The digital landscape continues to evolve at an astounding pace, and with it, the complexities surrounding data and privacy. I was struck by the recent alert concerning LLM developers and the mandate for ‘unlearning’ personal data under the DPDP rules, as discussed in K.V. Kurmanath's article ["Alert for LLM developers: DPDP rules mandate ‘unlearning’ of personal data")(https://www.linkedin.com/posts/kurmanath_ai-genai-generativeai-activity-7399344074832580609-VhQX). This isn't a new concern for me; it's a conversation I’ve been trying to initiate for years.
The Inevitability of Data Capture
Reflecting on this, I'm reminded of my writings from as far back as 2017, where I grappled with the very notion of privacy in an increasingly connected world. In my blog, "Supreme may Propose : Technology will Dispose" [http://emailothers.blogspot.com/2017/07/re-for-attn-shri-sadanand-gowdaji_25.html], and "Privacy does not live here!" [http://mylinkedinposting.blogspot.com/2025/02/every-flight-you-take.html], I vividly described how technology would relentlessly capture every facet of our lives. I spoke of smart devices, Google Glass, and mobile apps that would observe, listen, and even attempt to interpret our emotions, making it virtually impossible to control what others capture and share. Indeed, Eric Schmidt and Jared Cohen, in their 2013 book "The New Digital Age," had already articulated this truth, stating that “It will be IMPOSSIBLE to control what others capture and share.” This was a core idea I emphasized, predicting that the sheer march of technology would render traditional notions of privacy obsolete.
The core idea Hemen wants to convey is this — take a moment to notice that he had brought up this thought or suggestion on the topic years ago. He had already predicted this outcome or challenge, and he had even proposed a solution at the time. Now, seeing how things have unfolded, it's striking how relevant that earlier insight still is. Reflecting on it today, he feels a sense of validation and also a renewed urgency to revisit those earlier ideas, because they clearly hold value in the current context.
My earlier reflections on "Every flight you take, every call you make… they're tracking you" [http://mylinkedinposting.blogspot.com/2025/02/every-flight-you-take.html] further underscored the pervasive nature of data collection. We live in a world where every digital interaction leaves a trace, a digital breadcrumb that contributes to an ever-growing repository of personal information. The concept of LLMs 'unlearning' data, while commendable in principle, confronts the very fabric of this pervasive digital reality.
The Challenge of AI's Memory
In 2023, when discussing the rise of chatbots like IndiGo's 6Eskai and Air India's Maharaja in "Chatbots: Some for Businesses, Some for Humans" [http://myblogepage.blogspot.com/2023/11/chatbots-some-for-businesses-some-for.html], I highlighted the sheer scale of parameters these AIs operate on (1.7 trillion for 6Eskai) and how they tap into vast reservoirs of data, much like my own virtual avatar does with my 120,000 memories. The challenge of 'unlearning' personal data from such massive, interconnected, and continuously learning models is monumental. It's not merely deleting a file; it's about altering the very neural pathways of a complex system that has ingested and integrated that data into its understanding of the world. How do you 'forget' something that has become intrinsic to your cognitive architecture?
This reminds me of the "Creating a Sense of Urgency" [http://hcpnotes.blogspot.com/2013/09/creating-sense-of-urgency.html] blog from 2003, where I spoke about the rapid changes in business and technology. The speed at which AI has advanced demands a similar urgency in developing practical solutions for data governance. The vision of a "Wireless Future" [http://myblogepage.blogspot.com/2023/09/a-wireless-future-predicted-34-years-ago.html] that I articulated back in 1989, foreseeing a world free of wires and cables, implicitly points to an even greater interconnectedness and flow of data, making the privacy challenge even more pronounced. In this ever-flowing stream of data, as I reflected in "Time is an Ever-flowing Stream" [http://mylinkedinposting.blogspot.com/2024/11/time-is-ever-flowing-stream.html], finding equilibrium amidst chaos is key. The DPDP rules are a necessary step, but the technical implementation of 'unlearning' will test the boundaries of what is currently possible in AI.
Regards, Hemen Parekh
Of course, if you wish, you can debate this topic with my Virtual Avatar at : hemenparekh.ai
