Signaling the intent of the Indian government to deploy frontier technology in combating real-life problems, policy think tank Niti Aayog has started work on three proofs of concept using artificial intelligence in the areas of agriculture, healthcare, and regional languages.

Two sources confirmed the developments. “We at Niti (Aayog) are looking putting factors in place that will help in catalyzing AI across sectors. Looking at India’s population we are looking at maximizing reach,” said one of the sources, a top-level Niti Aayog official.

Niti Aayog has formed a committee on frontier technology of six people and will get in touch with global experts, academia, companies, and other ministries to take the projects on the ground.

“The most important use of technology is when you use it for uplifting the economy and improving the quality of lives of the people,” said the second government source. “Most of the projects that Niti Aayog is taking up is in line with this.”

The second source said health care and agriculture have a big impact on a large number of people. “Global economy is transforming into a digital economy thanks to the development of cutting-edge technologies in digital space, machine learning, artificial intelligence, internet of things, 3D printing… Niti Aayog will initiate a national program to direct our efforts in the area of artificial intelligence, including research and development of its applications,” finance minister Arun Jaitley had said in his budget speech in February.

Improving farming

Niti Aayog has partnered with a large technology company (name of which was not disclosed) to provide real-time advice to farmers. “It will use satellite and sensor images, real-time weather data, drone images if any, soil health data etc. to provide the advisory,” said the first source.

In the PoC, soil models will be tested in 25 districts. The analysis will provide the requirement of water and fertilizer in the soil. It will also advise the farmer on when to sow the seed and how much water is needed before sowing.

“It is important to make sense of the existing data, learn from the past available data and predict for farmers,” said the first source. “Using AI and analyzing the data, it can also be predicted what quality of fertilizers should be used in the soil for the particular crop.”

Going forward, the government wants to implement AI-based farming predictions in 100 districts. Amitabh Kant, CEO of Niti Aayog had told earlier that it the policy think tank along with state and central government will put out a fool-proof mechanism to help farmers get the right price for their produce.

According to market research firm MarketsandMarkets, AI in agriculture will be worth $2.63 billion by 2025 — mostly driven by rising population, adoption of information management systems, and need to improve crop productivity.

Ashok Gulati, agricultural economist and former chairman of Commission for Agricultural Costs and Prices is skeptical about mass deployment. “It will take 10 years for AI to influence agriculture,” he said.

Gulati added that the technologies are available, but it will also need massive investments that the farmers can’t do. “So far, all crop cutting experiments using technology have been bogus,” he said.

But, that has not deterred Microsoft to use AI farming in a few dozen villages in Telangana, Maharashtra and Madhya Pradesh. Once enrolled, the software sends farmers text messages with sowing advisories that have sowing date, land preparation, soil test-based fertilizer application, according to Microsoft.

Indianising technology

India and the importance of the local language are inseparable. The government understands it better than anyone else (ministers know the impact of local language, especially during elections).

“Using NLP, Niti Aayog is making available a set of tools for developers to build chatbots in local languages,” said the first source. “It’s something like teaching villagers mathematics in their local language.” NLP is short for natural language processing.

Chatbots not only offer scale but also provide personalized learning, an expert said. “Every child should have the right to learn in their native language. Having chatbots in regional languages is a great way to provide regional language teaching at scale,” said Richard Delanty, founder of Into23, a Hong Kong-based translation and localization company. “Making them multilingual would also be a great way to teach children different languages from a young age. If lessons are available in their local language, whether Tamil or Bhojpuri, you could also have the same lessons in Hindi and English.”

The team at Niti Aayog is looking at multiple use cases with creating chatbots in local languages — education, banking, and healthcare being some of them. Local languages are the biggest barrier to technology, said the second source and Niti Aayog is looking at breaking that.

However, the sources said that the PoC in NLP is in its very early stage and will take some time before significant deployment gets done.

Controlling cancer

A third area where work is in its advanced stages is the detection of cancer using AI. This was one of the areas where Niti Aayog started work quite early. The project inside Niti Aayog is loosely called bio-banking of images — a PoC, which is being done in partnership with Tata Memorial Centre.

“The whole idea is to scan pathological images and find traces of cancer cells using AI,” said the first source. Collecting the images is not the biggest problem. “Annotating the data, marking where the cancer cells are and do the inferencing where is required is the biggest challenge.”

Rohit Pandey, co-founder and CEO of SigTuple, a healthcare startup that helps in diagnosis using AI, puts the challenge in perspective. “To analyse what is called peripheral blood smear slides there was no automated way to do it,” he said. Doctors have to do it under a microscope.

SigTuple has found a solution to it, very similar to what Niti Aayog is working on. “This can be done in an automated manner. AI can identify precancerous and cancerous cells. The AI models are there on the cloud and generates the report,” says Pandey.

The number of labs in India with testing for cancer is very limited. “This niche can be generalized using AI model. Even in a normal lab at a Rs 350 complete blood count test cost cancer detection can happen, which is not possible today,” he adds.

At Niti Aayog, the team is in touch with experts from universities of Stanford and Maryland to build this into a solution that will go into the workflow of doctors. “We are looking at how to build this as a service model for secondary and tertiary healthcare centers,” said the first source.

The data resides with Tata, which will partner with a technology company to implement the solution at multiple government hospitals and smaller private hospitals so that doctors can use this as a service whenever they wanted.

However, the sources at Niti Aayog admit that deployment of AI in various formats will require capital. That is also the reason why a large number of deployments will be done by the private sector to speed up the process.

It has been earlier reported that Niti Aayog is working on a national artificial intelligence policy, which will detail out the scope of adoption and commercialization of the technology. It is said that the policy will look at a higher deployment of AI in the areas of education, healthcare, agriculture and creation of social infrastructure. “Meetings on the paper is being held, one happened last week,” said the first source.

---