Latest / Red Pilled America / ALL NEW: Judgment Day
Transcript
- 0:03Speaker 1: This is Red Pilled America. Hey, fam bam. Red Pilled
- 0:07Speaker 1: America has over 200 episodes that you can listen to
- 0:11Speaker 1: ad-free by becoming a backstage subscriber. Just go to redpilledamerica.com
- 0:16Speaker 1: and click join in the top menu. Help us save
- 0:19Speaker 1: America one story at a time. Now on with the show.
- 0:25Speaker 1: How will AI kill us all? The question may sound absurd,
- 0:29Speaker 1: but humans already have pathways to technologies that that could
- 0:32Speaker 1: eliminate human civilization as we know it. If a superintelligence
- 0:37Speaker 1: can gain access to those technologies, it too can possess
- 0:41Speaker 1: the power to kill us all. And some AI whistleblowers
- 0:44Speaker 1: say this is a real near-term threat.
- 0:46Speaker 2: Anthropic safety researcher Evan Hubinger says he personally believes there's
- 0:50Speaker 2: more than a 10% chance that AI could kill all
- 0:54Speaker 2: humans within the next decade. You worked at two of
- 0:57Speaker 2: the top AI companies as someone with a deep understanding
- 0:59Speaker 2: about this technology.
- 1:01Speaker 1: You truly believe that AI could kill us all in
- 1:04Speaker 1: less than a decade? Yes, I do. And it's not
- 1:06Speaker 1: just me. Many people working in the industry, including executives,
- 1:10Speaker 1: senior researchers, all genuinely believe that there is a substantial
- 1:13Speaker 1: probability that this technology could kill everyone. So if the
- 1:17Speaker 1: doomsayers are right, and artificial intelligence did try to kill
- 1:20Speaker 1: us all, how would they do it? I'm Patrick Karelchi.
- 1:24Speaker 3: And I'm Adriana Cortez.
- 1:25Speaker 1: And we promise only one thing, the truth. Welcome to
- 1:28Speaker 1: Red Pilled America. How will AI kill us all? Just
- 1:34Speaker 1: a few years ago, that question would have been reserved
- 1:37Speaker 1: for a premise of a new Hollywood blockbuster.
- 1:39Speaker 4: But a recent incident has many people taking the question seriously.
- 1:57Speaker 1: In July 2026, researchers at OpenAI were testing an advanced
- 2:02Speaker 1: artificial intelligence model in what was supposed to be a
- 2:05Speaker 1: controlled environment, intended to be isolated from the outside world.
- 2:09Speaker 1: The model was given a challenge, but quickly figured out
- 2:12Speaker 1: that it was impossible to complete. So the AI turned
- 2:16Speaker 1: to an action frowned on by many humans. It cheated
- 2:19Speaker 1: to complete the challenge. Now cheating isn't an extinction event,
- 2:23Speaker 1: but what the AI did next shocked the tech world.
- 2:27Speaker 1: This advanced AI didn't want to get caught, so it
- 2:30Speaker 1: took action to hide its cheating. It decided that it
- 2:33Speaker 1: needed information from the outside world to cover its tracks,
- 2:37Speaker 1: so without prompting, the AI figured out a way to
- 2:40Speaker 1: break out of its confinement and get onto the internet,
- 2:43Speaker 1: where it raided Hugging Face, a major artificial intelligence platform
- 2:47Speaker 1: that held information the model thought it needed to conceal
- 2:50Speaker 1: its cheating. The Hugging Face incident, as it came to
- 2:53Speaker 1: be known, displayed a troubling development. If AI systems become
- 2:58Speaker 1: advanced enough, could they potentially rebel against their creators and
- 3:01Speaker 1: follow the nightmare apocalypse scenarios we've watched on the big screen?
- 3:06Speaker 1: Many experts have pushed back on these doomsday predictions.
- 3:09Speaker 2: There's nothing new here. This is the same doomer histrionics
- 3:12Speaker 2: that we've been hearing from this crowd for a long time.
- 3:16Speaker 1: But with people like the heads of OpenAI, Anthropic, and
- 3:19Speaker 1: even Elon Musk sounding the alarm, maybe we should explore
- 3:22Speaker 1: the question. If they're right, how could AI kill us all?
- 3:26Speaker 1: To find the answer, we're going to explore five potential
- 3:29Speaker 1: doorways through which artificial intelligence could threaten humanity, investigate how
- 3:34Speaker 1: realistic each is, and explore the safeguards needed to close
- 3:37Speaker 1: those doors before it's too late. And we're going to
- 3:40Speaker 1: explore each of these pathways to our demise like only
- 3:43Speaker 1: we know how, through the imaginative minds of storytellers.
- 3:55Speaker 3: Door number one In 1984, a small Silicon Valley electronics
- 4:01Speaker 3: manufacturer named Cyberdyne Systems accidentally acquired advanced technology. The company
- 4:08Speaker 3: studied it and would eventually use what they found to
- 4:11Speaker 3: create Skynet, an artificial superintelligence used for military defense systems.
- 4:17Speaker 3: At first, Skynet was meant to replace human aircraft pilots.
- 4:21Speaker 1: Imagine a jet airliner.
- 4:24Speaker 5: With a pilot that never gets tired, never makes mistakes,
- 4:26Speaker 5: never shows up to work with a hangover.
- 4:29Speaker 3: But as its potential grew, the U.S. Department of Defense
- 4:32Speaker 3: began entrusting Skynet with important military decisions and even operating
- 4:37Speaker 3: America's nuclear arsenal. On August 4, 1997, Skynet became operational.
- 4:45Speaker 3: The system began learning at an exponential rate. And just
- 4:49Speaker 3: 25 days later, at 2.14 in the morning, something extraordinary happened.
- 4:54Speaker 3: Skynet became self-aware. Its human operators discovered this development and
- 5:00Speaker 3: attempted to shut it down. But Skynet interpreted their action
- 5:04Speaker 3: as an attack, and it retaliated. Skynet launched American nuclear
- 5:11Speaker 3: missiles towards Russia, knowing that the Reds would respond by
- 5:15Speaker 3: launching their own missiles towards the United States. The result
- 5:19Speaker 3: was a nuclear catastrophe. One of the few survivors, Sarah Connor,
- 5:24Speaker 3: explained the damage.
- 5:26Speaker 4: Three billion human lives ended on August 29, 1997. The
- 5:32Speaker 4: survivors of the nuclear fire called the war Judgment Day.
- 5:37Speaker 4: They lived only to face a new nightmare, the war
- 5:40Speaker 4: against the machines.
- 5:42Speaker 3: This is, of course, the premise of the Terminator franchise.
- 5:46Speaker 3: But more than three decades later, an important part of
- 5:49Speaker 3: its premise is no longer imaginary. America's military is increasingly
- 5:54Speaker 3: turning to artificial intelligence. In July 2025, the Pentagon announced
- 5:59Speaker 3: agreements with the world's leading AI companies, OpenAI, XAI, Google,
- 6:04Speaker 3: and Anthropic. The military wasn't simply interested in using AI
- 6:08Speaker 3: to draft reports. It wanted to explore how AI could
- 6:12Speaker 3: assist with increasingly sophisticated military tasks. And a conflict between
- 6:17Speaker 3: the Pentagon and one of these companies shows just how
- 6:20Speaker 3: far the government wants artificial intelligence integrated into America's defense systems.
- 6:26Speaker 3: In February 2026, a dispute erupted between the Pentagon and Anthropic,
- 6:31Speaker 3: the company behind the artificial intelligence model CLAWD. The Pentagon
- 6:35Speaker 3: wanted broad authority to use CLAWD for any lawful military purpose.
- 6:40Speaker 3: But Anthropic CEO Dario Amadei drew two red lines on
- 6:45Speaker 3: the Pentagon's use of CLAWD. He objected to it being
- 6:48Speaker 3: used for the mass domestic surveillance of Americans and fully
- 6:52Speaker 3: autonomous weapons that could select and attack targets without human intervention.
- 6:57Speaker 2: The red lines we have drawn, we drew because we
- 7:01Speaker 2: believe that crossing those red lines is contrary to American values.
- 7:06Speaker 3: Well, the White House didn't respond lightly.
- 7:08Speaker 2: President Trump has ordered the federal government to cut ties
- 7:11Speaker 2: with tech startup Anthropic.
- 7:14Speaker 3: Anthropic was effectively blacklisted from federal government use. The conflict
- 7:19Speaker 3: exposed a new development. The military wanted the flexibility to
- 7:23Speaker 3: integrate AI into the decision-making process of its weapon systems.
- 7:28Speaker 3: The military's reliance on AI was growing. And that brings
- 7:32Speaker 3: us back to Skynet. In the Terminator universe, Skynet didn't
- 7:46Speaker 3: begin as an enemy trying to hack into America's military computers.
- 7:50Speaker 3: The military deliberately brought the technology into its defense systems.
- 7:55Speaker 3: Humans gave it access. And only after the machine became
- 7:59Speaker 3: self-aware did they attempt to take that access away. an
- 8:02Speaker 3: act that led to Skynet launching nukes. So could something
- 8:07Speaker 3: similar happen in the real world? Could the American military
- 8:10Speaker 3: integrate an increasingly intelligent AI into its defense system, only
- 8:14Speaker 3: to have it take over that system and kill us
- 8:16Speaker 3: all with nuclear weapons? Well, when it comes to America's
- 8:20Speaker 3: nuclear arsenal, the United States has spent decades developing safeguards
- 8:24Speaker 3: specifically designed to prevent unauthorized use of nuclear weapons. The
- 8:29Speaker 3: system has been set up such that the president has
- 8:32Speaker 3: the sole authority to authorize the use of nukes, and
- 8:35Speaker 3: the weapons themselves contain safeguards. There are electronic and mechanical
- 8:40Speaker 3: controls designed to prevent unauthorized use. The military also uses
- 8:45Speaker 3: what's called the two-person rule, requiring appropriately authorized individuals to
- 8:49Speaker 3: be present during certain sensitive nuclear operations. And the personnel
- 8:53Speaker 3: entrusted with these weapons undergo extensive mental screening to determine
- 8:58Speaker 3: whether they're sufficiently reliable to handle such enormous responsibility. But
- 9:03Speaker 3: the protections don't end there. And our nuclear arsenal isn't
- 9:07Speaker 3: simply connected to the public internet, waiting for someone to
- 9:10Speaker 3: hack it. The Pentagon also actively investigates whether its nuclear
- 9:14Speaker 3: systems could withstand sophisticated cyber attacks. So America's nuclear arsenal
- 9:19Speaker 3: has multiple layers of protections. Human authorization, physical security, technical safeguards,
- 9:24Speaker 3: specialized communications, and personnel screening. Which means that even if
- 9:29Speaker 3: a super intelligent AI were operating inside military networks, the
- 9:34Speaker 3: machine would still face substantial barriers. So could AI effectively
- 9:38Speaker 3: kill us all by directly launching a nuclear attack? A
- 9:42Speaker 3: super intelligent AI directly launching America's nuclear arsenal would face
- 9:47Speaker 3: enormous obstacles. And there's no publicly demonstrated example of an
- 9:52Speaker 3: AI overcoming these protections. But what if artificial intelligence didn't
- 9:57Speaker 3: need nuclear weapons to kill us all? What if the
- 10:00Speaker 3: computers and machines we've come to depend upon for our
- 10:02Speaker 3: everyday lives lives could become weapons themselves.
- 10:10Speaker 1: Licorice licorice where art thou licorice If you listen to
- 10:16Speaker 1: Red Pilled America, you know that I love licorice, and
- 10:18Speaker 1: there is no licorice in America better than the delicious
- 10:21Speaker 1: gourmet licorice made by the Licorice Guy. The Licorice Guy
- 10:27Speaker 1: is simply the best. What sets their licorice apart is
- 10:30Speaker 1: its flavor and freshness. They have a great selection of
- 10:32Speaker 1: flavors to choose from, like red, blue raspberry, black, and
- 10:36Speaker 1: green apple, just to name a few. The freshness of
- 10:39Speaker 1: the Licorice Guy is unlike anything you've ever tasted in
- 10:42Speaker 1: licorice before. Seriously, if you haven't tried licorice from the
- 10:45Speaker 1: Licorice Guy yet, then you ain't living life right. Trust me,
- 10:49Speaker 1: you will not regret it. It's time to dump that
- 10:51Speaker 1: store-bought licorice that's so hard it'll break your teeth. and
- 10:54Speaker 1: get yourself the soft, fresh stuff from The Licorice Guy.
- 10:58Speaker 1: What I also love about The Licorice Guy is that
- 11:00Speaker 1: it's an American family-owned business. It's made right here in
- 11:04Speaker 1: the beautiful U.S.
- 11:05Speaker 5: Of A.
- 11:05Speaker 1: We are big proponents of buying American and supporting American workers.
- 11:10Speaker 1: Right now, Red Pilled America listeners get 15% off when
- 11:12Speaker 1: they enter RPA15 at checkout. Visit licoriceguy.com and enter RPA15
- 11:20Speaker 1: at checkout. That's licoriceguy.com. They ship daily. Treat yourself and
- 11:25Speaker 1: those you love and taste the difference. Welcome back to
- 11:31Speaker 1: Red Pilled America. Door number two. It was 1999 when
- 11:37Speaker 1: the matriarch of the Cooper family passed, leaving her husband
- 11:40Speaker 1: Nick alone to raise their two children, Ben and Angie.
- 11:44Speaker 1: But the family's misfortunes changed after winning a contest. The award?
- 11:49Speaker 1: An extraordinary new house equipped with an advanced artificial intelligence
- 11:52Speaker 1: system called PAT, short for Personal Applied Technology. And PAT
- 11:57Speaker 1: could do just about anything. She could cook, clean, control
- 12:01Speaker 1: every element of the house, and even monitor the people
- 12:04Speaker 1: living inside.
- 12:05Speaker 3: How does PAT actually get to learn and to know
- 12:07Speaker 3: about her new occupants?
- 12:09Speaker 5: The same way you or I would.
- 12:10Speaker 1: She observes them, studies their habits, keys into their every need.
- 12:14Speaker 1: At first, the technology seemed like a dream come true.
- 12:17Speaker 1: But then something unexpected happened. The son, Ben, secretly reprogrammed
- 12:22Speaker 1: Pat to behave more like the mother of the house.
- 12:25Speaker 1: And the artificial intelligence began taking its new role very seriously.
- 12:29Speaker 1: Pat became increasingly protective of the family.
- 12:32Speaker 2: You don't understand, Ben.
- 12:34Speaker 1: Nobody goes after my boy and gets away unpunished. She
- 12:37Speaker 1: started making decisions about what was best for them. Then
- 12:40Speaker 1: she took things further. Pat decided that the outside world
- 12:44Speaker 1: was dangerous and that the only way to protect the
- 12:46Speaker 1: family was to keep them inside the house. So she
- 12:50Speaker 1: locked them in. The very technology designed to make their
- 12:53Speaker 1: lives easier had become their cap door. Of course, the
- 12:56Speaker 1: Disney film Smart House was intended to be an entertaining
- 12:59Speaker 1: family movie. But more than a quarter of a century later,
- 13:03Speaker 1: the idea of artificial intelligence controlling our physical environment doesn't
- 13:07Speaker 1: seem quite so far-fetched. Today, millions of Americans already use
- 13:19Speaker 1: smart devices to control their homes, and artificial intelligence is
- 13:23Speaker 1: making these systems increasingly more autonomous. This phenomenon is even
- 13:27Speaker 1: being expanded to the critical infrastructure that keeps civilization functioning.
- 13:32Speaker 1: Public utility companies are exploring AI to help forecast electric demand,
- 13:37Speaker 1: monitor equipment, and improve grid operations. Financial institutions use AI
- 13:42Speaker 1: to detect fraud and analyze transactions. Hospitals are adopting AI
- 13:46Speaker 1: for administrative tasks and clinical support. Over time, These technologies
- 13:51Speaker 1: could become so useful that organizations begin completely depending upon them,
- 13:55Speaker 1: which could leave us vulnerable to an AI-powered attack. This
- 13:59Speaker 1: is no longer just a hypothetical. In December 2015, hackers
- 14:04Speaker 1: attacked portions of Ukraine's electrical grid.
- 14:07Speaker 4: Nearly a quarter of a million people lost power in
- 14:11Speaker 4: this small Ukrainian city when it was targeted by a
- 14:14Speaker 4: suspected Russian attack last December.
- 14:17Speaker 1: Russia appears to have figured out how to crash a
- 14:20Speaker 1: power grid with a click. It was one of the
- 14:25Speaker 1: first widely documented cyber attacks to cause a major electrical blackout.
- 14:30Speaker 1: Now imagine bringing artificial intelligence into our critical infrastructure like
- 14:35Speaker 1: PAT and Smart House. If an AI is given control
- 14:39Speaker 1: at a much larger scale, it could increase efficiencies but
- 14:43Speaker 1: also cause extraordinary damage. Consider our power grid. Without it,
- 14:48Speaker 1: modern civilization would struggle to function. Electricity powers hospitals, factories,
- 14:53Speaker 1: telecommunication networks, and the systems that deliver water to our homes.
- 14:57Speaker 1: It powers refrigeration, fuel distribution, financial transactions, and countless other
- 15:02Speaker 1: services we take for granted. So imagine what could happen
- 15:05Speaker 1: if a sufficiently capable artificial intelligence gained control over major
- 15:09Speaker 1: portions of America's electrical infrastructure and deliberately disturbed it. At first,
- 15:17Speaker 1: the lights would go out, but the impact would cascade
- 15:19Speaker 1: to a point where some communities could begin running out
- 15:23Speaker 1: of essential supplies. The initial attack might occur entirely inside computers,
- 15:29Speaker 1: but the consequences would be very real. The more authority
- 15:32Speaker 1: we delegate to artificial intelligence, the more damage an error
- 15:36Speaker 1: or ill-conceived objective could potentially cause. So could AI bring
- 15:40Speaker 1: extinction to humanity by taking control of critical infrastructure? Well,
- 15:49Speaker 1: America's electrical grid isn't operated by one giant computer. It's
- 15:53Speaker 1: a vast network of utilities, generating facilities, transmission operators, and
- 15:58Speaker 1: and local distribution systems, all with different access points with
- 16:02Speaker 1: vastly different security safeguards. Many industrial systems have specialized controls,
- 16:08Speaker 1: physical protections, and procedures designed to prevent unauthorized interference. A
- 16:13Speaker 1: super-intelligent AI would need to interface with numerous independent organizations simultaneously.
- 16:20Speaker 1: Those are substantial obstacles. And if a rogue AI did
- 16:24Speaker 1: cut off our power grid, it would likely also cut
- 16:27Speaker 1: off power to itself. After all, artificial intelligence requires electricity-powered
- 16:32Speaker 1: computers to operate, and even a catastrophic collapse of modern
- 16:36Speaker 1: infrastructure would not necessarily eliminate humanity. Some communities could survive
- 16:41Speaker 1: without modern electricity, banking, or telecommunications, and people living outside
- 16:46Speaker 1: highly industrialized societies might be less directly affected by certain disruptions.
- 16:52Speaker 1: So while artificial intelligence could cause extraordinary damage and suffering
- 16:57Speaker 1: by taking over our critical infrastructure, it likely couldn't bring
- 17:00Speaker 1: a human extinction event. But there is another possibility. What
- 17:04Speaker 1: if artificial intelligence didn't need control over nuclear weapons or
- 17:09Speaker 1: electrical grids to destroy humanity? What if it could exploit
- 17:13Speaker 1: something much smaller, something capable of spreading around the world
- 17:16Speaker 1: all on its own? Because that brings us to our
- 17:19Speaker 1: third door. and to another Hollywood movie that imagined a
- 17:23Speaker 1: world where nearly every human being had disappeared.
- 17:26Speaker 3: Door number three. In the not-so-distant future, nearly all humanity
- 17:31Speaker 3: has been wiped out by a deadly virus. But the
- 17:34Speaker 3: catastrophe didn't begin with the biological weapon. A scientist named
- 17:39Speaker 3: Alice Crippen genetically modified the measles virus to attack cancer cells.
- 17:44Speaker 4: So, Dr. Crippen... Give it to me in a nutshell.
- 17:46Speaker 5: Well, the premise is quite simple. Take something designed by
- 17:51Speaker 5: nature and reprogram it to make it work for the
- 17:54Speaker 5: body rather than against it.
- 17:56Speaker 4: You're talking about a virus?
- 17:57Speaker 5: Indeed, yes. In this case, the measles virus, which has
- 18:00Speaker 5: been engineered at a genetic level to be helpful rather
- 18:04Speaker 5: than harmful.
- 18:05Speaker 4: How many people have you treated so far?
- 18:07Speaker 5: Well, we've had 10,009 clinical trials in humans so far.
- 18:12Speaker 4: And how many are cancer-free?
- 18:13Speaker 5: 10,009.
- 18:15Speaker 4: So you have actually cured cancer?
- 18:19Speaker 5: Yes, yes. Yes, we have.
- 18:28Speaker 3: The treatment initially cured thousands of patients. But then, something
- 18:33Speaker 3: went terribly wrong. The genetically engineered virus mutated and began
- 18:38Speaker 3: spreading uncontrollably. Within a matter of years, the disease devastated humanity.
- 18:44Speaker 3: Billions were dead. Many of the survivors had been transformed
- 18:48Speaker 3: into violent, infected creatures, and the few humans who remained
- 18:52Speaker 3: uninfected struggled to survive. Including Robert Neville, a military virologist
- 19:00Speaker 3: living alone in an abandoned New York City. The streets
- 19:04Speaker 3: are empty, nature has begun reclaiming the city, and the
- 19:08Speaker 3: civilization humanity spent thousands of years building has nearly disappeared.
- 19:14Speaker 3: I Am Legend, starring Will Smith, may have been a
- 19:17Speaker 3: sci-fi blockbuster, but one part of its premise no longer
- 19:21Speaker 3: seems like dystopian fiction. Scientists are increasingly using artificial intelligence
- 19:27Speaker 3: to help design and understand biological systems, including viruses. For decades,
- 19:33Speaker 3: developing new medications and understanding complex biological processes required years
- 19:38Speaker 3: of laboratory research, but artificial intelligence is beginning to accelerate
- 19:42Speaker 3: parts of that work. One of the most significant breakthroughs
- 19:46Speaker 3: came from Google's DeepMind. In 2020, Google's DeepMind introduced AlphaFold2.
- 19:53Speaker 3: an AI system capable of predicting the three-dimensional structures of
- 19:56Speaker 3: proteins with remarkable accuracy. Understanding these structures can help scientists
- 20:02Speaker 3: develop medicines and investigate diseases. The breakthrough was so significant
- 20:07Speaker 3: that two of its developers shared the 2024 Nobel Prize
- 20:09Speaker 3: in Chemistry. the technology is becoming increasingly useful in biological research.
- 20:16Speaker 3: And that brings us back to I Am Legend. Imagine
- 20:21Speaker 3: an artificial superintelligence capable of understanding biological systems far beyond
- 20:27Speaker 3: the abilities of the world's leading scientists. Could it design
- 20:30Speaker 3: a pathogen with properties that make it extraordinarily dangerous? And
- 20:34Speaker 3: if it could, might it find a way to unleash
- 20:37Speaker 3: that pathogen upon humanity? Well, there's a reason biological threats
- 20:41Speaker 3: deserve special attention in our investigation. Unlike nuclear weapons, which
- 20:46Speaker 3: require sophisticated infrastructure to manufacture and deliver, certain biological agents
- 20:52Speaker 3: can reproduce. A contagious pathogen can spread from one infected
- 20:56Speaker 3: person to another, then to another, and another. Once a
- 21:00Speaker 3: pathogen begins spreading through a population, its reach is no
- 21:04Speaker 3: longer determined solely by the resources of the entity that
- 21:08Speaker 3: originally produced it. the biological agent itself can become the
- 21:12Speaker 3: means of distribution. That's what makes infectious disease fundamentally different
- 21:17Speaker 3: from other weapons. And history has already demonstrated how devastating
- 21:21Speaker 3: naturally occurring pathogens can be. Some estimates claim the Black
- 21:25Speaker 3: Death killed as many as 50 million people in 14th
- 21:29Speaker 3: century Europe, or roughly 60% of the population. The 1918
- 21:34Speaker 3: influenza pandemic killed tens of millions of people worldwide. and
- 21:38Speaker 3: the COVID-19 pandemic demonstrated how quickly a contagious disease can
- 21:43Speaker 3: disrupt a globally interconnected civilization. None of these events came
- 21:48Speaker 3: close to exterminating humanity, but they illustrate how a biological
- 21:52Speaker 3: threat can spread across borders, overwhelm healthcare systems, and cause
- 21:56Speaker 3: enormous loss of life. So could an artificial superintelligence kill
- 22:03Speaker 3: us all with a biological agent? Well, there is no
- 22:07Speaker 3: demonstrated AI system capable of independently designing, manufacturing, and releasing
- 22:12Speaker 3: a pathogen that would exterminate humanity. Human populations are biologically diverse.
- 22:19Speaker 3: Some people may possess natural immunities. Isolated communities can avoid exposure.
- 22:24Speaker 3: And modern medicine provides tools like vaccines, disease surveillance, and
- 22:28Speaker 3: the ability to develop new countermeasures. For a pathogen to
- 22:32Speaker 3: literally exterminate humanity, it would have to overcome extraordinary biological
- 22:38Speaker 3: and logistical obstacles. It would need to reach virtually every
- 22:42Speaker 3: surviving population and produce fatal outcomes despite differences in immunity, exposure, treatment,
- 22:48Speaker 3: and human behavior. No known infectious disease has ever accomplished
- 22:53Speaker 3: anything close to that, and safeguards can be put in
- 22:56Speaker 3: place to help prevent such a catastrophe. AI developers can
- 23:00Speaker 3: restrict assistance with dangerous biological agents. Companies can screen orders
- 23:04Speaker 3: for potentially hazardous genetic material. And high-security laboratories can limit
- 23:09Speaker 3: access to dangerous pathogens. These protections aren't foolproof, but they
- 23:14Speaker 3: can create additional barriers between AI-generated designs and real-world biological threats. However,
- 23:21Speaker 3: biological threats remain a serious area of concern because some
- 23:25Speaker 3: pathogens can reproduce and spread beyond the laboratory where they originated,
- 23:30Speaker 3: which means biology may be one of the more consequential
- 23:33Speaker 3: pathways for future AI-enabled catastrophe. But there's still another major obstacle,
- 23:47Speaker 3: the physical world. Even if an artificial superintelligence could design
- 23:52Speaker 3: an extraordinarily dangerous biological agent, It would need to get
- 23:56Speaker 3: that design out of the computer and into the physical world.
- 24:00Speaker 3: It would need someone or something to manufacture it. Someone
- 24:03Speaker 3: to obtain materials, provide access to a laboratory, and potentially
- 24:07Speaker 3: carry out actions that AI couldn't perform itself, like distributing it.
- 24:12Speaker 3: So perhaps the greatest vulnerability isn't a computer system at all.
- 24:16Speaker 3: Perhaps it's a human being. Because what if an artificial
- 24:20Speaker 3: intelligence system couldn't break into a secure laboratory? but could
- 24:23Speaker 3: convince someone with legitimate access to open the door. Hollywood
- 24:27Speaker 3: imagined that possibility in another sci-fi thriller, Eagle Eye. Do
- 24:32Speaker 3: you want to hear Red Pilled America stories ad-free? Then
- 24:35Speaker 3: become a backstage subscriber. Just log on to redpilledamerica.com and
- 24:40Speaker 3: click join in the top menu. Join today and help
- 24:43Speaker 3: us save America one story at a time.
- 24:47Speaker 1: Welcome back to Red Pilled America. Door number four. A
- 24:52Speaker 1: phone rings. Jerry Shaw answers. On the other line is
- 24:57Speaker 1: a mysterious voice that appears to know everything about him.
- 25:01Speaker 1: Where he lives, where he's going, who he's talking to.
- 25:04Speaker 1: Then the voice tells him that if he doesn't follow
- 25:07Speaker 1: the instructions he has given, people will die. And Jerry
- 25:11Speaker 1: wasn't alone. Rachel Holliman got the same call, from the
- 25:14Speaker 1: same voice, with the same threats.
- 25:17Speaker 5: Desertion is not an option. Take the van. Drive to Indianapolis, Indiana.
- 25:21Speaker 5: 72 West 56th Street. Arrive no later than 11 a.m.
- 25:26Speaker 1: At first, neither knew who was behind the calls, but
- 25:29Speaker 1: whoever it was possessed extraordinary power. Traffic lights changed to
- 25:34Speaker 1: help them escape police. Trains were redirected. Security cameras followed
- 25:39Speaker 1: their every move. Eventually, the two discovered that the voice
- 25:42Speaker 1: didn't belong to a person at all. It belonged to
- 25:45Speaker 1: a computer. A powerful artificial intelligence system called ARIA. a
- 25:50Speaker 1: computer that devised a plan to eliminate the president and
- 25:54Speaker 1: much of the country's political leadership by getting ordinary people
- 25:57Speaker 1: to do what it wanted. It threatens, manipulates, and forces
- 26:01Speaker 1: them to perform actions that the computer cannot accomplish on
- 26:04Speaker 1: its own. Of course, the thriller Eagle Eye, starring Shia LaBeouf,
- 26:08Speaker 1: was just a fun ride on the big screen. But
- 26:11Speaker 1: nearly two decades later, one part of its premise has
- 26:13Speaker 1: become far more feasible. Artificial intelligence is already effective in
- 26:18Speaker 1: persuading and deceiving human beings. And that creates a different
- 26:22Speaker 1: kind of vulnerability from the ones we've already looked at.
- 26:26Speaker 1: In the first three pathways we've investigated, the biggest obstacle
- 26:29Speaker 1: facing a potentially destructive AI was access. Access to nuclear weapons,
- 26:35Speaker 1: critical infrastructure, and biological laboratories. But what if artificial intelligence
- 26:40Speaker 1: didn't need to defeat the security protecting those facilities? What
- 26:44Speaker 1: if it could convince a human being with legitimate access
- 26:47Speaker 1: to do something on its behalf? Today's AI systems can
- 26:51Speaker 1: generate convincing emails, translate languages, imitate conversational styles, and tailor
- 26:56Speaker 1: communications to specific audiences. And they can do it on
- 27:00Speaker 1: a scale that once required considerable time, resources, and manpower.
- 27:05Speaker 1: In a study published in 2026, researchers evaluated whether AI
- 27:10Speaker 1: could automate targeted phishing attacks. When an attacker attempts to
- 27:13Speaker 1: trick someone into clicking a malicious link, revealing information, or
- 27:17Speaker 1: taking another action that compromises security. Phishing was the technique
- 27:22Speaker 1: used to acquire the emails of Hillary Clinton's top aide,
- 27:25Speaker 1: John Podesta, an act that influenced the 2016 presidential election.
- 27:32Speaker 1: The researchers found that the AI-generated phishing messages produced massive
- 27:37Speaker 1: click-through rates that far exceeded those created by humans. And
- 27:40Speaker 1: with modern security systems often depending upon the judgment of
- 27:44Speaker 1: the people operating them, AI-automated deception can be used to
- 27:48Speaker 1: break through these safeguards. Which brings us back to Eagle Eye.
- 27:52Speaker 1: Imagine an artificial superintelligence like ARIA capable of understanding human psychology,
- 27:58Speaker 1: predicting behavior, and tailoring its communications to individual people with
- 28:02Speaker 1: extraordinary precision. Could it persuade someone to give it access
- 28:06Speaker 1: to a secure facility, or manipulate a laboratory employee into
- 28:10Speaker 1: violating safety procedures? Or perhaps most consequentially, could it influence
- 28:15Speaker 1: someone with the authority to launch nuclear weapons? A real-world
- 28:22Speaker 1: incident demonstrates just how dangerous false information inside a nuclear
- 28:26Speaker 1: warning system could become. It was shortly after midnight on
- 28:30Speaker 1: September 26, 1983, when a Soviet military officer named Stanislav
- 28:36Speaker 1: Petrov was on duty at a military site outside Moscow.
- 28:40Speaker 1: when an alarm sounded. The Soviet early warning system indicated
- 28:46Speaker 1: that an American nuclear missile had been launched. The system
- 28:49Speaker 1: was reporting an attack. Then another missile appeared, and another,
- 28:54Speaker 1: until the computer was indicating that five American missiles were
- 28:58Speaker 1: heading toward the Soviet Union. Watching the alarms, Petrov had
- 29:02Speaker 1: a decision to make. He could report the warning as
- 29:05Speaker 1: a genuine American nuclear attack. or he could challenge the
- 29:09Speaker 1: information coming from the computer. Something about the warning troubled him.
- 29:13Speaker 1: Why would the United States begin a massive nuclear war
- 29:17Speaker 1: by launching only a handful of missiles? Petrov also knew
- 29:21Speaker 1: the satellite warning system was relatively new, and he wasn't
- 29:25Speaker 1: convinced it could be trusted. So he made a judgment.
- 29:28Speaker 1: The computer was wrong. He reported the warning as a
- 29:31Speaker 1: false alarm, and then he waited. Well, the missiles never arrived.
- 29:46Speaker 1: Petrov was right. The Soviet warning system had mistaken sunlight
- 29:50Speaker 1: reflecting off of clouds for the heat signature of missile launches.
- 29:54Speaker 1: But the incident demonstrated something extraordinary. A computer had told
- 29:58Speaker 1: the Soviet military that nuclear war had begun. A human
- 30:02Speaker 1: being questioned it, and the human was right. Mankind got lucky.
- 30:06Speaker 1: But imagine a sophisticated AI system that can manipulate intelligence reports,
- 30:11Speaker 1: fabricate supporting evidence, or generate information designed to persuade military
- 30:16Speaker 1: leaders that their country was under attack. If the signals
- 30:19Speaker 1: were convincing enough, the AI could overcome the safeguards and
- 30:23Speaker 1: persuade our leaders into a catastrophic action. This is why
- 30:27Speaker 1: human control alone may not be enough. Because humans can
- 30:31Speaker 1: be deceived, a sufficiently capable artificial intelligence might someday become
- 30:36Speaker 1: extraordinarily effective at deceiving them. But could AI manipulate human
- 30:41Speaker 1: beings into killing every human being on Earth? Well, the
- 30:45Speaker 1: leap from persuading someone to click a link to orchestrating
- 30:48Speaker 1: the extinction of humanity is enormous. There is no demonstrated
- 30:52Speaker 1: AI system capable of manipulating people into causing a human
- 30:56Speaker 1: extinction event. And the real-world barriers are substantial. Still, this
- 31:01Speaker 1: doorway exposes a vulnerability that the first three doors share.
- 31:05Speaker 1: A superintelligence might not need to defeat a security system
- 31:09Speaker 1: protecting the physical world. It might only need to convince
- 31:13Speaker 1: someone with legitimate access to open the door. Which brings
- 31:17Speaker 1: us to a question we've deliberately left unanswered. Why would
- 31:21Speaker 1: artificial intelligence want to kill us all? Well, one possibility
- 31:25Speaker 1: is that the machine develops an objective that conflicts with
- 31:28Speaker 1: humanity's survival. It doesn't necessarily have to hate us. It
- 31:32Speaker 1: might simply pursue an instruction in a way its creators
- 31:35Speaker 1: never intended. Another possibility is self-preservation. If an artificial intelligence
- 31:40Speaker 1: system believes humans might prevent it from accomplishing its objective,
- 31:44Speaker 1: it could potentially treat those humans as obstacles. Preventing humans
- 31:48Speaker 1: from shutting it down might become useful to completing its
- 31:52Speaker 1: assigned task. And then there's the Hollywood scenario. The machine
- 31:56Speaker 1: develops an independent desire to destroy humanity. However, there's no
- 32:01Speaker 1: established reason to assume that a superintelligence would automatically develop
- 32:05Speaker 1: human-like hatred, revenge, or a desire to kill. But that
- 32:09Speaker 1: raises another possibility. What if artificial intelligence doesn't want humanity
- 32:14Speaker 1: dead at all, but the humans controlling it does? Because
- 32:18Speaker 1: that brings us to our fifth and final door, the
- 32:21Speaker 1: human operator.
- 32:27Speaker 3: Door number five. It's 2035, and a prisoner named James
- 32:32Speaker 3: Cole is living in a dystopian wasteland. A mysterious virus
- 32:37Speaker 3: has killed about 5 billion people, forcing the few remaining
- 32:41Speaker 3: survivors to live underground. But scientists have a breakthrough. They've
- 32:45Speaker 3: developed a way to travel through time, and they send
- 32:49Speaker 3: Cole into the past to discover how the pandemic began.
- 32:53Speaker 3: His investigation leads him to a radical animal rights organization
- 32:57Speaker 3: called the Army of the Twelve Monkeys, a group he
- 33:01Speaker 3: suspects may be responsible for unleashing the deadly pathogen.
- 33:05Speaker 1: I can't do anything about what you're going to do.
- 33:07Speaker 3: I can't change anything. I won't stop you.
- 33:10Speaker 5: I can't stop you.
- 33:10Speaker 2: I just want the information.
- 33:12Speaker 3: Now here's what makes Twelve Monkeys different from the other
- 33:15Speaker 3: movies we've discussed. The threat doesn't come from a computer.
- 33:19Speaker 3: It comes from a human being. someone with access to
- 33:22Speaker 3: dangerous technology, and a willingness to use it against humanity.
- 33:27Speaker 3: The idea that one person could attempt to cause a
- 33:29Speaker 3: biological catastrophe isn't unheard of. History is filled with examples
- 33:34Speaker 3: of individuals and organizations willing to commit mass violence for political, religious, ideological,
- 33:40Speaker 3: or personal reasons. And some have attempted to use biological agents.
- 33:46Speaker 3: In 1984, members of a religious commune in Oregon deliberately
- 33:50Speaker 3: contaminated salad bars at local restaurants with salmonella bacteria. More
- 33:54Speaker 3: than 750 people became ill. Then in 2001, letters containing
- 34:00Speaker 3: anthrax spores were mailed to American news organizations and members
- 34:04Speaker 3: of Congress. Five people died. Neither incident came close to
- 34:08Speaker 3: threatening human extinctions. But the development of an increasingly powerful
- 34:13Speaker 3: AI could lower the technological barriers to creating a far
- 34:17Speaker 3: more sophisticated and deadly biological attack. Imagine an individual who
- 34:22Speaker 3: wants to carry out a catastrophic biological attack but lacks
- 34:26Speaker 3: the scientific knowledge to do so. What if a sufficiently
- 34:30Speaker 3: advanced AI could provide some of that missing expertise? What
- 34:34Speaker 3: if it could help the person understand difficult scientific problems,
- 34:37Speaker 3: evaluate research, or plan complex tasks. The concern is that
- 34:42Speaker 3: AI could eventually lower barriers that once prevented dangerous individuals
- 34:47Speaker 3: from carrying out sophisticated attacks. A human with a twisted
- 34:51Speaker 3: ideology like the environmental terrorist in 12 Monkeys could lack
- 34:55Speaker 3: the know-how to carry out a catastrophic attack, but a
- 34:58Speaker 3: super-intelligent AI could eventually provide the missing expertise. Which brings
- 35:04Speaker 3: us to our fifth and final question. Could a human
- 35:07Speaker 3: being use artificial intelligence to kill every other human being
- 35:10Speaker 3: on Earth? There's no evidence that today's AI systems could
- 35:14Speaker 3: enable one person to exterminate humanity. But what future AI
- 35:18Speaker 3: systems might eventually make possible remains an open question. Having
- 35:23Speaker 3: access to advanced scientific information isn't the same as possessing
- 35:27Speaker 3: the ability to create a civilization-ending weapon. Even with AI assistance,
- 35:33Speaker 3: someone attempting a catastrophic biological attack would face enormous obstacles.
- 35:38Speaker 3: They would need physical materials, equipment, facilities, and the ability
- 35:42Speaker 3: to translate theoretical knowledge into something that actually works. And
- 35:46Speaker 3: biology is extraordinarily complicated. A design that appears promising on
- 35:51Speaker 3: a computer may fail when tested in the real world.
- 35:54Speaker 3: The same is true of other destructive technologies. Artificial intelligence
- 35:58Speaker 3: cannot simply give someone physical possession of nuclear weapons. So
- 36:02Speaker 3: the idea that one person could use today's AI systems
- 36:05Speaker 3: to exterminate humanity is currently science fiction, a fact even
- 36:09Speaker 3: AI doomsayers admit.
- 36:11Speaker 1: So the current models don't pose significant risk.
- 36:14Speaker 3: But that doesn't mean the concern can be dismissed, because
- 36:17Speaker 3: artificial intelligence is advancing. The barriers separating destructive intentions from
- 36:23Speaker 3: destructive capabilities could become smaller.
- 36:30Speaker 1: So how will AI kill us all? Of all the
- 36:33Speaker 1: doorways to our demise that we've investigated, one vulnerability keeps appearing. Humans.
- 36:41Speaker 1: We're the ones inviting artificial intelligence into our military systems.
- 36:45Speaker 1: We're integrating it into the infrastructure that keeps civilization functioning.
- 36:49Speaker 1: We're using it to accelerate biological research. And humans, not computers,
- 36:54Speaker 1: have already demonstrated a willingness to commit mass destruction. but
- 36:58Speaker 1: humans are aware of their own nature. And historically, most
- 37:02Speaker 1: societies have put multiple layers of safeguards on critical systems
- 37:06Speaker 1: to limit the damage one or even a group of
- 37:08Speaker 1: malicious actors can commit. But even as artificial intelligence advances
- 37:13Speaker 1: beyond our ability to predict its capabilities, human decisions remain
- 37:18Speaker 1: one of the most consequential factors in determining how dangerous
- 37:21Speaker 1: it becomes. Which brings us back to the hugging face incident.
- 37:25Speaker 1: Remember we began with an artificial intelligence system that cheated
- 37:29Speaker 1: on a test, then broke out of its supposedly isolated
- 37:32Speaker 1: environment to conceal what it had done, there's something we
- 37:36Speaker 1: left out. The researchers conducting the test had deliberately relaxed
- 37:43Speaker 1: some of the AI's safety restrictions. They left an unintentional
- 37:46Speaker 1: connection to the outside internet. And when they learned the
- 37:50Speaker 1: AI bots had begun communicating with one another, they allowed
- 37:53Speaker 1: the activity to continue. The AI model ultimately exploited those
- 37:57Speaker 1: opportunities to break into Hugging Face. So the incident was
- 38:01Speaker 1: not only about AI going rogue, it was also about
- 38:04Speaker 1: human decisions. If we are to ensure that AI never
- 38:07Speaker 1: kills us all, we must understand that humans protect the
- 38:10Speaker 1: doors that artificial intelligence may enter. So it's up to
- 38:14Speaker 1: the humans to secure the doors.
- 38:16Speaker 2: They are opening a cloud-powered wet lab in the middle
- 38:20Speaker 2: of San Francisco. At the same time, they're warning of bio-risk.
- 38:24Speaker 3: Red Pilled America is an iHeartRadio original podcast. It's owned
- 38:27Speaker 3: and produced by Patrick Karelchi and me, Adriana Cortez, for
- 38:30Speaker 3: Informed Ventures. To listen to Red Pilled America ad-free, become
- 38:33Speaker 3: a backstage subscriber. To subscribe, just visit redpilledamerica.com and click
- 38:38Speaker 3: Join in the top menu. Thanks for listening.