India, April 25 -- Pioneering Yale astrophysicist Priyamvada Natarajan proved that black holes can be formed by unstable gas. Now she's looking at the invisible universe using new technologies. In popular imagination and in science fiction, black holes are places of intense gravity, kind of like cosmic vacuum cleaners that suck in everything around us. They are formed when stars explode, become supernovae and leave behind deep punctures in the fabric of the universe. They're hard to understand, even for Priyamvada Natarajan, professor of astronomy and physics at Yale University, who has studied the phenomenon for decades. In 2006, Natarajan proposed a radically different idea on how black holes in the early universe are formed. Not by a star exploding, but through a direct collapse of gas. She theorized that gas in the early universe became unstable and funnelled into the centre very fast, kind of like pulling a plug in a bathtub. This bathtub action created massive black holes, as big as 10,000 times the mass of the sun, in a jiffy. In late 2023, two space telescopes proved all her predictions and her theory right. Today, we know about direct collapse black holes and UHZ1, an ancient galaxy which has this, thanks to her. We catch up with her at the Indiaspora 2026 conference in Bengaluru to understand her fascination with invisible space, competition for space telescopes and how artificial intelligence is reshaping her work as an astrophysicist. Edited excerpts: We needed to think about another way how black holes form than stars exploding because we were seeing massive black holes a million times the size of the sun in early universe and had no explanation for it. My calculations showed that in the early universe, massive black holes could form through a bathtub action where gas becomes unstable and stars start funnelling into the centre very fast. Thanks to sophisticated computers, we were able to build a concrete prediction in 2017 which theorised exactly what the observations of the real universe would look like if this kind of a black hole existed. When gas heats up and glows around an event horizon of a black hole, it can be seen in the X-ray or optical or ultraviolet. Seeing X-ray emissions from the centre of a galaxy is often a pretty good sign that you have a black hole there. However, something that was emitted in the optical at the very early time of the universe, will be seen by us today in the mid-infrared range as the wavelength of light would have stretched out because our universe is expanding. James Webb has cameras that detect infrared radiation. Our prediction was that for these black holes to have formed in the early universe, not only should you see them in X-ray, but also in infrared radiation. So the same black hole needed to be seen by Chandra X and by James Webb simultaneously to prove our theory that black holes can be formed by this method. We predicted six different unique properties, the spectrum, the amount of energy that comes out, even a shape. Only if all six were satisfied by an object, you could clearly say that there's convincing evidence that this was a new way that black holes were created. As scientists, you compete with each other for data allocation of these telescopes. You submit your proposal and it's anonymously evaluated by peers. The most interesting idea wins. For our prediction, we needed James Webb to look in the same direction for several hours, but with Chandra, we needed several days of observation as it is capturing things that are faint and far away. When the black hole we predicted about was found, Chandra had been recording the same patch of sky for 24 days. Light is our cosmic messenger, but dark matter or dark energy doesn't interact with light at all. Their visual absence means they're physically present. I find them intellectually very seductive because you have to look at these objects indirectly to infer things about them, to understand their nature. It's like detective work. Dark matter, which I'm studying currently, does deflect light. Light rays travel on the fabric of our universe, going up and down on matter and dark matter which creates kinks or potholes in this fabric. We physicists record these imprints of deflected light to understand dark matter. In case of black holes, only light beyond its event horizon is absorbed. Anywhere outside the event horizon, you can receive signals and radiation. For example, when gas gets pulled into a black hole, it gets heated up on its way before it reaches the event horizon. It glows. That's how we see, measure and map to study black holes. I'm exploring how dark matter interacts with black holes as they both seem to be present in centres of galaxies. Another direction I'm exploring is what construes dark energy, which is so far not understood but is responsible for the rapid expansion of our universe. Both of these are open frontier questions. Large LLM models have digested everything that's published in our area, right? So they come armed with expertise of peer-reviewed published material. Can an AI agent interrogate a new idea much more finely than a bunch of scientists? Can it identify our blind spots? Can it in future become a peer reviewer? Can it actually work in a lab, in the messiness of science first hand? Can an AI actually write a scientific paper from start to finish, generating the idea, developing it, writing it in a coherent way? We're definitely in the time of exciting intellectual upheavals. We haven't fully grasped how AI is going to change all dimensions of our reality....