‘Of Us’ and ‘For Us’: Humanity under the AI microscope

That AI gives us faster and sometimes better answers is one of its superpowers and the source of considerable angst. That may be warranted. But in the latest episode of our “Learning Machines” podcast, where we explore how our interactions with AI can help us learn, love, and lead better, we focused on the other side of the ledger: the questions AI asks of us and raises for us.

Our recent discussion with Dr. Suhas Mahesh, a physicist and scientist at Schmidt Sciences working on AI and scientific discovery, ranged widely from scientific discovery and the hidden processes behind human judgment to the future of work, the limits of optimization, the meaning of intelligence and empathy, and even love.

The omnipresence of AI in personal, communal, and professional spheres is increasingly asking questions “of us”: what do we mean by intelligence, rationality, empathy, care, thoughtfulness, and judgment? AI may therefore be as useful a lens for ourselves by forcing in greater precision on what we mean by distinctly human capacities as it is a tool for solving problems.

“The thing that they tell you AI is going to do is that it is going to make me more productive,” Dr. Mahesh noted. “But that is truly not what the excitement around AI should be, because this is a new kind of intelligence. The question you should be asking yourself is: what is AI going to help us do? That is, what are the new sets of questions that AI can ask for us?”

Dr. Mahesh turns to the microscope to illustrate his point that transformative technologies do not simply solve existing problems faster; they reveal phenomena we could not previously perceive and therefore create new categories of questions.

The microscope was not invented to discover cells, microorganisms, or disease. The earliest microscopes emerged around 1600 from advances in lens-making, particularly the craft of spectacle makers. Their precise origins are disputed and some early compound microscopes are associated with Dutch lens makers.

The original problem was essentially optical: how can lenses let us see things too small for the naked eye? Its development preceded the scientific disciplines that eventually depended on it. Its inventors could hardly have set out to create fields that did not exist yet: microbiology, histology, pathology, etc.

By 1665, English scientist and inventor Robert Hooke described observing thin cork through a microscope and identifying tiny compartments he called “cells.” Like many other imaging devices that followed, the instrument revealed that familiar objects contained structures humans had not known were there.

Within a decade, the Dutch tradesman and self-taught microscopist Antonie van Leeuwenhoek, using powerful single-lens microscopes, observed microorganisms — including bacteria and protozoa — as well as sperm cells and blood cells. An entirely living world previously invisible to humans became observable. By making previously invisible things observable, the microscope enabled us to discover entire categories of questions and worlds.

By the late 1700s, the microscope had become sufficiently powerful and user-friendly that the Encyclopedia Britannica gushed: “among all the inventions that ever appeared in the world, none, perhaps, can be found so constantly capable of entertaining, improving, and satisfying the mind of man.” This definition could apply to how we view AI today.

READ: Sreedhar Potarazu and Carin Isabel Knoop | Humans vs. AI: Thinking in epochs (Part 1) (December 10, 2025)

However, Dr. Mahesh believes that our vision is still too narrow. Today, with AI, we might tend to ask: Can AI help cure cancer? Can it discover better material? Can it make a scientist or author 10 times more productive? But he suggested that we may still be thinking too narrowly. “AI is essentially a different kind of intelligence that does things that the human mind cannot. It can analyze datasets at scale, and there are also lots of interesting questions hiding in the large datasets that we have collected.”

According to a recent paper in which Dr. Mahesh was involved, the promise is enormous: “When it comes to retrieving information,” the scientist noted. “There have been incredible advances. They have made interdisciplinary research as well as understanding what the state of current knowledge is extremely easy.”

In science, AI can improve and accelerate research and output; Dr. Mahesh also notes that it can also make the process of science itself more visible. “Where the problems really lie is that we have not in the past systematically built science to be verifiable. It has always been a system of trust. The publication record itself tells you very little about the quality of science.”

But AI brings the possibility of building a system where every artifact science produces can be linked and verified. “Every step, we maintain a record of the scientific process itself,” Dr. Mahesh notes. “The end data, the outputs that come out, are actually available in plenty because we have published them. What we do not have is the process of doing science, what happens internally, what the scientists actually do in the lab. That has been a black box because we never report on that.”

However, Dr. Mahesh argues that these may be the equivalent of asking what existing problem a microscope could solve. The larger transformation came when the microscope revealed a previously inaccessible world and science began asking questions prompted by what it could suddenly see. To return to that analogy, AI helps us produce more science but can also make visible parts of the scientific process that previously remained hidden, partly because we lacked a good way to capture them.

This matters because scientific publications generally report methods and results, but necessarily capture only part of the sequence of judgments, choices, failed attempts, and adjustments needed in producing them.

By making the process of science more visible, AI can also help us consider a more uncomfortable question: how much do we understand, and how visible is the process by which we ourselves reach conclusions, often about ourselves?

As humans, we rarely pause before a decision to ask how we actually arrived there — and we rarely extend that curiosity to s, others, focusing instead on the result or what is immediately visible.

Consider a disagreement with someone of different political persuasion. Our instinct may be to challenge the conclusion. A more revealing inquiry might begin elsewhere: What were the steps in that person’s intellectual and emotional process that brought them to this position? And, equally importantly, what were ours? We see the conclusion or the position, much as we see the published scientific paper, while much of the process that produced it remains hidden.

Our judgments, however, emerge from what might be thought of as a human intellectual and emotional laboratory: accumulated knowledge and experience, certainly, but also assumptions, emotions, values, genetic signatures, intuitions, and biases that are rarely apparent even to ourselves. Research across psychology and behavioral economics has documented many ways in which judgments are shaped by processes that are not fully accessible to conscious introspection.

READ: Sreedhar Potarazu and Carin Isabel Knoop | How we learn and why we get stuck (Part 2) (December 22, 2025)

Examining how machines arrive at answers should encourage us to become more curious about how we do. Much as observing how properly prompting AI can help us become better communicators, examining how machines arrive at answers may therefore encourage us to become more curious about how we do.

When Dr. Mahesh spoke of the new questions AI might ask “for us,” one of us had already filled in the blank in their head and expected the question to be: What new questions might AI ask “of us”?

Misunderstandings, especially literal ones like this one, which can easily occur in video exchanges, can become sources of understanding.

Questions AI asks “for us” concern discovery: what might this new form of intelligence allow us to perceive, investigate, or understand that we could not before?

Questions AI asks “of us” concern self-examination: what does the existence of another form of intelligence force us to reconsider about ourselves?

For centuries, we have described people as intelligent, rational, thoughtful, empathetic, or caring without necessarily defining precisely what those words mean. The arrival of AI is forcing us to define them more precisely. If we speak of artificial intelligence, what exactly do we mean by intelligence? If a machine appears empathetic, what distinguishes empathy from its simulation? Does artificial empathy have a similar impact on humans as “natural” empathy? Many people, including business people on platforms that did not use to veer to the philosophical, are asking what constitutes judgment, rationality, or thoughtfulness and which parts of those qualities, if any, are distinctly human?

In an upcoming paper, Dr. Mahesh et al. examine how terms such as intelligence and rationality are used in the literature and whether an underlying consensus exists about what these abstractions mean. The lack of agreed definitions also complicates between human and AI; there is some difficulty in comparing humans and machines when there is little agreement of what judgment and intelligence mean.

Just like the microscope moment, then, AI may open up entirely new scientific fields as well as bring into focus assumptions about ourselves that have been there all along. And some of the most consequential may not be the questions AI can ask for us, but the questions AI’s existence asks of us if we care to listen. According to our guest, “The most exciting science is open-ended science where humans have this thing called surprise.” This could be a useful way to think about our present dance with AI.

All of us are part of this giant AI experiment on humankind in some way. We do not yet know what it will reveal about us. We should leave room for positive surprises.