Rethinking AI: Trust and Transformation in the Digital Age

Focus on Trust in AI Development

Artificial Intelligence (AI) is revolutionizing industries across the globe, but as former Infosys CEO Vishal Sikka points out, the next phase of progress relies heavily on trust. In a recent podcast, Sikka discussed the inherent randomness of large language models (LLMs), the technology that fuels popular chatbots, including ChatGPT.

The Probabilistic Nature of AI

“LLMs are probabilistic,” Sikka noted. “They don’t always know what they don’t know.” This statement underscores a critical flaw in these systems: they can appear confident yet provide inaccurate or misleading information. As enterprises begin to deploy these technologies, it is essential to implement safeguards to prevent misuse.

The Need for Control Over AI

“You can’t just let them run wild,” Sikka warned. “You have to put a leash around AI.” Understanding the uncertainties inherent in AI is crucial; organizations must be aware of when these models fabricate information, and they should know when to defer to human judgment.

Designing Trustworthy AI Systems

Creating “trustworthy and reliable” systems involves developing AI models that can recognize their limitations and express uncertainty when necessary. “In our systems, we design models that can express uncertainty,” Sikka explained.

A Safety Boundary for AI

Elaborating on the concept of a “leash,” Sikka described it as a “safety boundary.” He emphasized that in their AI architecture, models are designed to acknowledge their own uncertainty. “When the model doesn’t know, it says so,” he remarked. This is a shift from the current approach, where many LLMs simply complete the next token without regard for truthfulness.

Infosys’s Early Support for OpenAI

In a rare disclosure, Sikka revealed that Infosys was among the first supporters of OpenAI in 2015, long before AI became a buzzword in tech circles. “When Sam [Altman] and Elon [Musk] were starting OpenAI, Infosys was one of the first supporters,” he said, highlighting their vision for creating safe and beneficial artificial general intelligence.

A Commitment to Responsible AI

During its early collaboration with OpenAI, Infosys was focused primarily on research rather than commercial applications. “It was about long-term impact, not commercial value,” Sikka noted. He stressed the ongoing relevance of these principles as AI technologies permeate various industries.

Anticipating the Future of Work

Sikka, who has led transformative technology initiatives at both SAP and Infosys, foresees a dramatic shift in workplace dynamics as intelligent systems become more prevalent. “It’s inevitable. The 20 people doing something today will become five people plus AI tomorrow,” he asserted.

An Evolution, Not a Doomsday

This transformation, according to Sikka, is not a doomsday prediction but an inevitable evolution akin to past technological advancements like automation and cloud computing. “AI will make this shift faster and deeper because it touches cognition, not just execution,” he explained.

Preparing for Challenges Ahead

However, Sikka cautioned that both companies and nations must adapt to this shift. “Companies that don’t figure out how to reskill people and design AI-plus-human workflows will lose competitiveness,” he said.

Uplifting, Not Displacing

For countries like India, the challenge lies in ensuring that this transition benefits people rather than displacing them. “The services model has to change. It can’t just be ‘people doing tasks’ anymore — it has to become ‘AI plus people delivering outcomes,’” he concluded.

Conclusion: The Path Forward

As AI continues to evolve, building trust and ensuring responsible implementation will be key to leveraging its full potential. By recognizing the limitations of AI and creating frameworks that prioritize human oversight, companies can navigate the new digital landscape effectively.