W Power 2024

Building AI solutions for an inclusive world is the need of the hour

By Harichandan Arakali
Published: Jun 24, 2020

A group of experts came together recently to discuss the role of AI at the intersection of enterprise and society and how such technologies could be harnessed for the benefit of all. The experts participated in a tweetathon organised by Forbes India in partnership with Microsoft India.

They discussed a wide range of topics including AI for everyone, pre-requisites for building good AI solutions, the challenges of building and adopting such solutions, AI for inclusion and helping persons with disabilities. and various use cases.

AI for all

“In the current environment, AI adoption in healthcare would be my no. 1 choice — whether it is diagnostics for cancer, lung health, diabetic retinopathy — there are a number of interesting solutions already available that need to be scaled,” said Sangeeta Gupta, senior VP and chief strategy officer at Nasscom.

Apart from improving productivity at an individual level, AI can enhance quality, improve accessibility and reduce costs of services to every citizen, said Vijay Bhaskaran, Partner – Artificial Intelligence, EY India. Some examples are telemedicine and diagnostics, personalised learning, content creation, public services, safety and security, customised financial planning, fast claims/loans processing and so on, Bhaskaran said.

Cross-cutting areas which impact all of us include public health, education, disaster prevention, conservation, and agriculture. Every citizen benefits if AI is deployed across these sectors, promoting diversity and inclusiveness, with eyes on fairness and democratisation, said Adnan Masood, Chief Architect of AI and Machine Learning, UST Global.

AI is needed the most in every sphere of development and social sector, delivering value to citizens, said Shanti Raghavan, founder of Enable India. From livelihoods to learning solutions, AI can make an impact, Raghavan said. Just as e-commerce sites prompt users with messages like ‘customers also bought the following things’ imagine AI enabling a confused fresher who is still finding her way in the world with information such as “people with profiles like yours have chosen the following jobs and careers.”

Challenges of adoption

Challenges in the way of building strong AI solutions are many. Among them are lack of comprehensive AI policies with respect to data protection, privacy and ethics; lack of adequate open data sets to build and train AI models to achieve the desired results; enterprises unable to link AI to revenue growth or sustainable business growth; lack of adequate academia/education funding for AI research; and providing teachers adequate training to adapt themselves with the new ways of AI-enabled teaching; most of the schools in India lack basic IT infrastructure to support education through AI, said EY’s Bhaskaran.

There are challenges to adoption of AI in India — Apart from access to data, adoption of AI solutions in India is still low. Startups building AI solutions have to go through many proof-of-concepts before the solutions can be commercialised. “There is a need for innovative approaches for leapfrogging to an AI-enabled world,” Nasscom’s Gupta said.

Operationalisation challenges which get amplified in high-stake decision making are lack of MLOps (Machine Learning and Operations) discipline, scarce data science skillset, opportunity costs of AI investment, change management and AI adoption concerns by end users, and not having AI governance strategy, UST Global’s Masood said.

Important challenges in early education include missing focus on foundational literacy and numeracy, rote learning with a lack of conceptual understanding and application, and the need for pragmatic teacher development, said Anustup Nayak, Director of Classroom Instruction Practice, Central Square Foundation.

Covid-19 impact

The experts put their discussion in the Covid-19 context, which is an urgent challenge. Covid-19 has pushed innovations from labs to solving real-world problems. Avenues are multiple—digital health platforms, citizen services with bots, digital commerce, remote working, and thermal attendance, all of it enabled through the cloud, said Geeta Gurnani, Director, ISV Partnerships, Microsoft India.

AI is enabling companies to rapidly adapt to the new norm with Covid-19. In healthcare, for instance, organisations are using AI for: Command centre for pandemic management, facial scans to identify patients with fever, accelerated drug discovery, disease forecasting and so on. India's Arogya Setu app has enabled cloud and analytics to identify hotspots and community spreads; it has helped India to provide early warnings for better preparedness to the pandemic, Bhaskaran said.

AI and cloud computing have been of great use for tackling the Covid-19 pandemic—from modelling the protein folding of the virus to fighting fake news, from drug discovery to Covid-19 war rooms and telemedicine. Nasscom, in partnership with leading IT companies, has built an AI based Covid-19 dashboard that can help state governments track diverse data sources and plan both the unlock and steps for healthcare infrastructure, Gupta said.

The pandemic has brought both unparalleled setbacks and unforeseen opportunities. Many of the gains in access to schooling in the last few decades will see setbacks. The learning loss for the poorest children will take years if not decades to recover from, Central Square Foundation’s Nayak said.

Yet for the first time, the pandemic has created a new habit of teaching and learning online, even if it is restricted to a thin sliver—a habit that will alter learning over the longer term, he added. Tech can facilitate inclusion with targeted learning at the level of each child, low-tech and low-effort access points (voice, text, radio, tv).

Fostering inclusion

A critical imperative in the adoption of AI and related technologies is to deploy them to make the world much more inclusive for persons with disabilities. There are a billion people with disabilities in the world, and the panelists tweeted about the need for inclusion.

This is an important imperative and AI can play a key role in building an enabling environment. Assistive glasses, virtual assistants, spoken description of objects for visual impairment are some of the solutions available already, Gupta said.

“We run a rural social networking platform called Enable Vaani for persons with disability (PWD),” Raghavan said. It is an IVR based moderated platform that doesn’t require an internet connection and people can use basic feature phones to use it. “We have reached 1 million calls with a user base of 30,000 users across 20 States. And with AI, can I expect to scale it to the next level?”

At Enable India, technology is used in every aspect of Inclusive solutions—from concept stage to executing to monitoring the outcome. Enable India has a learning solution 'EYE tool' for persons with vision impairment to learn computers on their own. This is used by many visually impaired across India and the world. They are now working with leading companies such as Microsoft India to improve this using AI.

They are also working with Microsoft India to deploy a Q&A maker tool which can answer questions posed by candidates from across the country. This will help us to be responsive to different candidates with a disability and their parents. We have so much data and case studies, it is impossible for a small support team at the help desk to answer the unique questions. We only hope to solve this with AI, she said.

Vijay Bhaskaran added: Differently-abled people often are not treated as equals by many. This is expected to change as AI is not just aiding them but unlocking their hidden potential by enabling them to participate in the job economy and innovate in ways previously unknown to society. For instance, extended reality (XR) enables employees with motion disabilities to be virtually present in the site to carry out tasks just like how a normal person would do. (e.g. Microsoft Remote Assist, Vuforia Chalk).

For differently-abled people, AI for Accessibility has extended beyond MSR's Seeing Eye. He saw inclusion across the spectrum with auto-captioning, autonomous vehicles, facial recognition assist with low vision, language facilitating comprehension for cognitive disabilities, Masood said.  All of this helps differently-abled people with unemployment rate more than twice their non-disabled peers. Outcomes change with improved workplace culture and inclusive hiring.

Inclusion helps drive people-centric innovation by using technology to unlock equal opportunity for everyone, Microsoft’s Gurnani said, adding that fundamentally AI can help remove biases, for example when people are at different levels of spoken and written language skills. AI tools within Microsoft365 are an example of how to weave in inclusive ways. To drive the right inclusion culture one needs to put people at the centre of the process and then bring in technologies.

Technology use-cases

Beyond inclusion, there were several use-cases that the panel of experts tweeted. According to Gupta, AI and IoT can help in maximising energy generation while minimising maintenance costs and planning optimal usage; enable efficient supply chains; tap weather data with AI for crop management; traffic management and so on.

Citizen inquiries, GST refunds / ITR payments, retrieval and maintaining land transaction records that are time consuming most of the times are the immediate areas where AI can play a role, Bhaskaran said. In the long term, AI will enable governments to draft policies that cater to all segments of the population by gathering insights from the past policies, census data and other macroeconomic data.

In education, tech and AI can help with personalised, data-driven approach to instruction and assessment, and self-paced learning. The role of tech and AI is not to replace teachers but to augment their capabilities—automate repetitive aspects, analyse performance gaps, and recommend content, said Nayak.

AI continues to bring data and digital technologies together in aggregating environmental data around the world and put it to work in a new ‘Planetary Computer.’ which can be used in enhanced environmental decision-making, Gurnani said.

Pre-requisites

The right use cases, data, talent, policy and, most importantly, a well thought through execution strategy are among prerequisites for building AI infused solutions, Nasscom’s Gupta said. These are not necessarily available in India—one must especially focus on data, talent and policy.

Posing the right questions before building AI solutions is critical: Does the use case identified have clear social impact; Do you have access to the right data? Do you have the right capabilities to develop it? Services that had earlier limited reach will be more accessible and affordable with AI, Bhaskaran said.

Being mindful of bias in data and models, algorithmic transparency, power and information asymmetries is key for positive AI impact on economic inequality. Tools like FairLearn and InterpretML help, said UST Global’s Masood.

Masood concluded: Notably Microsoft Team Data Science Process addresses Business Understanding, Data Acquisition, Modelling, Deployment and Customer Acceptance. TDSP is an agile, iterative data science methodology to deliver intelligent solutions. Leveraging this digital adoption to promote and improve public health, environmental sustainability, deployed in an ethical, inclusive, and accountable manner would be a pioneering step towards AI for addressing the diverse challenges faced by citizens.

Disclaimer: The views, suggestions and opinions expressed here are the sole responsibility of the experts. No Forbes India journalist was involved in the writing and production of this article.

©2019 New York Times News Service

Post Your Comment
Required
Required, will not be published
All comments are moderated