(Rachel, a 'Replicant', from the 1982 sci-fi thriller Blade Runner)
Former prime minister Manmohan Singh, who passed away recently, heralded the era of economic liberalisation in India with the quote: “No power on earth can stop an idea whose time has come.” It was perhaps Singh's keen sense of history that made him append those Victor Hugo lines to the end of his famous 1991 Budget speech in Parliament.
The forces unleashed at that moment were responsible for the social, political and economic changes that marked the past 34 years of India. Economic freedom helped lift millions out of poverty. It made wealth, until then looked down upon as a sign of dubious morality, fashionable. Above all, it spawned a vibrant middle class of managers, professionals, innovators, entrepreneurs, creators, and many more pursuing a multitude of new vocations.
White-collar jobs, fat salaries, aspirations, new tastes, and uncommon pursuits drove the economic engine for three decades. Conventional wisdom suggests that the expansion of this class helps a country improve its stock in the world in many ways. The middle class greatly contributes to the national economic status, social progress, and transnational influence with their creativity, innovation, and drive. Their hive mind powers the nation and influences its trajectory.
Now, the time for another idea—Artificial Intelligence (AI)—has arrived, and it may not bode well for the middle class, which in countries such as India would mean the very backbone of the economy.
The Disruption
As a phenomenon, AI swept the public imagination in November 2022 when OpenAI unleashed ChatGPT on an unsuspecting world. But the slow creep of the tide began arguably a decade before with AlexNet, an artificial neural architecture built by founders Geoffrey Hinton, Alex Krishevsky, and Ilya Sutskever, the pioneers of AI as we know it. What they demonstrated was training machines to recognise and classify images. That was made possible largely because of Nvidia re-engineering its ultra-fast graphic processing units built for gaming for general-purpose computing. Its programming model and parallel computing platform could process data at supercomputer speeds. Nvidia's processors kept getting faster, each generation in lockstep with skyrocketing demand when the industry discovered that it could deploy the processors to teach machines how to analyse words, voice and images, still and moving. The growth of the industry was so fast that within two years of ChatGPT's release, the Nvidia stock rose nearly 800% in value, helping it pip Apple to become the most valuable company in the world.
Speech recognition and medical imaging arrived close on the heels of AlexNet demonstrating image recognition, followed quickly by ChatGPT and generative AI capable of creating content. So far, so good. It is from the next iteration that AI's relationship with humans really begins to get complicated.
Delivering his keynote address in January at the electronics trade show, CES 2025 in Las Vegas, Nvidia CEO, Jensen Huang, showed a glimpse of what's coming. Next up is Agentic AI, which can perceive, reason, plan and act. The one after that is Physical AI, which could drive vehicles and make general robotics functional. Those would be theoretical forerunners of Replicants in the Harrison Ford-starrer Blade Runner, the 1982 film adaptation of Philip K. Dick's dystopian sci-fi novel Do Androids Dream of Electric Sheep. Maverick billionaire Elon Musk's Tesla is working on an advanced generation of a humanoid robot, which it calls Optimus, that displays remarkable mobility skills even on difficult terrain and real-life situations.
The Upending
Goldman Sachs CEO David Solomon described to the Financial Times how AI is commoditising 95% of documentation work that now takes an army of skilled executives at his bank weeks. In the worst-case scenario, AI could gut Wall Street. The same would happen in other industries.
The number of scientists at drugmakers or metallurgists at material producers could shrink as AI-powered machines discover drugs and new alloys faster. A skeletal group of engineers could potentially run entire public works departments. It is a well-recognised fact that factory robots would make a number of low-skilled jobs redundant, shutting out one conventional road to economic progress and graduation to the middle class.
The domino effect of automation is already reflected in the dwindling number of skilled workers such as fitters, welders, and painters on shop floors. Past technological leaps such as steam and electricity produced more net jobs and better wages as they helped labour productivity and generated new tasks. Occupations that spawned new jobs sustained jobs growth in the post-war US, economists Daron Acemoglu and Pascual Restrepo argued in a 2019 paper. However, that trend is trailing off. Employment and wage growth suffered in the first two decades of this century as new tasks failed to materialise, they wrote. And for the first time in the history of the world, AI will not help net job addition.
The impact of these rapid leaps in technology, if not calibrated, would be devastating, dystopian even. It may sound alarmist but intelligent robots and thinking systems could potentially clear out entire office complexes. A large number of white-collar jobs would be eliminated and innumerable tasks commodity-fied. The requirement of ancillary services, some as simple as fast food and taxis near offices, would reduce as a result. An ecosystem of products and services that has evolved around the working individual since the beginning of the Industrial Revolution would undergo dramatic changes.
Honey, Shrink The Middle Class
Countries such as India will struggle to lift people out of poverty as traditional paths for individual economic prosperity would vanish. New avenues might open up, but those are likely to be fewer and require early investment in education or skilling. India is the world's most populous country and youngest among major economies. It needs an expanding middle class, not a shrinking one. Chapter 13 of the Economic Survey 2024-25 acknowledges this, yet bets on history; "Demand for new products and services due to new technology in a market with unmet demand has created new, auxiliary tasks where labour has a competitive advantage."
The survey rightly proposes creating stewarding institutions to manage the disruption but wrongly suggests that the state should tax profits generated from the replacement of labour with technology to mobilise resources if companies do not behave responsibly. This is not to say that companies should not be sensitive to local conditions, but taxing them would be counterproductive and will result in them moving bases where machines will be welcomed.
One of the deterrents to the adoption of AI would have been cost. The big US corporations developing the technology had so far emphasised the need for billions of dollars as investment. In January, however, Chinese startup DeepSeek shook the world when it released R1, a model that has a radically different way of learning, runs on relatively less powerful chips and costs a tiny fraction of OpenAI's or Google's models. It surpassed OpenAI's ChatGPT downloads on the Apple Store within a week of release. At just $0.55 per million input tokens, the R1 beats OpenAI's o1, which charges $15 per million input tokens. The model reportedly clocks a 97% success rate at coding, leaving o1 behind. DeepSeek could trigger a race to make LLMs accessible and cheap.
The Deliberation
Various stakeholders, including states, international institutions and organisations, business coalitions, and civil society groups, which have been trying to design governing principles and mechanisms for the development and deployment of AI, are congregating in Paris on February 10 and 11 to delve into some of these scenarios. The Artificial Intelligence Action Summit is expected to be somewhat of an emergency meeting, akin to the Earth Summit in Rio in 1992, which brought the environment and sustainability into the mainstream global discourse. World leaders and those helming companies developing futuristic technologies are expected to deliberate on evolving a global AI framework. Future of Work is one of the five themes, which include Public Interest AI, Innovation and Culture, Trust in AI, and Global AI Governance.
Even though Europe has been at the forefront of regulating new technologies, it will likely face pushback from the new administration in the US led by President Donald Trump, who is supported by some of the leading AI players. Many of them dislike the European regulatory approach—central to which is individual dignity, personal security and privacy as opposed to the centrality of freedom in the US—as harsh and anti-innovation.
India is a key US partner in technology development. In May 2022, the two countries announced a US-India Initiative on Critical and Emerging Technologies. While the broad agreement encompasses strategic technology partnership and defence industrial cooperation, it has a specific clause on AI that is vital to India's interests. It says the two countries are committed to “drawing from global efforts to develop common standards and benchmarks for trustworthy AI through coordinating on the development of consensus, multi-stakeholder standards, ensuring that these standards and benchmarks are aligned with democratic values”.
India, A Goldmine
This broad understanding on aligning with democratic values should presumably include protecting jobs and livelihoods in India. After all, India's large, diverse population makes it a goldmine for the AI industry. Central to developing artificial intelligence and deep learning is tokens, which are broken down fundamental bits of data, akin to bytes or bits in normal computing lingo. AI models use these to process and learn from data. The more data a model consumes, the better and faster it gets. Tokens transform worlds into knowledge, turn ideas into videos. Tokens teach robots to move like their masters,” Huang's introductory presentation at CES 2025 said.
Earlier, in an October 2024 fireside chat with billionaire Mukesh Ambani in Mumbai, Huang said that Nvidia and Reliance Group had jointly upskilled 200,000 Indian engineers to work on AI. The two announced that they plan to scale AI infrastructure in India as it generates massive amounts of data, a prerequisite for creating intelligence. Ambani pointed out that Jio had delivered 16 exabytes of data in one year alone. That is Indians hanging out on social media, listening to audio and watching videos on streamers, and shopping on e-commerce platforms generating massive amounts of data. Huang envisions everyone in India using AI, creating momentum for an AI flywheel. Or turning India into a token factory for future Replicants, the economic benefits of which will flow to only a few.
(Dinesh Narayanan is a Delhi-based journalist and author of 'The RSS And The Making Of The Deep Nation'.)
Disclaimer: These are the personal opinions of the author
