cloud needs has helped greatly. “We enable our clients to build an application once and run it on any cloud, and integrate data and applications across multiple clouds—public or private—from any vendor,” he says.
With digital transformation at their doorsteps, companies need a reliable partner that can guarantee security of their data and operations at scale. Who better to do that than the Big Blue?