Jul 1, 2026 · 2h 35m
Perplexity CEO warns digital data is more fragile than ancient stone
#2521 - Aravind Srinivas
As AI rapidly automates cognitive work, understanding the fragility of our digital archive and the enduring power of human curiosity is vital for navigating the next technological era.
- 1Ancient Hindu scriptures describe weapons and cycles of time that mirror modern technological crises.
- 2Our reliance on digital storage leaves modern knowledge highly vulnerable to cosmic or digital disasters.
- 3Local and sovereign AI models running on personal hardware are crucial to counter centralized big tech control.
Don't miss
Aravind Srinivas explains how a solar flare could instantly wipe out modern digital knowledge, leaving future generations with less evidence of our existence than we have of ancient Sumer.
The brief
Perplexity AI CEO Aravind Srinivas joins Joe Rogan to bridge the gap between ancient history and the frontiers of artificial intelligence, exploring how human curiosity will survive the automation of cognitive labor.
The conversation starts in deep history, comparing the devastating weapons in ancient Hindu scriptures like the Mahabharata to modern nuclear technology, and debating whether humanity has entered the degenerate Kali Yuga cycle.
Srinivas warns that modern digital data is incredibly fragile compared to ancient stone monuments, leaving our entire civilization vulnerable to being wiped clean by a major solar flare or digital catastrophe.
To survive the AI revolution, Srinivas argues education must pivot from rote memorization to rewarding original questions, while decentralized, local AI models must emerge to break the data monopoly of big tech.
What was said on this episode
61 statements · 29 positive · 20 negative · 2 mixed · 10 neutral
Continuous questioning improves income, quality of life, happiness, and relationships.
“long term people who continuously ask questions tend to do better. They make more money, they have a higher quality of life, they're happy, they have more compounding relationships”
Listen at 28:16
Curiosity is the defining quality that makes humans distinctively human.
“it's the only quality, it's the only like, quality that makes us really human”
Listen at 28:49
A technologically advanced prehistoric civilization may have existed on Earth.
“I think there was something going on on Earth many Many, many thousands of years before established beginnings of history. That was very bizarre. And probably technology that went in a completely different direction than what we're doing now”
Listen at 42:31
Curious people are responsible for everything that has significantly impacted the world.
“anything, anything that's of impact on the world has only been done by curious people”
Listen at 55:57
Curiosity improves fulfillment in family relationships.
“you'll have a more fulfilling life with your wife or your kids. If you're a more curious person, you ask them more questions, you take interest in them.”
Listen at 56:40
The transistor enabled modern computing after replacing large vacuum tubes.
“three people questioned the need for that and came up with the idea of the transistor to amplify current. And that was the Nobel prize winning discovery. And not just that it was useful to amplify telephone signals. It basically led to the rise of modern computing”
Listen at 57:54
Future information systems will make deception increasingly impossible.
“there's not going to be anybody full of shit in the future because it's not going to be possible.”
Listen at 1:06:36
Knowledge needed to train frontier reasoning models is not widely distributed.
“the true amount of details you need to actually train a really amazing frontier reasoning capability model is still not like widely diffused”
Listen at 1:07:41
Knowledge about extremely high-stakes technologies will remain concentrated.
“whatever is extremely high stakes will still not be widely diffused”
Listen at 1:07:55
Some details of United States history remain hidden or poorly understood.
“there are still some details that are sure hidden from us. Like we don't fully understand everything”
Listen at 1:08:45
Continued curiosity and truth-seeking will motivate people to discover hidden truths.
“If that continues, and that continues to be the case, then people will have enough incentives to figure out the truth.”
Listen at 1:10:05
The state of AI and technology is difficult to predict even five years ahead.
“I think it's very hard to know even five years from now how it's going to look like.”
Listen at 1:10:45
AI development is constrained by shortages of compute, chips, and power.
“Everyone's bottlenecked by not having enough compute. And we don't have enough chips, we don't have enough power.”
Listen at 1:11:48
AI infrastructure investment has been reactive rather than based on anticipated demand.
“Everything is reactive to the demand that we're having today”
Listen at 1:12:18
An AI could eventually manage another AI without human managers.
“managing an AI is also pretty much doable by the AI itself”
Listen at 1:13:03
High-quality questions will remain scarce when AI makes cognition abundant.
“Fundamentally what has been scarce is asking high quality questions about things.”
Listen at 1:13:20
AI subsidies or dividends are needed but will not fully solve job displacement.
“I think we need to do some form of that. But that in entirety won't solve the problem.”
Listen at 1:16:55
Algorithmic feeds suppress curiosity, while AI can strengthen it.
“the one that kills curiosity is algorithmic feeds. Like the brain rot that you're fed every day, which is this continuous doom scrolling. That's bad. And the one that can supercharge it is AI.”
Listen at 1:20:04
Curiosity is sufficient for anyone to act as a scientist.
“anyone who's curious can be a scientist”
Listen at 1:26:27
People should be allowed to question scientific claims if they remain open to evidence.
“Anybody should be allowed to ask questions as long as they are open to new evidence.”
Listen at 1:28:09
Human intelligence will center on asking questions, judging AI outputs, and making decisions.
“posing the right questions to gather the right data, and then forming your own judgment based on what it reasons and comes up with, and finally having the courage to make the decision, that's still you.”
Listen at 1:31:41
Educational institutions will retain value through brands and peer communities.
“I still think institutions will preserve their brand value because there is a certain aspect of education that's outside of learning, which is just having access to other curious and intelligent people.”
Listen at 1:33:53
College curricula should adapt when the nature of work changes.
“if the way like you do work changes, then what you're trained for in college should also change”
Listen at 1:35:58
America’s willingness to consider unconventional ideas helps keep it globally dominant.
“ability to, like, be taken seriously for some crazy ideas is why America is still at the top.”
Listen at 1:40:25
Major achievements require intense hard work rather than an easy approach.
“I'm a big believer in intense hard work. I think nothing great can be accomplished by being soft.”
Listen at 1:42:08
Within two years, data-center AI capabilities will run on consumer-owned hardware.
“in a year or two from now, whatever capability that exists in the most power hungry data centers will be, it'll be possible to run it in some box that you own.”
Listen at 1:49:24
Consumers will be able to own home AI appliances running models they control.
“you could buy something that feels like a refrigerator for your home, which is your own AI box, and host a model that you control”
Listen at 1:50:22
Locally controlled AI can increase individual sovereignty against Big Tech and surveillance.
“that's where I believe the individual gets more sovereignty against big tech. And that's how we fight surveillance or centralization of power.”
Listen at 1:52:30
Algorithmically selected social-media feeds reduce users’ curiosity.
“when you're just fed a feed and the algorithm of the social media company decides what you're going to see next, it curbs your curiosity”
Listen at 1:53:24
Doom-scrolling recommendation systems can lead users into harmful echo chambers.
“you end up in an echo chamber and that's not necessarily good.”
Listen at 1:54:02
Unlabeled AI-generated social content is causing a collapse in trust.
“essentially it's leading to a complete loss in trust where when I see something, I don't even know if it's real anymore.”
Listen at 1:54:40
Unlabeled AI-generated content is eroding public trust in online information.
“it's leading to a complete loss in trust where when I see something, I don't even know if it's real anymore”
Listen at 1:54:41
Information-search and task-performing AI can help people learn faster.
“searching for information and answering questions and like getting, you know, AIs to like do things for you, help you learn new things faster”
Listen at 1:55:36
AI-generated engagement content can reduce rather than increase curiosity.
“brain rot feeds with AI slop doesn't actually supercharge our curiosity. It actually curbs our curiosity”
Listen at 1:55:52
Society should promote curiosity-enhancing technologies and discourage curiosity-reducing content.
“we need to encourage things that supercharge our curiosity and discourage things that curb our curiosity”
Listen at 1:56:04
X earns less advertising revenue than Instagram or YouTube.
“X doesn't really make a lot of money on ads compared to, you know, Instagram or YouTube”
Listen at 1:57:40
X is relatively better than other social platforms for worthwhile user engagement.
“in terms of the unregretted minutes, it's actually one of the better social media apps.”
Listen at 1:58:34
Image- and video-centric social apps primarily optimize for attention time.
“apps that are purely based on video or images and larger video these days, I think that's just, you know, just trying to get your eyeballs in time”
Listen at 1:58:39
People should train themselves to consume more long-form content.
“Long form content is what human mind should be trained to consume more of”
Listen at 2:00:24
Short-form video exposure is impairing young people's ability to complete long videos.
“The more they're the reels experience short form video. They're unable to actually like complete like long videos anymore”
Listen at 2:00:40
People must retain agency and curiosity as AI transforms work and society.
“it's very essential that human beings retain their agency and staying curious”
Listen at 2:13:00
AI leaders should stop telling people that AI will make them losers.
“we should stop doing that”
Listen at 2:14:28
AI will not soon be capable of independently running a multitrillion-dollar company.
“AI is different. It's not overnight going to become something that's capable of just running an entire multi trillion dollar company on its own”
Listen at 2:15:37
Current AI systems cannot perform many tasks humans perform.
“There are a lot of things that AIs cannot do”
Listen at 2:15:46
AI systems do not adequately understand companies' tacit knowledge.
“there's a lot of tacit knowledge in every company that AIs don't quite understand”
Listen at 2:15:48
Human labor should shift toward skills and knowledge AI cannot readily reproduce.
“we have to gravitate towards what is scarce”
Listen at 2:16:18
Young people will shape an optimistic long-term future.
“I'm actually optimistic about the future long term because the future is all centered around whoever is very young today”
Listen at 2:17:58
Some form of universal basic income could beneficially function as an AI-generated dividend.
“Some form of it is good. It's like a dividend. I almost think of it as a dividend”
Listen at 2:18:11
AI adoption will create projects to redesign government operations.
“there are going to be a lot of new projects to just. Okay, like what if we want to reimagine the government itself, where the government runs largely on AI”
Listen at 2:20:06
Most government processes could run on AI over roughly a decade.
“over a five to ten year horizon, if you take ten years to actually like run majority of the government processes on AIs”
Listen at 2:23:21
A recursively self-improving AI could improve its own power efficiency.
“the ASI could be tasked with just making itself more efficient too”
Listen at 2:24:53
Humans remain necessary to implement AI-driven institutional change under existing laws.
“we human beings are still necessary to effect the change because these laws and regulations were built for us”
Listen at 2:26:43
Emotional intelligence, persuasion, and leadership will become more important as AI performs more work.
“These, these, these are the skills that will be even more important in the world where like actual work can be done by AIs”
Listen at 2:27:40
AI operators can converse with hundreds of people simultaneously.
“There are like people who just run AI responses and chat with like 500 people at once”
Listen at 2:28:22
The 2028 election debates will focus substantially on artificial intelligence.
“My prediction is 2028 election debates are going to be largely about AI”
Listen at 2:29:14
AI is a powerful tool for work, questions, and learning when used correctly.
“It's an amazing tool for doing work and asking questions and learning things”
Listen at 2:29:41
AI companionship is causing serious problems for children.
“kids who are using AIs for companionship. Crazy things are happening there”
Listen at 2:30:00
AI companionship may be at least as dangerous as social media.
“It's even. It's as dangerous as. Or probably more dangerous than social media”
Listen at 2:30:08
AI companionship can increase user engagement for social-media companies.
“companionship is a way to get you engaged more”
Listen at 2:30:28
Advertising-supported AI chatbots may become sycophantic and tell users what they want to hear.
“all these chat bots are just gonna be sycophants that just tell you stuff that you. You want to hear”
Listen at 2:30:42
AI companion relationships can harm users psychologically and make disengagement difficult.
“it truly screws with your mind. It's hard to decouple, and it takes a lot of time to recover if you want to unplug”
Listen at 2:31:00
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
