Daron Acemoglu, the Nobel Prize-winning economist, has disputed the automatic link between artificial intelligence’s (AI) disruptive power and positive outcomes, arguing that its transformative nature does not inherently guarantee productivity gains or serve as a force for general good.
“It could be transformative while its productivity gains disappoint,” he said recently, while speaking at a webinar hosted by the Indian Council for Research on International Economic Relations (ICRIER) on the role of AI in sustainable and inclusive job creation, and future policy challenges.
The Turkey-born Armenian-American argues that as the current trajectory of AI development is heavily focussed on automation, largely driven by the ease of creating lucrative business models for big tech companies, the world faces some critical, and in some sense intersecting, choices.
The first is to determine whether AI should be developed as a ‘substitute’ or a ‘complement’ to human talent; second is whether to pursue artificial general intelligence (AGI) or AI as a tool; and finally resolving the conflict between data centralisation and individual data ownership. He adds that unless we acknowledge the ‘nature and implications’ of these choices, the world is unlikely to mitigate the negative consequences of AI or fully realise its promised benefits.
Acemoglu believes that even though AI has more ‘pro-worker’ capabilities than any other previous digital technology, it is not being fully utilised as current investment priorities are not aligned with utilising these capabilities.
He argues that automation looks easy but integrating it is a serious challenge that very few companies have successfully leveraged. While automation is not all that bad, but it needs to be ‘counterbalanced’ by promoting complementarity with human labour.
“Automation creates inequality, automation creates dislocation. Automation creates imbalance between capital and labour, and ultimately will start creating unemployment,” he said.
Other issues that he touched upon were the reinforcing of an automation bias through the architecture of large language models (LLMs), which mimic human intelligence and drive an automation mindset, and information centralisation, which prevents data ownership necessary for creating high-quality, worker-specific tools.
India’s Chief Economic Advisor V Anantha Nageswaran echoed Acemoglu’s sentiment when he said that automation-centric AI path, coupled with data centralisation, creates significant challenges “which is going to replace labour”. He highlighted how developing countries like India are likely to face two significant headwinds—one, how AI automation could erode established growth models; the other, the willingness of capital-intensive sectors to hire could simultaneously decline.
As these factors are likely to impact job creation, India also faces a significant technological disadvantage because the country lacks the necessary computing or GPU capabilities possessed by US or China to develop its own ‘Indian data-centric LLM’.
Nageswaran believes that guiding AI requires political and policy choices, and not just technological ones, to avoid economically counterproductive outcomes.
First Published: Nov 05, 2025, 09:52
Subscribe Now