“One thing to keep in mind is that just as the web brought information to the fingertips, AI is going to put knowledge at everyone’s fingertips,” said Mustafa Suleyman, at the Microsoft: Building AI Companions for India event in Bengaluru on Wednesday. The British Artificial Intelligence entrepreneur was responding to a question about how AI as a general-purpose technology enhances productivity in India or what could be its potential applications here. Suleyman explained that AI is going to put the knowledge that is synthesised, distilled, and personally tuned to the way that one wants to learn and use it. He added that this will be applied both in the workplace and at home. Further elaborating, he gave the example of Microsoft 365 Copilot that, according to him, does an “incredible job of reasoning” over work data. “It can draw on and then provide citations for any question that you would ask of it, referencing your email or referencing your calendar, looking at your Excel sheets and your docs, your company's HR records, your company's supply chain information,” Suleyman told S Krishnan, secretary, Ministry of Electronics and Information Technology, Government of India. According to the Microsoft executive, this lowered entry barrier to utilising knowledge is making a valuable contribution in the workplace. “For many knowledge workers, it is fundamentally about getting access to useful information that you can take action on. And so I think that is going to have a profound economic benefit for many of our industries,” he said. During the fireside chat, Krishnan emphasised that when it came to impact there was a need to be culturally specific and adapt according to social situations in different parts of the world. When asked how Microsoft was attempting to do that, Suleyman said that he was proud that India is one of the fastest-growing markets and that the company has one of its strongest teams worldwide based in Hyderabad. Explaining further Suleyman said, “I think there's extremely talented engineers and developers here. But also increasingly, we're involving social scientists, psychologists, therapists, scriptwriters, comedians—I mean really people who you might more often associate with. So I think that's an opportunity for us to synthesise more diverse perspectives and get a broader picture of people involved in the design operation process. Krishnan shared Prime Minister Narendra Modi’s vision of making AI accessible to all across India and said that the government was looking at ways in which it could adapt to various Indian languages, briefly mentioning Bhashini, an AI-led language translation system initiated by the government of India. When asked how Microsoft can work with this diversity in India, Suleyman said that voice is the ultimate way to make the tools accessible and widely available. “It's exactly the kind of investment in languages and translation that I think governments should be making.” While discussing his thoughts on making more government data available to help boost innovation, Suleyman said that the pre-trained models going forward would become largely commoditised and many of them will be available via API and, in some cases, open source. “Data that will be required for post-training or the last stage of training to adapt a model to a specific use case is actually very small. You only need a few hundred thousand examples of the good behaviour that you're trying to get your model to imitate or learn from at the post-training stage. I expect to see many thousands of different types of agents with different types of expertise, both linguistically, but also knowledge and grounding over different types of databases and knowledge corpuses,” he said, underscoring the significance of the need for more data to train AI models. Reflecting on his journey, Suleyman said that when he looked back in hindsight they got something right. The Microsoft executive said that he was very lucky to choose to start the company at the right time. “In 2010 we just caught the beginning of the deep learning revolution. Had we been a couple of years late or a couple of years early, we would have missed that first wave of deep learning starting to become very practical and useful,” he said, adding that timing is everything.
