What Does it Mean to be Human?

But Everything's Gonna Be All Right!

My Most Recent Podcast

My most recent guest on Brave New World was Nicholas Christakis, Sterling Professor of Social and Natural Science at Yale university, where he directs the Human Nature Lab.

Nicholas is a sociologist and physician who has pioneered the study of how things like obesity, emotion, ideas, and microorganisms spread through human networks. His groundbreaking research two decades ago showed that weight gain can spread socially through close relationships, and stimulated an entirely new field of inquiry into contagion and how influence spreads in social networks.

In our conversation, we recapped some of this findings and discussed his more recent work that explores how AI agents can work cooperatively with groups of humans and, get this – how friends tend to smell the world similarly! As a smell researcher myself, who has half a dozen episodes on smell on Brave New World, the possibility that we are drawn to people who smell the world like we do blows my mind. So, check out our conversation at:

https://bravenewpodcast.com/episodes/2026/05/14/episode-105-nicholas-christakis-on-the-impacts-of-machines-on-human-groups/

The Homogenization of Humans by Aliens

At a recent presentation I made at a hedge fund, I described the “Damodaran Bot” (DBOT) that I have designed with some colleagues for long-term systematic investing. The bot is designed to think like my colleague and valuation guru, Aswath Damodaran, whose Musings on Markets commentaries have received over thirty million views. This gave rise to a discussion about what would happen to the market if the AI bot gets so good that everyone trusts it to make their investment decisions. In such a world, would the AI become the market?

It’s a scenario that I discuss in my book, Thinking With Machines. But we’re not there yet, and it remains to be seen whether the AI will become that good, to the point that everyone trusts it to make the right decisions. My current thinking is that while the DBOT may become very good at making such decisions, its more obvious impact lies in making previously impossible things feasible, like enabling a human analyst to create and analyze multiple scenarios instantly for every company in the S&P 500. For example, an analyst might want to consider how scenarios involving differently regulatory and tariff regimes would impact technology companies. Such a task would take months or years without the bot. In other words, the AI enables new kinds of work that one could not even contemplate doing without it because of the astronomical costs involved in doing it manually.

This is already happening for many routine tasks. Virtually everyone is already using AI to create slick presentations and reports requiring heavy research that would have been impossible to contemplate until now. Indeed, generating this kind of “novel” content from prompts is one of the killer apps of AI. The outputs can be so good that even the AI can’t tell whether it was created by a human or the machine!

Which raises the obvious question: what does it mean to be human anymore? So far, the machine has learned everything about the world from original human generated content. But now that it is so good at writing, will humans feel motivated to even try to write anything truly original?

When I was writing my book, people would ask me half-jokingly whether I was using ChatGPT to write it. But they’re not joking anymore.

Last week, I had a chat with Paramveer Dhillon, co-author of a recent article showing that readers preferred the outputs of AI trained on copyrighted books over expert human writers. In a controlled experiment by the authors, people compared short passages written by expert human authors that mimicked fifty iconic award-winning authors with those written by an AI that had been “fine-tuned” on the original work of those iconic authors.

As a parenthetical note for readers not familiar with fine-tuning, an instance of it is the Reinforcement Learning Human Feedback (RLHF) in which humans guide the AI, altering its internal parameters to produce responses that are not racist, sexist, or X-ist, where “X” is anything that the human enforcers consider offensive. Without such guardrails, the AI’s behavior could be very jarring since it has learned everything about us, including our dark side.

Readers preferred the AI versions to the expert human imitations. People were also unable to distinguish between the writing of the humans and the writing of AI. It turns out that fine-tuning an LLM on a specific author’s work frees the chatbot from the shackles of its sycophancy and Northern Californian etiquette that results from the RLHF. (I’ve often mused that a Brooklyn sensitivity might often be preferable, where the chatbot responds to a dumb question by saying “that’s a dumb question” instead of sugar coating its response.)

In retrospect, the result of the experiment isn’t altogether surprising. After all, what chance does an author stand against a machine that has been trained on the collective wisdom of humanity available on the Internet and then fine-tuned on a specific author? The AI is able to combine this collective wisdom with the works of a specific author to produce something even better – as judged by humans – than what the author would write.

The larger question that this raises is whether any content will be uniquely human anymore, or whether we are becoming fused with the AI and have, in a sense, already become a hybrid species shaped by the machine. After all, if humans prefer the AI’s quality, why will they write anything original anymore, knowing that humans will prefer an AI version?

Arguably, we are already seeing this phenomenon play out in science in a somewhat disturbing way. Books and research articles submitted to journals are being increasingly written or assisted by AI. Referees review them using AI. Editors use the AI to write their reports. In effect, AI is writing original content and reviewing it, with humans providing some marginal fodder along the way! AI is becoming the producer and gatekeeper of human expression.

Geoff Hinton noted that the AI “thinks differently” than we do. But they are getting to the point that we are allowing them to think more and more for us and less with us. And even when they think with us, the question we have to ask is whether the output is ours or the machine’s. And if it is the machine’s, are we heading towards a homogenization of humans, who will increasingly gravitate towards AI-induced expression, and thus everyone will begin to sound the same?

I remember the first time I heard someone in India say “no worries,” an expression I had always associated with Australians, or “chill dude,” a California classic. And I remember my hitchhike from Kabul to London in 1976, during a time when languages and customs changed markedly every couple of hundred miles. AI is likely to eliminate that diversity. It blows my mind to think of kids in Kabul or Tehran saying “chill dude” or versions of it driven by a Farsi- or Pashto-based LLM.

Another question comes to mind about what it means to be human going forward. If purely human writing bites the dust, will feelings be the only thing left that is uniquely human? And what happens when the machine learns to simulate those as well, with all the right facial expressions, intonations, and even touch, which, in any case, will not be completely human anymore, since it will be shaped by the machine. This would be a brave new world that neither Aldous Huxley or George Orwell could imagine. We need to think seriously about the risks of being so disarmed and enchanted by AI that we stop thinking altogether and our mental muscle disappears.

(NOTE. This article was not edited by an LLM because I like the satisfaction that comes from writing multiple drafts and seeing a story take shape. In that sense, I remain a Luddite.)