Regulations and guardrails are essential for the use of AI, particularly as tools evolve, and so are data sets, according to Indian tech company Micro land’s Karthikeyan Krishnan There aren’t many tech companies that have seen, or can chart the evolution of technology, as well as Indian digital infrastructure company Microband can. Headquartered in Bengaluru and with a now 34-year-old legacy with extensive investments in many countries including the UK, Microband made a pivot in the services and solutions that they provide for enterprises and businesses post-Covid-19. AI has simply accelerated in what is likely just the latest chapter in a journey of consistent transformation. “We started by focusing on networking. If I must use that parlance, we started with coaxial cables, which are used for the cable TV connections even now. It’s fascinating,” Karthikeyan Krishnan, who is the senior vice president and Geo Leader – Europe, Middle East and Africa, told HT. Along the way were UTP cables, or unshielded twisted pair cables, used in telephone wiring and local area networks (LANs). “The next big shift happened when Internet Explorer and Netscape came. We were a Platinum Partner for Netscape at the time, and then became a Microsoft partner. In fact, we hosted Bill Gates’ first visit in India,” he said, talking about an extraordinary journey. Bill Gates’ first visit to India was back in 1997. In 2010, which is when the big cloud boom happened, Microband continued to evolve, with focus on specifically enabling digital platforms for their enterprise customers. Edited excerpts: IN the last couple of years, what is happening is incredible. It is not even like the internet and mobile. It’s much beyond that. What has made that change is generative AI. Mind you, AI has been around for many years. In fact, some of our platforms that we use to manage our customers’ digital platform are smart cloud operations. They all run on AI and machine learning, which is especially useful for predictive analytics. But generative AI is taking it to the next level, and we are making significant investments on that. Customers are also asking what more they can do leveraging this in our business. But that is one aspect. The second parallel aspect is what they can do to improve operational efficiency and productivity using those platforms with our tools.

Q. Is AI Mature Enough To Handle complex Processes Right.

now in workplaces? Or, would wait-and-watch be the best policy before eventual deployment? There are two parts to this. The first is in terms of why context setting is required. If you look at Chat GPT or any other chatbot, it is a foundational model, which means it starts with a blank sheet of paper. You can draw whatever you want. So, it doesn’t tell you what you want it to. Since it is a very broad spectrum, if you ask anything, it’ll be generic. It is not going to give you an answer as a journalist, an IT specialist, or as a doctor, because each of them could answer the same question in a different way. So you need to contextualise this. The last quantum data gets narrowed down to the next level. Then it narrows down further to look like objective knowledge base, and then tries to respond. So that’s broadly when you do a very broad question. That’s when hallucination happens. It doesn’t know where to get data and then tries to answer because it doesn’t say no. It doesn’t give you a ‘no’ for an answer, right? That’s the beauty of it and it is up to us on how much and what we need to do. So, context-setting is very important, and it need not be complex. There is a third component. For example, Chat GPTs are the public data sets. It’s not necessary that we need to use public data sets all the time. What happens if I wanted to use my private data sets and train the bot accordingly? Take the example of any specific company. So much knowledge that they may have been built over a period of time, which is not available probably on the internet, and they wanted to use that to train bots and generate out of that. They can use an LLM (Large Language Model) and connect it to their website. So, it is trained to now look at only my data. It has no public references. It gives the perspective as a consolidated perspective. Q. EU set the tone a few years ago with regulation, specifically the GDPR, which many other countries then followed through with over the next few years. Do you believe the UK is setting the tone for AI regulation, something more countries would now follow? It is very important to have the guardrails because the use cases are so diverse now. If you wanted to use Open AI’s chatbot and AI services through Microsoft Azure, they don’t give you access to it straight away. They will ask you for a use case. Europe has come out with early drafts. The next level, beyond generative AI, is auto-generating AI. You can feed in one input and define the output. It takes care of every step in between. I think Europe will come with the first one, followed by the US, and then there will be versions with India and other parts of the world soon.