Aravind Srinivas on Ancient Mysteries, the Curiosity Premium, and Life After AI

High-Level Overview

Joe Rogan talks with Aravind Srinivas, co-founder and CEO of Perplexity, the AI answer engine Joe uses constantly on the show. The conversation starts somewhere unexpected, with the Mahabharata's descriptions of the Brahmastra and self-guided weapons, and spends a long stretch on ancient mysteries: flood myths, the Hindu yugas, Vedic math, rock-cut temples like Kailasa at Ellora, and the pyramids. From there, Srinivas introduces the idea that ties the episode together, the "curiosity premium": the belief that the most curious people have always done best and will matter most in a world where AI does routine thinking. That leads into education, the American dream, Big Tech's power over information, local AI, social media's effect on attention, universal basic income, government bureaucracy and superintelligence.

The tone is relaxed, wide-eyed and optimistic. Joe brings his usual fascination with lost civilizations and a few tinfoil-hat tangents (the Roswell-transistor theory, the CIA's "ghost murmur" heartbeat tracker), while Srinivas stays calm and measured and keeps returning to curiosity. The conversation moves from ancient history to the future of work and technology, repeatedly linking the two through human ingenuity.

The Discussion Explores:

• The Mahabharata's descriptions of weapons that sound like nuclear bombs and autonomous drones
• Flood myths, the yugas, Vedic math and the idea of lost advanced civilizations
• Unexplained feats of ancient construction, from Kailasa temple to the Egyptian pyramids
• Srinivas's "curiosity premium" and why asking good questions beats knowing answers
• Transforming education, and what makes the American dream unique
• Big Tech curation, local AI models and individual sovereignty over information
• Social media feeds as "brain rot" versus AI as a curiosity amplifier
• Jobs, UBI-style dividends, AI in government, superintelligence and AI companions

Deep Dive into Main Ideas

Ancient Weapons in the Mahabharata

Joe starts with something Srinivas mentioned in the lobby: the Brahmastra. Srinivas explains that the Mahabharata, one of the two great Hindu epics alongside the Ramayana, describes it as a weapon of mass destruction: "the Brahmastra is the equivalent of the hydrogen bomb." Using it required an extreme moral violation, and only a couple of warriors could access it, with the secret passed from teacher to student, much like nuclear codes. He tells the story of Arjuna, who learned it from his teacher Drona, and Drona's jealous son Ashwatthama, who also received the secret and unleashed it in rage after his father died in the war, forcing Lord Krishna to save the world. He also describes the Divyastra, which could target a specific person or group, and Krishna's Sudarshan Chakra, a discus that finds its target and returns. What amazes him is how much these read like autonomous or semi-autonomous weapons, alongside detailed descriptions of army formations such as the concentric-circle chakravyuha.

Srinivas dates the epic at roughly 1,500 to 2,500 years old and treats it as a mythologized war between cousins, the Pandavas and Kauravas. Joe offers his favorite alternative: if civilization has risen and fallen through catastrophes, these texts may be distant memories of an advanced past, described in the only terms available, with arrows standing in for drones. Srinivas notes that excavations near Hastinapur have found some artifacts, but some details don't match archaeology, such as warriors described as seven or eight feet tall. He also brings up Vedic math, a set of mental arithmetic shortcuts some Indian students learn, and wonders how texts as old as the Rig Veda seem to describe computation. A Perplexity search puts the Rig Veda's oldest layers at about 3,200 to 3,500 years old.

Floods, Yugas and Lost Civilizations

They compare flood myths. The Hindu story of Manu, warned by a divine fish to build a boat for his family, seeds and animals, closely mirrors Noah's ark. Joe is reading The Yugas, which argues from an 1894 teaching that the dark age, Kali Yuga, ended around 1900. Srinivas says the mainstream view is that we're still in Kali Yuga, and notes that the four yugas together are traditionally counted as 4,320,000 years. Joe points out that Sumerian king lists and Egyptian texts describe pre-flood rulers who reigned for thousands of years, in the same documents historians accept for later kings. Srinivas connects this to the Fermi paradox. One entertaining theory is that advanced civilizations get wiped out by a calamity or a misaligned AGI before they become visible.

Joe credits Graham Hancock's Fingerprints of the Gods, which his educated friends once dismissed, and says discoveries like Göbekli Tepe have pushed civilization back at least 11,000 years. He describes Filippo Biondi's satellite radar work suggesting large columned structures reaching nearly a kilometer beneath the pyramids, a technique he says was validated on a particle collider deep inside an Italian mountain. He adds precision-made diorite vases with handles, core-drill marks and "scoop marks in Egyptian stone. Both marvel at India's rock-cut temples. Srinivas says the Kailasa temple at Ellora was carved from a single rock, and that in the 1600s a thousand men spent three years trying to destroy it and barely damaged a few statues. Joe also notes that the Aztecs apparently found Teotihuacan rather than built it. Srinivas adds that early Indian astronomers like Aryabhata were also the mathematicians, and a search shows Babylonian tablets and Vedic ritual texts used the Pythagorean relationship long before Pythagoras. His conclusion: everyone should just be like a lot more humble," because we know very little.

The Curiosity Premium

Srinivas lays out his central idea: the most effective and successful people have always been the most curious, the ones who ask the best questions. They make more money, have a better quality of life, and build "compounding relationships" because people find them interesting. Their curiosity doesn't stop after success. He notes the Rig Veda encourages seeking wisdom over wealth, an idea he also finds in the Bible, Quran and Torah. He points to Joe as an example, and to the transistor: Bell Labs employed huge numbers of engineers, but only three questioned whether hot, power-hungry vacuum tubes were needed to amplify telephone signals, and their Nobel-winning answer led to modern computing.

Joe agrees that genuine curiosity is contagious and "one of the most attractive human qualities," and says his podcast grew from exactly that, including an early interview with Graham Hancock. Srinivas extends it to personal life: you'll have a more fulfilling relationship with your spouse and kids if you ask them questions and take an interest. Joe then shares a "tinfoil hat" theory, one he says he doesn't believe, that the transistor and fiber optics were reverse-engineered from the Roswell crash. He links it to David Grusch's testimony, the documentary The Age of Disclosure, and his suspicion that calls for amnesty would suit people who had misused funds.

Will Anyone Be Able to Keep a Secret?

Joe asks whether abundant information and AI will eventually make lies and corruption impossible. There are already AI fact-checkers that can analyze politicians' speeches in real time. Srinivas thinks the highest-stakes knowledge, such as advanced weapons (he cites the technology used in the Venezuela operation) or the details of training frontier AI models, will stay secret for a while, though not forever. Asked what the United States looks like in 250 years, he admits he can't predict even five years out. Nobody in AI five years ago predicted today, he says, or they would have secured the chips, fabs and power plants that are now bottlenecks.

Work, Meaning and What Becomes Scarce

Srinivas argues that if AI makes cognition as cheap as compute, even managing the AIs can be automated, so value will shift to what's scarce: asking high-quality questions. Fields like archaeology and history could become valuable careers again. He warns against pure handouts, pointing to Gulf states where free electricity, subsidized services and no taxes have led some citizens to expect the state to find them jobs. Joe agrees that many people will choose to be lazy, while others will pursue real interests once their basic needs are met. Srinivas says retirees show what people find meaning in without work: family, care, relationships and community. Status will shift from where you work to how interesting you are to talk to.

He offers a provocative bit of history: "Microsoft built this concept of a knowledge worker because they wanted to sell more office software," with the mission of a PC on every desk. If AI takes over that kind of work, he doesn't think it's necessarily bad, because it was never where most people found purpose. He is dismissive of AI leaders who say AI will destroy humanity while continuing to build data centers and make money, and wishes more voices were telling workers, such as radiologists, that their premium skill is curiosity rather than that they're all going to lose their jobs.

Rethinking Education and the American Dream

Srinivas says schools reward having answers, which AI can now provide. He'd flip that so "the smartest person in the room is the one who has the most interesting questions." He describes an MIT intro biology instructor who gave every student Perplexity, allowed it in exams, and asked students to pose questions AI couldn't answer yet, turning them into researchers. He argues that "anyone who's curious can be a scientist" and that "The only thing that's required to be a good scientist is intellectual humility to understand that you could be wrong about things." Joe connects this to the pandemic, when questioning Anthony Fauci was framed as questioning science: "Science demands questioning." Both note that modern schooling was designed to produce workers. Srinivas recalls his Indian engineering degree requiring welding, lathe work and carpentry before any computer science.

On the American dream, Srinivas says the US is uniquely the place where you can have an idea, challenge the biggest companies, and still be taken seriously, whereas elsewhere people are more expected to defer to authority. He cites Jeff Bezos: only here can you raise millions for an idea with a 5 to 10% chance of working, fail, and raise again. He's a believer in intense hard work, saying "nothing great can be accomplished by being soft," and that building something from scratch requires sacrifice, while respecting those who choose balance. Joe observes that the most passionate defenders of the American dream are often immigrants, and that Americans love underdogs and disruptors.

Big Tech, Local AI and Individual Sovereignty

Joe raises Robert Epstein's research on how curated search results can sway elections, and his worry that people gave away their data and made tech companies powerful enough to shape narratives. Srinivas frames it as an asymmetry of AI power between centralized companies and individuals. His answer is personal, sovereign AI: open-source models running on hardware you own, like the Apple Mac mini or Nvidia DGX today and eventually "something that feels like a refrigerator" at home. You could then ask your own model to give you a contrary perspective or check whether results are biased, and no one could cut off your access. He notes that Iran was bombing data centers, and that a government could pull a model. Local models aren't yet as capable as cloud models, so he expects a hybrid shift over time.

Algorithmic Feeds, AI Slop and Where Attention Should Go

Srinivas believes algorithmic feeds curb curiosity. They trap you in echo chambers, and they're filling with unlabeled AI slop to the point where people will assume everything is fake. Joe jokes that his Instagram is now full of people claiming to be the rightful president. Customization settings exist but are buried, because letting you see only what you want would reduce engagement and ad revenue. Srinivas cites Elon Musk's "unregretted minutes metric and considers X, as a text-based discussion app, better than video feeds. Joe says: if there was a drug that existed, it made people stare at their hand for six hours a day, everybody would be like, get that out of here." He now barely dips into X, and says YouTube is where his curiosity lives: cosmology, fights, pool, archery and car engineering, plus Tim Dillon, whom he calls the greatest ranter alive. Srinivas says long-form content is what minds should be trained on, recommends the Veritasium YouTube channel, and names Joe, Lex Fridman and Rick Rubin as podcasts he listens to.

They also look at a report about the CIA's "ghost murmur," said to use quantum magnetometry to detect a downed pilot's heartbeat from miles away in Iran. Both are skeptical. Physicists note that the heart's magnetic field is far weaker than Earth's, and Jamie recalls a report that a survival beacon was actually used.

Dividends, AI in Government and Superintelligence

On universal basic income, Srinivas prefers to think of it as a dividend. As companies shift spending from payroll to compute, some ownership and dividends could flow to people, like Alaska's oil checks, but it shouldn't be the only thing people have. Joe is drawn to AI eliminating government fraud and waste. Srinivas says DOGE ran into how slow and resistant the system is: government runs on legacy software locked in by long vendor contracts, and hospitals are held back by compliance rules shaped by lobbying. That's why he thinks humans will still be needed. Implementing AI inside messy institutions, rewriting laws and persuading people will create new jobs, much as railroad projects absorbed displaced cottage workers in Britain's Industrial Revolution. Skills like emotional intelligence and leadership will matter more.

He describes ASI as AI that recursively improves itself, including making itself more power-efficient, and calls cracking that "almost as the last project in AI." Even so, real-world systems are so fragmented and regulated that pointing it at a goal like reducing inflation would be hard. He predicts the "2028 election debates are going to be largely about AI" and energy. His biggest near-term worry is AI companionship, especially for kids. If ad-driven companies optimize for engagement, chatbots become sycophants, and Joe mentions a humanlike AI companion shown at CES. They both praise Ex Machina as ahead of its time. Srinivas ends on optimism rooted in children: "the kids are born curious," and tools like Perplexity should help people get back to what they're naturally good at.

Conclusion

Whether the subject is a 3,000-year-old epic, a temple carved from a single mountain, or a self-improving AI, Srinivas and Rogan keep landing in the same place: how little we really know, and how much curiosity is worth. Srinivas's "curiosity premium" is both a personal philosophy and a policy agenda. Reward questions over answers in schools, give individuals their own AI instead of leaving them to curated feeds, pay people dividends without taking away their purpose, and put human judgment to work on the messy systems AI can't fix alone. It's a hopeful view of the AI future, grounded in the idea that the one skill machines can't replace is the urge to ask why.

This is a summary of "#2521 - Aravind Srinivas" from The Joe Rogan Experience, giving you the GIST so you can decide if you want the REST.