#231 – Paul Scharre on how AI-controlled robots will and won't change war

#231 – Paul Scharre on how AI-controlled robots will and won't change war

In 1983, Stanislav Petrov, a Soviet lieutenant colonel, sat in a bunker watching a red screen flash “MISSILE LAUNCH.” Protocol demanded he report it to superiors, which would very likely trigger a retaliatory nuclear strike. Petrov didn’t. He reasoned that if the US were actually attacking, they wouldn’t fire just 5 missiles — they’d empty the silos. He bet the fate of the world on a hunch that his machine was broken. He was right.

Paul Scharre, the former Army Ranger who led the Pentagon team that wrote the US military’s first policy on autonomous weapons, has a question: What would an AI have done in Petrov’s shoes? Would an AI system have been flexible and wise enough to make the same judgement? Or would it immediately launch a counterattack?

Paul joins host Luisa Rodriguez to explain why we are hurtling toward a “battlefield singularity” — a tipping point where AI increasingly replaces humans in much of the military, changing the way war is fought with speed and complexity that outpaces humans’ ability to keep up.

Links to learn more, video, and full transcript: https://80k.info/ps

Militaries don’t necessarily want to take humans out of the loop. But Paul argues that the competitive pressure of warfare creates a “use it or lose it” dynamic. As former Deputy Secretary of Defense Bob Work put it: “If our competitors go to Terminators, and their decisions are bad, but they’re faster, how would we respond?”

Once that line is crossed, Paul warns we might enter an era of “flash wars” — conflicts that spiral out of control as quickly and inexplicably as a flash crash in the stock market, with no way for humans to call a timeout.

In this episode, Paul and Luisa dissect what this future looks like:

  • Swarming warfare: Why the future isn’t just better drones, but thousands of cheap, autonomous agents coordinating like a hive mind to overwhelm defences.
  • The Gatling gun cautionary tale: The inventor of the Gatling gun thought automating fire would reduce the number of soldiers needed, saving lives. Instead, it made war significantly deadlier. Paul argues AI automation could do the same, increasing lethality rather than creating “bloodless” robot wars.
  • The cyber frontier: While robots have physical limits, Paul argues cyberwarfare is already at the point where AI can act faster than human defenders, leading to intelligent malware that evolves and adapts like a biological virus.
  • The US-China “adoption race”: Paul rejects the idea that the US and China are in a spending arms race (AI is barely 1% of the DoD budget). Instead, it’s a race of organisational adoption — one where the US has massive advantages in talent and chips, but struggles with bureaucratic inertia that might not be a problem for an autocratic country.

Paul also shares a personal story from his time as a sniper in Afghanistan — watching a potential target through his scope — that fundamentally shaped his view on why human judgement, with all its flaws, is the only thing keeping war from losing its humanity entirely.

This episode was recorded on October 23-24, 2025.

Chapters:

  • Cold open (00:00:00)
  • Who’s Paul Scharre? (00:00:46)
  • How will AI and automation transform the nature of war? (00:01:17)
  • Why would militaries take humans out of the loop? (00:12:22)
  • AI in nuclear command, control, and communications (00:18:50)
  • Nuclear stability and deterrence (00:36:10)
  • What to expect over the next few decades (00:46:21)
  • Financial and human costs of future “hyperwar” scenarios (00:50:42)
  • AI warfare and the balance of power (01:06:37)
  • Barriers to getting to automated war (01:11:08)
  • Failure modes of autonomous weapons systems (01:16:28)
  • Could autonomous weapons systems actually make us safer? (01:29:36)
  • Is Paul overall optimistic or pessimistic about increasing automation in the military? (01:35:23)
  • Paul’s takes on AGI’s transformative potential and whether natsec people buy it (01:37:42)
  • Cyberwarfare (01:46:55)
  • US-China balance of power and surveillance with AI (02:02:49)
  • Policy and governance that could make us safer (02:29:11)
  • How Paul’s experience in the Army informed his feelings on military automation (02:41:09)

Video and audio editing: Dominic Armstrong, Milo McGuire, Luke Monsour, and Simon Monsour
Music: CORBIT
Coordination, transcripts, and web: Katy Moore

Episoder(333)

The 80,000 Hours Career Guide (2023)

The 80,000 Hours Career Guide (2023)

An audio version of the 2023 80,000 Hours career guide, also available on our website, on Amazon, and on Audible.If you know someone who might find our career guide helpful, you can get a free copy se...

4 Sep 20234h 41min

#162 – Mustafa Suleyman on getting Washington and Silicon Valley to tame AI

#162 – Mustafa Suleyman on getting Washington and Silicon Valley to tame AI

Mustafa Suleyman was part of the trio that founded DeepMind, and his new AI project is building one of the world's largest supercomputers to train a large language model on 10–100x the compute used to...

1 Sep 202359min

#161 – Michael Webb on whether AI will soon cause job loss, lower incomes, and higher inequality — or the opposite

#161 – Michael Webb on whether AI will soon cause job loss, lower incomes, and higher inequality — or the opposite

"Do you remember seeing these photographs of generally women sitting in front of these huge panels and connecting calls, plugging different calls between different numbers? The automated version of th...

23 Aug 20233h 30min

#160 – Hannah Ritchie on why it makes sense to be optimistic about the environment

#160 – Hannah Ritchie on why it makes sense to be optimistic about the environment

"There's no money to invest in education elsewhere, so they almost get trapped in the cycle where they don't get a lot from crop production, but everyone in the family has to work there to just stay a...

14 Aug 20232h 36min

#159 – Jan Leike on OpenAI's massive push to make superintelligence safe in 4 years or less

#159 – Jan Leike on OpenAI's massive push to make superintelligence safe in 4 years or less

In July, OpenAI announced a new team and project: Superalignment. The goal is to figure out how to make superintelligent AI systems aligned and safe to use within four years, and the lab is putting a ...

7 Aug 20232h 51min

We now offer shorter 'interview highlights' episodes

We now offer shorter 'interview highlights' episodes

Over on our other feed, 80k After Hours, you can now find 20-30 minute highlights episodes of our 80,000 Hours Podcast interviews. These aren’t necessarily the most important parts of the interview, a...

5 Aug 20236min

#158 – Holden Karnofsky on how AIs might take over even if they're no smarter than humans, and his 4-part playbook for AI risk

#158 – Holden Karnofsky on how AIs might take over even if they're no smarter than humans, and his 4-part playbook for AI risk

Back in 2007, Holden Karnofsky cofounded GiveWell, where he sought out the charities that most cost-effectively helped save lives. He then cofounded Open Philanthropy, where he oversaw a team making b...

31 Jul 20233h 13min

#157 – Ezra Klein on existential risk from AI and what DC could do about it

#157 – Ezra Klein on existential risk from AI and what DC could do about it

In Oppenheimer, scientists detonate a nuclear weapon despite thinking there's some 'near zero' chance it would ignite the atmosphere, putting an end to life on Earth. Today, scientists working on AI t...

24 Jul 20231h 18min

Populært innen Fakta

fastlegen
dine-penger-pengeradet
relasjonspodden-med-dora-thorhallsdottir-kjersti-idem
rss-strid-de-norske-borgerkrigene
mikkels-paskenotter
foreldreradet
rss-bisarr-historie
treningspodden
jakt-og-fiskepodden
sinnsyn
rss-kunsten-a-leve
hverdagspsyken
ukast
rss-sunn-okonomi
rss-bak-luftfarten
fryktlos
tomprat-med-gunnar-tjomlid
lederskap-nhhs-podkast-om-ledelse
gravid-uke-for-uke
hagespiren-podcast