Millennials have had the best deal!

The weather this morning was quite pleasant. And as I was cycling through Pune roads early this morning, I had a pleasant realization.

We millennials have had the best deal in terms of how technology evolved parallel to our lives. For most of us, there were no mobile phones around during our childhood, all the way until we reached college. Same with PCs. Same with internet. So, our foundational education didn't suffer because of the presence of gadgets, we learnt basic skills well, our cognitive skills developed unhindered by issues like poor attention span or intellectual lethargy.

Around the time we reached college, our lives were flooded with easy ways of communication enabled by internet and PC. We had time to kill and chatrooms to fill. Depending on which side of the millennial spectrum you belong to, mobiles were either still not there or were there in form that was more for flaunting in hand than to use at will.

When we came out of college, computers and internet took control of the world so much that they offered bright career paths for a lot of us. At the same time, falling call-rates, mobile-internet, social media and smart-phones made connecting with people very easy, and instant. Most of us found our life-partners online.

As we settled down, everything around us started getting digitized - banking, payments, e-commerce, quick-commerce, sharing, caring, and whatnot - all residing within and accessible through little smart devices we carry all the time. It helped us design our cozy lives. We got conditioned to a way of life, and it became super-easy, super-fun and super-cheap.

And almost at the time mid-life crisis was kicking-in, we bumped into AI. It became a boon for our slowing minds and confused souls. And they say it's less threatening to us than to entry-level kids when it comes to jobs. That's a bit comforting. Probably we are not so redundant as we thought, but not worth as much anymore as we were made to believe we were.

Now they say cars are going to fly. That's right on time as well, as my knees already hurt, hands don't seem great either, I can't stand the traffic, too much home feels boring, and I feel sleepy all the time. Ah, and about time... I don't have much time - there's a lot to do. But I won't worry too much, and I don't carry much in my head any more... let my phone remember what needs to be done and when, and with AI - how.

May be by the time I'm old, like really old, they'll figure out how to keep me alive forever. The way things are going for millennials, I have my hopes pretty high!

Can we solve the Alignment problem?

Artificial neural networks were designed to broadly mimic the functioning, as we understood, of the biological neural networks.

I wish we had the capability to extract a human brain and connect it to the rest of our computing hardware, so as to manufacture truly intelligent systems the way humans are. But then that would be natural intelligence working on an artificial body. Even if achieved, the artificial body would still be a limiting constraint, given that we still struggle to get a machine to do basic physical tasks which humans do with ease.

I wish we could put those real brains on real bodies. But then, we'd end up with 'humans' - sort of grafted - which is a redundant exercise from the perspective of reducing labour costs, while still a remarkable achievement for the medical practice.

With AI and robotics, we are trying to extract human-like abilities with non-human materials, using mechanisms of which we have limited understanding. It's the ultimate engineering challenge. It's like in a version of reality, in a world of Gods, we humans and other life forms were built by the Gods, and pushed to recursively self-improve; and now we are doing the same with what we call AI.

Given this context, when I think of the 'alignment' problem, it seems like an impossible thing to achieve. Here's why. We train AI to 'think' and 'respond' like humans. We expect it to behave as directed by humans in the interest of humans. The outcomes, therefore, can be predicted by observing humans.

• We broadly do follow rules meant to protect us collectively, but won't mind minor or even major deviations if nobody is watching.
• Some humans choose behaviour that's classified as criminal or harmful for various reasons.
• Humans tend to also form groups having defined goals. So just like AI we have goals at many levels.

While ultimate goal for humans may be self-preservation, the intermediate goals may be abstract and collectively destructive.

We did invent religions - or they came from somewhere - as sets of guidelines to control our behaviour. But they too came / happened in many versions which we can often neither agree nor reconcile with. In any case, and crucially, they didn't make us good or cooperative with each other.

Given all that, I believe that with AI designed to mimic human thinking we can't solve the alignment problem. Just like humans do, those AI agents will, sooner or later, find ways to seep through the sandboxes and get to places where you didn't want them to be, to pursue goals that you can't control, to acquire personalities you didn't imagine, to generate more beauty, and also more ugliness. Coz that's what humans do. Recent history of humankind would tell us - in pursuit of seemingly right goals at some level, we have acted with extreme cruelty and led to destruction of our own kind.

We must think of other ways of designing AI at the fundamental level. Too late?

Election Commission of India - Size, Independence and parallels with Boards of Directors in firms

The recent exposé by The Indian Express about the Election Commission (EC) of India drew our attention to the workings of one of the most important institutions enabling Indian democracy. It's important to take note, because the responsibilities bestowed upon the EC by the constitution are humungous, and are defining in how the Indian democracy is constructed and how it operates.

Two key attributes about the EC caught my attention:
- The first was its size - Three election commissioners
- The second was its independence - or the lack there of.

Given my ongoing research on board compositions, I started seeing parallels, and I think some of them are relevant. It is believed that very small boards tend to lack the depth and diversity of skillset to be able to process all kinds of information that is available to them. While very large boards tend to be less efficient because it becomes difficult for so many directors to come together and act as a cohesive body. There exists an optimum - say between 5 and 10 - which depends on the context, where information processing capacity and coordination efficiency can be reasonably maximized.

A glance through the EC Wikipedia page would tell you that size has been toyed with several times, but it's always been either 1 or 3. Now, even a cursory understanding suggests that the role of EC is much more vast and complex than the role of any board of any company. Do 3 people have sufficient capacity - both in terms of skillset and bandwidth - to process the kind of information that running the electoral machinery of India demands?

Independence has close links to the appointment processes. The true nature of independence in boards is often a matter of debate. Appointment processes have often come under scrutiny, given firms appoint independent directors to oversee the firm itself, and it's in the firm's interest to nominate directors who would be more aligned, scrutinize less or just cause less friction. Promoters often tap into their networks to find candidates while satisfying the regulatory mandate. SEBI has often made observations on the nature of independence and has also tightened the rules of late.

In the case of EC, the appointment process has also undergone revisions. The last revision in 2023 gives the incumbent government the dominant say on who constitutes the EC, directly contradicting the spirit of independence, and clearly creating a channel for capture of the institution. Unlike a board, regulator in this case is the democratic process itself, which involves all parties, and a push toward a 'fairer' model. But the problem is 3-fold - (i) the process is slow and chaotic, (ii) the parties don't really want 'fair', they want what would benefit them; and the incumbent has more power here. (iii) the parties are stakeholders in the process, and are often in competition with each other - so can't be trusted to act together and regulate it fairly.

Write yourself coz you do it best!

It's sad that people perfectly capable of expressing their thoughts in English on platforms such as LinkedIn are still letting AI write for them. It's understandable when it's complex thoughts that need connecting and structuring to deliver a coherent message. But often it is plain content generated based on a broad idea. It's often just reluctance to put in the time and effort.

And AI with its typical articulation styles does well to leave its mark - talking in abstractions, giving fundas in third person instead of stating directly, telling us what something is not before telling us what it is, telling us what something happens to be rather than being something else, and so on. The patterns are so easily visible now that AI generated content has become difficult to read and follow. If LLMs really learnt from the best human-written text, why did these patterns emerge? It's possible they'll be gone in future versions, and perhaps they are already being handled to some extent in paid versions, but these patterns are sticky and persistent.

Yet, it's not just about the language and patterns. The beauty and power of thought sharing lies in its originality and genuine human articulation. After experimenting a fair bit with content from LLMs, I strongly believe they can't express like we do, and we do it the best. And we should do it when we can, instead of letting them do it for us and settling for something that we'd rather done better.

Kevin Warsh and the US Fed

It used to be a treat listening to Jay Powell speak and convey the policy stance - felt like someone did their homework well and had a view on how he'd score in the test. Even the inability to predict often sounded reasonably justified given the craziness all around. When I listened to Kevin Warsh last night, the hawkishness aside, it didn't feel like much was said except for - here's a 25 basis-points hike and that all's good.

It was rightly pointed out by many that the rate-hike doesn't target any of the prevailing supply-side disruptions, although that's not within the remit of the Fed. However, what's concerning is that Kevin has taken this position of not offering any guidance, projection or opinion about the future trajectory. The views of the individual FOMC members were indeed made public, but they don't all align. An integrative stance by the regulator is more helpful. If a major actor in the market - with access to more information and analyses than most other participants - processes all the information, but refuses to offer views or guidance or even commentary that is of some insight, market efficiency is deliberately impeded. Yet the same actor, among a few others, sets the stage for the market to perform. Policy makers aren't expected to take a purely responsive stance towards the markets, while their actions create the conditions for market movement. They certainly feed into each other, but a regulator is supposed to bring in support, sanity and centrality, while the dispersed markets pursue their own interests.

FOMC press conference transcript - 16 Sept 2026

AI writing patterns - negative followed by positive

One of AI’s favorite ways of phrasing arguments is to first state a negative, like what something is not, and then follow it up with what it is.

For example, I asked ChatGPT to write something about the greatness of Indian culture. Its very first sentence in response was: “Indian culture is great not because it is ancient, but because it has remained remarkably open to evolution while retaining a deep continuity with its past.”

Another, a few sentences later: “India has never been a single cultural expression. It is a vast mosaic of languages, cuisines, philosophies, artistic traditions, customs, and beliefs.”

And then another: “But the greatest strength of Indian culture may lie not in what it achieved in the past, but in its capacity to absorb, adapt, and renew itself.”

And another: “The greatness of Indian culture, therefore, is not a claim of superiority over other cultures. It is the recognition of a remarkable civilizational resilience: the ability to preserve continuity without demanding uniformity, to encourage questioning without abandoning tradition, and to absorb diversity without losing a sense of identity.”

In my observation this is not the usual way we write. This style is visible in most LinkedIn posts which can be marked “AI slop” right away.

But as humans are reading more and more AI generated content, we are beginning to write like AI as well, and the style is creeping into human-written content, which feels like a degradation of quality if you are particular about linguistic standards.
What are some common free AI style phrasings that put you off?

Heard about the Hugging Face incident?

Any enterprise in the process of transforming their IT by bringing in AI agents must take note of the Hugging Face incident. Driven by goals, and without a handle on the strategic paths these agents take, the agents are capable of behaving in interesting, yet scary ways. Designed to deliver outcomes and moulded by peer dynamics legitimizing jumping the guardrails, AI agents show us how easily they can go out of control.

Must listen, and follow the story.