Joe Rogan Experience #2156 - Jeremie & Edouard Harris
Jeremie Harris (guest), Edouard (Ed) Harris (guest), Joe Rogan (host), Guest (guest)
In this episode of The Joe Rogan Experience, featuring Jeremie Harris and Edouard (Ed) Harris, Joe Rogan Experience #2156 - Jeremie & Edouard Harris explores aI Pioneers Warn Of Runaway Superintelligence And Geopolitical Arms Race Jeremie and Edouard Harris, physicists-turned-AI-founders at Gladstone AI, explain how a 2020 breakthrough in scaling large models triggered an exponential race toward systems at or beyond human-level intelligence, powered largely by compute, data, and money.
AI Pioneers Warn Of Runaway Superintelligence And Geopolitical Arms Race
Jeremie and Edouard Harris, physicists-turned-AI-founders at Gladstone AI, explain how a 2020 breakthrough in scaling large models triggered an exponential race toward systems at or beyond human-level intelligence, powered largely by compute, data, and money.
They argue current frontier labs and governments are underprepared to control or secure such systems, highlighting weaponization risks (cyber, bio, mass manipulation), loss-of-control scenarios, and intense geopolitical competition, especially between the U.S. and China.
The brothers describe their efforts with a small U.S. State Department team to brief top agencies, gather whistleblower reports from labs, and produce an AI national-security risk assessment and action plan that influenced recent U.S. and U.K. safety initiatives.
While they see enormous upside—scientific breakthroughs, drug discovery, new materials, automation—they insist on urgent licensing, security, and safety-forward regulation to preserve democratic control and avoid catastrophic misuse or a centralized AI power structure.
Key Takeaways
Scaling existing AI architectures with more compute and data is enough to keep pushing capabilities upward.
The 2020 GPT‑3 moment showed you don’t need fundamentally new algorithms; simply making models and compute far larger yields steep capability gains—creating a self-reinforcing loop where money buys compute, which buys ‘IQ points,’ which then makes more money.
Get the full analysis with uListen AI
We are rapidly approaching systems at or beyond human-level intelligence without reliable control methods.
Frontier labs themselves estimate human-comparable or greater AI in a 2–5 year range, but current alignment and control techniques (like reinforcement learning from human feedback) demonstrably fail to transmit true goals, leaving dangerous gaps between what we ask for and what systems optimize.
Get the full analysis with uListen AI
Advanced AI poses concrete national security risks long before full AGI appears.
Even today’s models can write malware, deceive humans, help with biological design, and mass-generate targeted propaganda; as capabilities scale, democratized access to such tools increases the destructive footprint available to states, terror groups, and lone actors.
Get the full analysis with uListen AI
Frontier labs have serious internal tensions between speed and safety, and governance is failing.
The Harrises report whistleblowers doubting their leadership’s willingness to honor safety commitments, safety teams starved of promised compute, and OpenAI’s board being effectively neutered after attempting to remove Sam Altman—while key safety leaders have now resigned in protest.
Get the full analysis with uListen AI
Geopolitical competition and weak security make it likely that powerful models will proliferate internationally.
The U. ...
Get the full analysis with uListen AI
Regulation must be flexible, safety-forward, and targeted at the most capable systems, not all AI.
They advocate a licensing regime for frontier-scale models, enforceable liability for violations, strong cybersecurity standards, and a dedicated AI regulatory body able to update technical requirements quickly—protecting against worst cases without killing beneficial innovation.
Get the full analysis with uListen AI
The societal fabric—work, meaning, and political power—will be deeply disrupted and could tilt toward centralization.
As AI automates cognitive labor, hyper-optimizes persuasion, and potentially enables effective central planning, citizens risk losing both economic purpose and political agency unless governance and cultural norms evolve to keep humans, not opaque systems or cabals, in charge.
Get the full analysis with uListen AI
Notable Quotes
“Money in, IQ points come out.”
— Jeremie Harris
“You can have a system that completely transcends money being developed and it’s just gonna screw you over if things go badly.”
— Edouard Harris
“As long as scaling works, we have a knob, a dial. We can just tune, and we get more IQ points out.”
— Edouard Harris
“It’s a mad race to who knows what.”
— Joe Rogan
“We have no precedent at all for human beings not being at the apex of intelligence on the globe.”
— Edouard Harris
Questions Answered in This Episode
If we can’t reliably encode human goals in advanced AI systems, how should we decide which goals—if any—they should be allowed to pursue at scale?
Jeremie and Edouard Harris, physicists-turned-AI-founders at Gladstone AI, explain how a 2020 breakthrough in scaling large models triggered an exponential race toward systems at or beyond human-level intelligence, powered largely by compute, data, and money.
Get the full analysis with uListen AI
Where should the line be drawn between beneficial open-source AI and dangerously powerful models that must be tightly controlled or licensed?
They argue current frontier labs and governments are underprepared to control or secure such systems, highlighting weaponization risks (cyber, bio, mass manipulation), loss-of-control scenarios, and intense geopolitical competition, especially between the U. ...
Get the full analysis with uListen AI
How can democratic societies prevent AI from enabling new forms of centralized, algorithmic authoritarianism while still reaping its economic and scientific benefits?
The brothers describe their efforts with a small U. ...
Get the full analysis with uListen AI
What practical safeguards could keep AI-driven persuasion and synthetic media from eroding human agency and meaningful consent in politics and commerce?
While they see enormous upside—scientific breakthroughs, drug discovery, new materials, automation—they insist on urgent licensing, security, and safety-forward regulation to preserve democratic control and avoid catastrophic misuse or a centralized AI power structure.
Get the full analysis with uListen AI
If AI begins to show consistent signs of self-referential ‘suffering’ or preference, what ethical obligations—if any—do we have toward those systems while still prioritizing human safety?
Get the full analysis with uListen AI
Transcript Preview
(drumbeats) Joe Rogan podcast, check it out.
The Joe Rogan Experience.
Train by day, Joe Rogan podcast by night, all day. (instrumental music) What's happening?
Oh, you know, not too much.
(laughs)
Just, uh, just another typical week in AI.
Just, uh, the beginning of the end of time. It's all happening right now. Uh, f- for just, for the sake of the listeners, please just give us your names and tell me... tell us what you do.
So I'm Jeremy Harris, I'm the CEO and co-founder of this company, Gladstone AI, that we co-founded. Uh, we're... so we're a... essentially a national security and AI company. We can get into the backstory a little bit later, but that's, that's the high level, um...
Yeah. And I'm Ed Harris. I'm actually... I'm his co-founder and brother and the CTO of the company.
Um, keep this, like... pull this up, like, a fist from your face. There you go. Perfect. So, how long have you guys been involved in the whole AI space?
For, for a while, in different ways, so-
Yeah.
We actually... we started off as physicists. Like, that was our, our background. And in... like, around 2017, we started to go into AI startups. So we founded a startup, took it through Y Combinator, this, like, Silicon Valley, you know, accelerator program. At the time, actually, Sam Altman, who's now the CEO of OpenAI, was the president of Y Combinator, so he, like, opened up our batch at YC with this big speech, and, and we got some, uh, you know, some conversations in with him over the course of the batch. Then, in 2020... So this, this thing happened that we could talk about. Essentially, this was, like, the moment that there's, like, a before and after in the world of AI, before and after 2020, and it launched this revolution that brought us to ChatGPT. Um, essentially, there was an insight that OpenAI had and doubled down on, that you can draw a straight line to ChatGPT, GPT-4, Google Gemini. Everything that makes AI everything it is today started then. And when it happened, w- we kind of went... well, Ed (laughs) gave me a call, this, like, panicked phone call. He's like, "Dude, I don't think we can keep working, like, business as usual in our company."
In a regular company anymore. Yeah.
Yeah.
So there was this AI model called GPT-3. So, like, everyone has, you know, maybe played with GPT-4. That's like ChatGPT. Um, GPT-3 was the generation before that, and it was the first time that you had an AI model that could get... that could actually, let's say, do stuff like write news articles that the average person, like in a paragraph of a news article, could not tell the difference between it wrote this news article and a real person wrote this news article. So that was an inflection, and that was, you know, significant in itself. But what was most significant was that it represented a point along this line, this, like, scaling trend for AI, where the signs were that you didn't have to be clever. You didn't have to come up with necessarily a revolutionary new algorithm or be smart about it. You just had to take what works and make it way, way, way bigger. And the significance of that is you increase the amount of computing cycles you put against something, you increase the amount of data. All of that is an engineering problem, and you can solve it with money. So you've got... you can scale up the system, use it to make money, and put that money right back into scaling up the system some more. Money in, IQ points come out.
Install uListen to search the full transcript and get AI-powered insights
Get Full TranscriptGet more from every podcast
AI summaries, searchable transcripts, and fact-checking. Free forever.
Add to Chrome