Science and Technology › Computing and AI
AI and robotics: technology and India's capability
Artificial intelligence and robotics as technology and as national capability: how large language models work, computing power and foundation models, uses in public life, robots, autonomous vehicles and drones. Prelims has asked how large language models work and what artificial intelligence can effectively do.
UPSC has asked
- Prelims 2026: how Large Language Models work
- Prelims 2020: what artificial intelligence can effectively do at present
10 Sep 2026 · 1 Sep 2026 · Prelims and Mains
BHASHINI
India's languages are the first problem any Indian artificial intelligence service has to solve, and BHASHINI (Bhasha Interface for India) is the platform built to solve it once for everybody. It is a layer other services call rather than a product a citizen opens, which is why its reach is counted in websites as well as languages.
- Natural language processing
- The branch of artificial intelligence that lets computers understand and generate human language, spoken or written.
36Indian text languages
23Indian voice languages
35international languages
800+government websites powered
- BHASHINI runs under the Digital India Corporation of the Ministry of Electronics and Information Technology (MeitY).
- BhashaDaan crowdsources speech, text and image data for it.
- Shrutlekh uses natural language processing for real time translation and transcription.
- BHASHINI Sangam, its international workshop, was held in Kathmandu on 1 September 2026 to widen work with Nepal.
- Maharashtra discussed a proposed MahaBHASHINI at a workshop in the State secretariat on 10 September 2026.
PIB, 10 Sep 2026: Digital India BHASHINI Division Organises Workshop at Maharashtra Mantralaya; Advances Multilingual AI for Inclusive Digital Governance (opens in a new tab) · PIB, 1 Sep 2026: Digital India BHASHINI Division Organises BHASHINI SANGAM Workshop in Nepal; Strengthens India–Nepal Collaboration on Multilingual AI (opens in a new tab)
14 Aug 2026 · Prelims and Mains
SraVaani
A language reaches a machine through speech long before it reaches it through text. Most Indian dialects had never been heard by a model at all.
- Scheduled languages
- The 22 languages listed in the Eighth Schedule of the Constitution.
- The Indian Institute of Science (IISc) released SraVaani, a speech recognition model covering 65 Indian languages and dialects.
- Built with ARTPARK and Google, it is free and open source.
- It reaches 20 scheduled languages and 45 regional languages and dialects, over 40 of them unsupported until now.
- That is wider than the 23 Indian voice languages BHASHINI carries.
31 Aug 2026 · Prelims and Mains
The Cyber-Physical Systems mission
The mission that funds India's work where software drives machinery is also where much of its artificial intelligence research sits. It does that spending through hubs rather than through its own departments.
- Cyber-Physical System
- A system in which software, sensors and networks control a physical process, as in a driverless car or an automated factory line.
- Section 8 company
- A company registered under Section 8 of the Companies Act, 2013 for a charitable or public purpose, which must apply its income to those objects instead of paying dividends.
- The National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS) was launched in 2018 for nine years and runs through 25 Technology Innovation Hubs.
- Each hub is registered as a Section 8 company rather than run as a department of its host institute.
- The hubs sit in academic institutions, including the Indian Institutes of Technology (IIT) and the Indian Institute of Science.
- The Centre raised its four best performing hubs to Technology Translation Research Parks.
- They are IIT Kanpur for cyber-security, IISc Bengaluru for robotics and artificial intelligence, IIT (ISM) Dhanbad for mining and IIT Indore for digital health.
- The reward for a strong hub is a wider mandate rather than only more money.
31 Aug 2026 · Prelims and Mains
Param-2
A foundation model is trained once on broad data and adapted afterwards to many tasks, so a country that has one of its own is not borrowing somebody else's. India's came out of the mission's institutes rather than out of a company.
- BharatGen, led by IIT Bombay, released Param-2, a model of 17 billion parameters covering all 22 scheduled languages.
19 Aug 2026 · Prelims and Mains
Women in India's artificial intelligence workforce
The share of women falls at every step from the science degree to the senior artificial intelligence job, and it falls furthest where the decisions are made. The narrowing starts before the degree, with who has an internet connection of her own.
- Foundation model
- A large artificial intelligence model trained on broad data that can then be adapted to many tasks, such as a large language model.
- Parameter
- One of the numbers inside a model that training adjusts; the count of parameters is a rough measure of how large the model is.
Women's share, stage by stage
Science, technology, engineering and mathematics graduates43 per cent
Technology workforce26 per cent
Advanced artificial intelligence roles12 per cent
Senior artificial intelligence leadership10 per cent
- Only 57 per cent of women have independent internet access, against 72 per cent of men.
- Algorithmic bias from skewed or incomplete training data can disadvantage women in credit scoring and in maternal health tools.
- The IndiaAI Mission is MeitY's programme for computing capacity, datasets, skills and safe and trusted artificial intelligence.
- A column argued that women must take part in it as researchers, engineers and policymakers, not only as users.
Indian Express, 19 Aug 2026: To build AI for all, India must build it with all (opens in a new tab)
- Algorithmic bias
- Systematic unfairness in an artificial intelligence system's outputs, coming from skewed or incomplete training data or from the design choices made around it.
Also filed elsewhere
- The national innovation system · on Research funding, institutions and awards
A column in August 2026 argued that leadership in frontier artificial intelligence comes from a national innovation system rather than from a funding announcement.