AI regulations have to be risk-based, driven by use case: IBM’s Sriram Raghavan

Sriram Raghavan, VP, IBM Research AI

Regulations on artificial intelligence (AI) have to be risk-based, driven by use case, and connected to existing industry-based regulatory frameworks, said Sriram Raghavan, Vice President, IBM Research AI.

His comments come at a time when countries, including India and the US, are working on regulatory guidelines and policies for AI amidst the ongoing surge in enterprise adoption of the technology.

Raghavan is an AI expert and an IBM veteran with nearly 20 years with the company.

Speaking to Moneycontrol on the sidelines of IBM’s flagship event Think in Mumbai, he said: “AI regulations should be risk-based, use case driven, connected to existing regulatory frameworks, and should strike the right balance between holding deployers of AI responsible and accountable without stifling open innovation.”

IBM has been an active participant with various governments on AI regulations. IBM’s global AI ethics leader Francesca Rossi has been a part of the European Union’s expert panel. Last month, in India, the company’s chairman and CEO Arvind Krishna had met Minister of State for IT and electronics Rajeev Chandrasekhar.

Krishna had said, “I am a firm believer that every country ought to have some sovereign capability on AI, including large language models, and generative AI.”

Explaining IBM’s risk-based approach, Raghavan said: “An AI engine that is going to recommend the next shoe to buy and an AI (engine) that's going to recommend whether or not you could get a loan don't have the same risk profile. The regulation has to be tied to risk, not a blanket statement around whether let's regulate AI or not.”

IBM has been involved in several critical enterprise AI use cases for its clients, such as helping NASA with satellite data, working with banks on mission critical workload and sensitive information and deploying generative AI for drug discovery for a company like Moderna, to name a few.

“We also believe that there is a strong role for open innovation, which has been true for our hybrid cloud strategy and AI. We are happy to participate in the dialogues across the world. There has to be a balance between holding deployers of AI responsible, without stifling the innovation that you want to invest in,” Raghavan added.

Building watsonx

IBM is betting big on its AI and data platform watsonx, which was launched in May 2023.

IBM’s Watson, as a software, first came out nearly a decade ago when the computer system famously competed on TV quiz show Jeopardy in the US against its two all-time champions, gaining attention.

Now, in its newest avatar, watsonx will have a multiple model approach. According to Raghavan, there are three parts to this platform -- .ai and .data which were released in July, and .governance, which will be released in the last quarter of this year.

“Watsonx had three pieces to it, and .ai is going to help you do everything you need around the models, from fine tuning to inference to deploying them.  As far as .data is concerned, it is a lakehouse-based repository to manage data for using AI-governed repositories. And .governance is our end-to-end governance story,” he explained.

Interestingly, a majority of the watsonx platform and its various components were researched and developed in India.

Raghavan said that about 80 percent of the research team in India is focused on AI, specifically on areas around watsonx assistant, for both existing capabilities, conversational AI and newer capabilities.

While he did not divulge exact numbers, according to media reports, nearly one-third of IBM’s total global workforce of 2.8 lakh employees is based in India.

Sandip Patel, managing director, India and South Asia, IBM, said: “We are truly the embodiment of innovating and developing in India -- for India and for the world. So a lot of what you saw with watsonx, including development work and the design work, happened in India through the research and software labs, and we continue to drive that innovation.”

IBM’s acquisition strategy

In 2023 alone, IBM has acquired seven companies in the hybrid cloud and AI space, apart from several major investments including Hugging Face.

When asked if IBM was going for an acquisition-led AI strategy to grow, Raghavan clarified that the company is looking at both organic and inorganic approach. Watsonx was entirely built in-house, he said.

Along with building their own platforms, the company looks at external capabilities to complement their technologies which is usually fulfilled through acquisitions.

“The strategy is always organic plus inorganic, as long as they add value to our overall portfolio in the market. Then obviously, there are business considerations, amount of money spent on all of those considerations. But both in AI and hybrid cloud is an organic plus an inorganic strategy,” he said.