Gen AI: What Can It Generate?

After the telecom revolution, the artificial intelligence (AI) revolution has swamped the world in just the last 12 months. It has penetrated every sector of the economy. I will segment my thoughts into eight different hot areas of debate, discussion and application.

Experts believe that AI can contribute USD 15 trillion to the global economy by 2030. India’s contribution alone is gauged at about USD 1 trillion. As of today, the AI market is captured by the US to the extent of 40 per cent, followed by China at 31 per cent, France & UAE at 9 per cent and Japan and the UK each at 4 per cent. Mastercard AI chief has very aptly suggested that AI is mastery of data, and it has minimal value if it is descriptive, large value when it is predictive, and colossal value when it is prescriptive. 

For a country like India, with a total tech force of 5.4 million, there is a huge need for AI skilling and Infosys has taken the lead, by hiring Nvidia to train 50,000 of its employees. Talking about the next development in AI, Srikanth, co-founder of Fractal Analytics says, AGI or artificial general intelligence, which can equal human abilities and can mentally do what we can do and hence can be a darling or a devil. The biggest use of it is in “predicting health issues and customising preventive care and treatment” to enhance longevity. Its strongest curse is in “surveillance overreach and autonomous weaponry". 

Hence he advocates the creation of a much-needed “ministry of Artificial Intelligence”, which cannot be an adjunct of MEITY. This will enable India to create solutions for use cases specific to its “demographic and socio-economic" environment. He who masters AGI is the future world‘s hegemon, he predicts.

AI developments in India: Digital India has already created 16 digital public goods (DPGs) and 22 are partially developed. In addition, 54 are under different phases of testing, and hence in a few years, India could lead the world with 100 DPGs, which could be replicated globally. India has created 24 science and technology parks and 24 technology business incubators for research and development. 

As a result, India has deep tech startups, one of which came into prominence recently, when it created a 3D-printed rocket and got it launched by ISRO. IITs have recognised the value of research and created research parks in all its over 20 plus locations. Two budgets ago, India announced the creation of three centres of excellence for AI. In the interim budget this year, a provision of Rs. 10,000 crores was made, for supercomputing and quantum computing capabilities to be created on a plug-and-play basis. 10,000 graphics processing units (GPUs) are hence on the anvil, in the next five years through PPP, to be leased out at subsidised rates to startups and researchers. This need was felt when individual startups were importing GPUs to build BharatGPT and indigenous generative AI platforms. 

VizzhyGPT, a large language model for the health sector, tapping data from hospital chains, was enabled by IIT Mumbai, in collaboration with the BharatGPT ecosystem. The extent of research and development taking place in India can be gauged from the fact that OpenAI acquired database analytics startup Rockset, for a few 100 million dollars. It was developed locally by two alumni of NIT Trichy and BITS Pilani. This revolutionary startup develops real-time search and analytics databases and deploys AI in various applications, including chatbots and anomaly detection.

India is collaborating with 29 member countries of Global Partnerships in AI (GPIA) for AI applications in healthcare, agriculture and other areas. Tejpreet Singh Chopra an AI hands-on expert, has used his imagination to connect wind turbines with sensors, which provide online data, for AI to prescribe clear dates for preventive maintenance. He has used the same reasoning for cranes used in large ports, fitting them with sensors, and ensuring that AI predicts their maintenance schedules. 

At Bengaluru airport, he has got 500 vision cameras installed to improve airport efficiency and management, through data collected every second. Use cases of this type are enormous where sensors can collect data from machines, and AI then predicts when a particular machine needs its repair and overhaul. LAM-powered devices came on the market in April 2024, providing AI agents to perform complex tasks.

Ethics In AI: 
The USA has developed AI protection guidelines and notified them a few months ago. EU also has created global guidelines which 18 countries have endorsed, and 17 more are in the process of doing it. The USA and UK are also setting up AI safety institutes to regulate any negative effects of AI applications.

India needs to look at these precedents and formulate its own AI ethics policy soon, which is safe, trusted, open, and accountable for all AI use cases. It should also work closely with cybersecurity institutes like CERTIn and the National Critical Information Infrastructure Protection Centre (NCIIPC).

The most important element in such a policy should be its flexibility to deal with different types of situations that may arise. It should be able to curtail any bias or discrimination in data use for AI& it must also include standards of platform design, through transparent means, to ensure the public is not duped. All this is becoming pertinent and imperative, because Corpus data captures not only of sound and the written word, but also images and gestures, and hence alters the data ecosystem radically.

Fears Regarding AI: 
Press has reported several challenges including security and privacy-related issues concerning AI. There is discussion and debate in research circles about environmental sustainability being affected. Some experts also feel that it would enhance litigation as trademarks, copyrights and trade secrets would be at peril. 84 per cent of respondents in a survey expressed their fears that AI’s strength getting concentrated in a few hands, can hit the competition and cause market imbalances.

Conservative thought on AI is that it should be banned for predicting human behaviour, predictive policing and emotion recognition systems in schools, as these would affect the freedom of the individual and violate the Constitution.

AI Applications For Public Benefit In India: 
Through concerted research and development, several AI-based apps have been created in the country to meet the needs of different sectors of the economy. About eight prominent ones are Bhaashini,a translation app that has enormous value even for politicians' propaganda in different parts of the country. The Kisan Chatbot reaches 110 million farmers, acquainting them with their entitlements. Cotton Ace can predict for the farmers, the precise time for pesticide spraying. In agriculture the Sowing App works on the past 45 years' rain data, to predict the exact week when different crops should be sown. 

Urbania is a municipality app based on traffic data to guide better traffic management. Atman App works on past chest X-ray data and gives red signals if pneumonia trends are visible. Dozee App, uses vital past patient data, giving early warning for patients to take quick medical help. Rabbit App is a large action model providing AI-enabled choices. Shiksha Copilot has been developed locally to improve learning outcomes and empower teachers and is already operational in 30 rural and urban schools in Bengaluru. Similarly, in research, teaching and academic circles, lots of apps are being developed all over the country, based on the use cases that confront the community, requiring AI-based solutions.

AI And HR: 
The greatest fear among students in the higher education institutes of India today is the loss of jobs with the advent of AI. Since there is uncertainty and lack of clarity on this subject, and various postulates are floating all over social media, the confusion is getting enhanced. Wall Street Journal's recent report indicates that with ChatGPT and platforms alike, the number of freelance jobs posted on social media platforms has already dropped by 21 per cent. But some experts believe that these technologies cannot replace human creativity, and will enhance and enlarge the avenues for creative expression and out-of-box solutions.

My personal view on this matter is that AI will eventually replace all routine jobs and hence several professions like accountants will get wiped out. But creative pursuits can never be replaced by AI and in fact will get emboldened due to better data availability, which can then be calibrated and attuned to generate far greater creative expression. In addition, in areas of HR and recruitment, AI can play a significant role in shortlisting the right candidates from thousands of applicants, a task which is done manually today. A day is not far when AI may choose the final candidate also, but I still believe that it should be restricted to shortlisting valid candidates, with pre-listed criteria, and then an interview- virtual or physical- is necessary to select the best candidates since non-verbal cues are a very important component, in final selection.

World Trends In AI: 
Nvidia, the creator of hardware that powers most generative AI software, has jumped in market valuation from USD 300 billion in 22 to USD 2.3 trillion today. TSMC, the Taiwan semiconductor manufacturing company, builds the GPUs that Nvidia designs and commands a market cap of USD 1.5 trillion, being the most advanced factory. ASML, at USD 370 billion, is Europe’s most valuable company creating complex machines like Twinscan Exe:5000 which prints nanometer-long silicon wafers that make AI models possible. These developments have made a journalist coin the term that “AI hardware rather than software, is gobbling the AI world”.

JP Morgan is giving AI prompt training to all its new employees and equipping them with the art of communicating effectively with AI systems. A large bank like JP Morgan is then able to take all the customer-related queries and initiate action quickly. It has been estimated that training 70,000 such employees, across various departments, can generate about USD 1-1.5 billion in value. Envisioning AI’s future, the CEO of JP Morgan says that AI will revolutionize longevity, and cut working time to half a week, from five days a week.

Google has upgraded its most advanced gen AI model Gemini 1.5 Pro enabling answers to be generated from thousands of pages of text or video content, facilitating such consumers enormously. It has also developed Project Astra, which is a real-time multi-model AI-assistant prototype that comprehends its surroundings, answers questions and serves as a universal AI agent bringing enormous comfort in daily life.

AI’s risks have brought China and the US together in Geneva recently to discuss and manage these risks. The USA is concerned about China’s misuse of AI while China is worried about the restrictions and pressures on AI put by the USA. One can easily see AI becoming a geopolitical tool,& hence both countries are into large funding for AI research, to prevent being left behind in this race.

In the field of health, AI deployment has immense potential for improving patient outcomes, optimising resource utilization and advancing medical research to its full potential, it requires the highest adherence to strict data protection too. The intricacies of integrating data and collaborating with different hospitals are essential, for efficient data flow and interoperability but a tough task to achieve. User training and patient awareness are important, but even more significant is the readiness and abilities of doctors and allied staff, to effectively deploy AI, without compromising its ethical use. 

Comprehensive user training manuals, rigorous testing and ongoing evaluation are critical, to ensure AI solutions are accurate, robust and reliable. The developments in Minimal Residual Disease (MRD) are very impressive because they enable decisions about when to escalate therapies, and when to safely deescalate treatments based on real-time data. MRD testing is growing rapidly in cancer detection of all kinds, and hence an era of a new standard of care and treatment, of several diseases is visible.

India is the second largest user base of ChatGPT at 7 per cent, outside the US. It has become the most commonly used, easily reachable and freely accessible tool. The world's largest software developing platform, GitHub, added 3.5 million new developers in India last year, making a total of 13.2 million and expects to surpass the US by 2027. GitHub Copilot is revolutionising the work of developers and making their code faster by at least 50 per cent.

Suggestions For Expanding AI Use In India:
Realising that 70 per cent of the 31,000 tech startups in India are using AI to augment productivity, I feel its largest use can be exploited for the 100 smart cities that India has embarked upon and which must all deploy AI for maximisation of smart benefits. Data on filth areas around the country needs to be collated and then AI used effectively to identify action areas regularly, to create a real Swacch Bharat, rather than leave it as a mere slogan.

Free water and free power are freebies, each state is extending without correct data. Using beneficiary's data, AI use can prevent misuse of taxpayers' money. Rs 5.1 lakh crores of cost overruns have been identified in 438 projects by MoSPI. AI use can track why, where and how it is happening, and what remedial measures need to be taken to save this massive loss on public exchequer.

Using WTO trade data globally, AI can be usefully deployed to identify demand and supply gaps for goods and services all over the world. This can help the Ministry of Commerce and Industry devise a strategy to identify relevant items to boost exports and fill such gaps. With a USD 85 billion trade deficit with China, comprising many non-essential items, the time has come to use AI to cull out all such items, which can then be substituted locally, to create the Atmanirbhar Bharat, a major slogan right now.

Most data records in hospitals today in India are in hardcopy which patients have to keep in storage for future use. Digital India cannot afford its hospital infrastructure to lag and hence a directive should be issued to all hospitals to digitize their patient records by a fixed deadline. This will enable the segregation of critical patients and extend quality care to them, besides making hospitals paperless.

While EVMs have made elections easy, electoral rolls need to be linked with Aadhaar, so that AI can ensure no fraudulent entries find a place in the polling process. It will also enable the reduction of staff strength from 8 to 2 per booth, with one handling the control unit of EVM, and the other using his laptop for identifying the voter and inking his finger.

Court pendency today is running into crores, without an assessment of what can be easily screened out and resolved. Good data coders are required for effective use of AI in separating wheat from chaff and then using simpler methods to resolve minor disputes, through mediation. 5,000 millionaires left India last year and 4,300 this year. While this is not a priority, there is no harm in using AI to find out reasons for their exit and identify how they can be brought back to save and invest within the country. PM Awaas Yojana has been operational for the last 40 years and a lot of data is available from the government. AI must be deployed here to find out whether duplication and triplication are taking place, and why we are not able to reach a finality in this longest functional scheme of all governments.

National science centres are 25 in number and they must create videos on AI to spread public awareness about its manifold use, in all spheres of human existence. Niti Aayog must assist the private sector by empanelling AI firms, after due diligence, so that they can be utilized by the private sector, especially the MSME, for any use cases that they can identify.

All examinations are now susceptible to fraud due to the use of AI. Hence, I have the following seven suggestions to make for all Controllers of exams, throughout the country

Facial recognition devices will have to be put outside most exam halls, to stop impersonation.

CCTV cameras inside exam halls would become necessary, to check if cellphones have been clandestinely smuggled inside, and are being used nefariously.

Banning of cellphones would require body scanners at the entry of each exam hall, to maintain the integrity of the system.

Professional invigilators will have to be deployed, with certified integrity, and a watch will have to be kept over them, through command and control centres to be established for all major exams.

Paper setters will have to be told that no question should ever be repeated, otherwise AI can predict question papers based on past data.

All examiners will have to be cautioned ,that if identical answers are found anywhere, they must be vigilant about ChatGPT,and take appropriate action to mark such papers accordingly.

Global impressions indicate that the pace of AI adoption has been too quick, causing enormous challenges for decision-makers, business entities and employees. This challenge is getting complicated, without a secure and responsible strategy in place. Education, after health, is the most intensive user of AI and 60 per cent of all educators are already using it, as per the latest data. Transformation of learning experiences is the biggest positive feedback, along with preparation, teaching and student engagement benefits. 

Hence, it is becoming critically important to skill the students in emerging technologies, if they are to be made ready for tackling the future. The time has come for the Bureau of Indian Standards (BIS) and Standardization Testing and Quality Certification (STQC) to collaborate with the US National Institute of Standards, to develop AI safety and trustworthiness standards, for wider reach and greater spread of AI technology throughout India’s large population. Finally, looking at India’s huge unemployment figures, the Labour Ministry needs to hire experts immediately to estimate likely job losses and ways and means for adapting such workers through job substitution. The economic survey touched onbtjix point. Microsoft CEO, Nadella, was in India in February 2024 and announced skilling two million Indians by 2025. Perhaps this figure may fall far short of India’s requirements.

J K Dadoo

Guest Author J K Dadoo is a retired IAS official