Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Indian technology grandee Nandan Nilekani expects companies around the world will increasingly build their own smaller-scale artificial intelligence models to streamline operations and boost productivity, dampening hope of a substantial enterprise payday for more powerful generative products.
The chair of IT services major Infosys told the Financial Times he was “not so sure” companies would want to shoulder the high costs and the potential “black box” of data and copyright liabilities associated with large language models behind popular applications, such as OpenAI’s ChatGPT.
“When you look at the large firms they’re all saying: ‘How do we take charge of our AI destiny?’” Nilekani said in an interview in Bengaluru, the Silicon Valley of India. “Small language models trained on very specific data are actually quite effective . . . everybody will build models, but I think they don’t have to build these gigantic ones.”
Nilekani said enterprise customers would call in LLMs for specific applications and the technology would fuel competition between the giants, such as Google and Apple, which offer online products to consumers.
But his prediction raises questions about the revenue model for start-ups that have invested in LLMs. Facing hesitation from businesses, the likes of Apple, Google, Meta and Microsoft have all recently released AI models with fewer “parameters” — the number of variables used to train their systems and shape output.
Infosys, the nearly $19bn annual revenue software and consulting business that Nilekani co-founded in 1981, is positioning itself to be an AI provider to the clients it serves across more than 50 countries, helping them to organise their data and train their own models.
Nilekani highlighted Infosys’ company launch last month of two small language models in partnership with AI chip group Nvidia. They were trained on Infosys data and integrated in products such as its digital banking software Finacle.
“We are actually offering a service to our clients to build a model . . . there’s a lot of interest in that because we are demystifying this whole model-building stuff,” he said. “The whole technology has become so easy that you can build models in a matter of months.”
Some analysts have raised concerns that India’s outsourcing industry will suffer as more companies adopt technologically sophisticated AI models, with value flowing to software providers and the so-called hyperscalers of cloud services. If companies build their own models, however, then the technology will be an opportunity and not a threat for companies such as Infosys.
Nilekani, a fatherly figure of Indian tech who has been central to efforts shaping the government’s digital policies and landmark Aadhaar biometric identity system, said AI would replace some functions in his country’s tech industry — but new roles marshalling its power would crop up.
The Indian services sector has struggled with subdued growth and has slowed down hiring in recent years as customers, including Wall Street banks, cut back on IT spending.
Nilekani did not expect a large pick-up in headcount across India’s tech services industry, which employs more than 5mn people. Partly because of advances in AI, but also due to a weak global economy, they “may not grow as they had grown in the past”, he said.
The industry is one of India’s few large sources of white-collar work in a country of 1.4bn people that is struggling to create meaningful jobs to soak up its young and vast labour force.
At the same time, the Infosys chair said Donald Trump’s re-election might spur the industry, even though Indian IT companies were affected by his last crackdown on H-1B visas, which companies in the US use to bring in skilled foreign workers for a limited number of years.
“The bull case is that this will unleash market deregulation, businesses will grow, people will do more acquisitions and therefore there will be more activity and therefore IT guys will benefit from it,” he said. “My own sense is that they will probably support legal migration, which makes sense. I mean really high-quality talent, why would you not do it?”
, Infosys chair bets companies will develop their own AI models