Marie Curie didn’t start as a groundbreaking physicist.
She was a lab assistant—working under Pierre Curie, dutifully toiling away with test tubes and formulas, learning as she went, until her work on radioactivity outpaced his and earned her two Nobel Prizes.
And for that matter, Charles Darwin didn’t arrive at the theory of evolution in isolation either—his early days were spent under Adam Sedgewick, a geologist whose mentorship seemed vast until Darwin dwarfed it with his own monumental theory.
Or consider Francis Crick, who started his career alongside Linus Pauling before co-discovering the double helix structure of DNA—the very code of life.
These early lab assistants, mere helpers at the outside, became the unsung collaborators of history, the ones who learned from, critiqued, and expanded upon their mentors.
But in the context of the 21st century, as we face problems that exceed the capacity of our human faculties—complexities so vast they can barely be fathomed by even the most brilliant minds—one can’t help but wonder if the real assistant science has always needed wasn’t human at all.
Generative AI isn’t just an assistant; it’s the superpower science never knew it could have.
As Stefan Harrer, Director of AI for Science at CSIRO, puts it, “This is not a dream.”
He explains that this is a new reality—one where AI agents that learn, reason, and collaborate with scientists, not in the narrow, task-bound way that we’ve come to expect, but in a fundamentally different, more expansive role.
In Harrer’s view these agents won’t just automate tasks—they’ll redefine the scale at which science operates. They don’t need specialized training for every task, nor do they struggle under the weight of complexity. They can take in the full scope of data, synthesize it, and apply scientific methods across every layer of biological understanding.
“Generative AI had gotten close to ingesting pretty much all publicly available digital information there was,” Harrer notes. And this is where AI’s potential as a partner in discovery comes into play. These agents are not simply running pre-programmed calculations—they are creating new models, generating ideas that move science forward in ways that were previously unimaginable.
We’ve all known the scientific method: observe, hypothesize, experiment, analyze. But the scale of modern science—particularly in biology—has long outstripped the methods we’ve used to understand it. Biological systems are vast and messy, with data that no single scientist could ever hope to fully process in their lifetime.
Harrer paints a picture of this enormity: “One single human cell contains about 20 million protein molecules. And there are 36 trillion cells in a human body.” The complexity of life, the sheer scope of data involved, renders the scientific method almost obsolete in the face of such scale. It’s like trying to map an entire continent while standing on a patch of grass, hoping to somehow take in the entirety of the landscape.
Ah, but AI doesn’t get overwhelmed. It sees the landscape. It doesn’t need to grasp every detail—it synthesizes it into patterns and insights.
This is where generative AI—AI agents specifically—take over. They don’t just assist—they redefine the scale at which science operates.
They don’t need specialized training for every task, nor do they struggle under the weight of complexity.
They can take in the full scope of data, synthesize it, and apply scientific methods across every layer of biological understanding.
And this is where AI’s potential as a partner in discovery comes into play. These agents are not simply running pre-programmed calculations—they are creating new models, generating ideas that move science forward in ways that were previously unimaginable.
This is just a forward-looking topic either, it’s already happening: One prominent example is the use of AI to assist in protein folding. AlphaFold, developed by DeepMind, has shown how AI can predict the 3D structure of proteins, solving a problem that has perplexed scientists for decades. By leveraging deep learning techniques, AlphaFold achieved a breakthrough that enabled biologists to better understand diseases, design drugs, and predict protein functions.
Another example comes from drug discovery, where AI agents are employed to simulate molecular interactions and suggest potential candidates for pharmaceutical development. In 2021, Insilico Medicine, using AI agents, developed a new drug candidate for fibrosis, reducing the usual timeline for drug discovery from years to months.
These advancements demonstrate how AI agents are evolving from mere assistants to full-fledged collaborators in research, empowering scientists to explore new possibilities at unprecedented speeds.
What’s remarkable here isn’t just the speed at which these AI agents can process data—though they can reduce analytical time from months to days, as Harrer suggests. What’s transformative is the scope of the work they enable.
For the first time in history, AI agents are collaborating with scientists to generate truly novel hypotheses.
They are teaching us to think in ways we’ve never been taught before, to expand the boundaries of experimentation beyond the human mind’s natural limitations.
In a recent experiment, scientists using AI agents as brainstorming partners produced more novel ideas than those working alone. “AI agents had become creative—from hypothesis generation to experimental design to analyzing outputs,” Harrer explains.
The assistant is no longer just answering questions. It’s asking them. It’s driving the research forward with the relentless forward momentum of a collaborator that never sleeps.
The implications of this are, frankly, staggering. It’s not just a matter of scientific acceleration, though the potential to halve research and drug discovery timelines is transformative enough. No—this is about unlocking new frontiers of knowledge.
As Harrer notes, we’re now capable of exploring the full depth of life’s mysteries: “For the first time, scientists can study life on its entire spectrum—from single molecules to entire organisms.” The AI agents are no longer just performing tasks—they are expanding the scientific horizon itself. They will allow biologists, for the first time, to explore biological data at a molecular scale—the very building blocks of life. And this is just the beginning.
Generative AI is not the end of science as we know it, but it’s certainly a critical inflection point. It won’t replace scientists. It will make them exponentially more capable. Human creativity, the kind that fueled groundbreaking discoveries, will remain at the heart of science.
But now, it will be paired with an intelligence that can amplify that creativity, accelerating the path to understanding and discovery.
The AI isn’t the assistant anymore—it’s the collaborator. The questions it raises, the hypotheses it generates, and the insights it provides are reshaping how science operates at the most fundamental level.
And so we stand at the threshold of a new scientific revolution. One where generative AI agents become the indispensable partners of every researcher, amplifying human potential and pushing the boundaries of what we can understand.
With AI’s help, we’re not just studying life—we’re learning how to create new frontiers for it. This is science redefined, driven by human intellect and powered by artificial ingenuity. The task now is to transition, something we’ll cover deeply in the months ahead here at Shared Everything.