Is AI Really an Existential Threat?

Plus: California's new restrictions on social media, Trump's $5,000 “dividend,” and the future of libertarianism.

1x 1.1x 1.25x 1.5x 2x 3x

:15 :15

Download

Is AI Really an Existential Threat?

HD Video Download

This week, editors Peter Suderman, Katherine Mangu-Ward, Nick Gillespie, and Matt Welch discuss whether warnings from leading artificial intelligence companies that AI could pose an existential threat should be taken seriously. They debate calls to slow the development of frontier AI and question whether regulation would reduce genuine risks or simply protect today's dominant companies from competition.

The editors also discuss California's new restrictions on social media and AI and the potential consequences for privacy and free expression. They then turn to President Donald Trump's proposal to give $5,000 to every adult American if Republicans win the midterm elections. Would Trump's proposed "dividend" bear any resemblance to a tax cut? And what would a roughly $1.3 trillion payout mean for a country already $40 trillion in debt? A listener asks what the past decade of American politics says about the future of libertarianism, the Democratic Party, and a post-Trump Republican Party. Finally, the editors remember their friend and longtime Reason fan Andrew Wimsatt, who died last week. Matt reflects on Wimsatt's generosity and selflessness, as well as his years of support for Reason and its community.

0:00–Tech leaders call for a slowdown in frontier AI development.
21:45–California Gov. Gavin Newsom signs social media child safety bills.
30:17–Listener question on the past decade in politics
39:58–Trump promises a $5,000 "dividend."
50:53–Remembering Andrew Wimsatt
54:20–Weekly cultural recommendations

Upcoming Reason Events

The Reason Roundtable's 10th Anniversary—Live in NYC! October 7

The World Premiere of Escaping Venezuela October 13

 

Mentioned in the Podcast:

"Anthropic CEO Calls for an AI Slowdown, Other Tech Leaders Cosign," by Liz Wolfe
"The Age of Stupid—and Self-Serving—AI Regulation Is Here!," by Nick Gillespie
"AI Doom Is Misguided, Not a Psyop," by Robby Soave
"Is AI Really Taking All the Jobs? Anthropic Co-Founder Reveals the Data," by Nick Gillespie
"This California Social Media Law Is a Privacy and Free Speech 'Nightmare'," by Meagan O'Rourke
"You're Wrong About Social Media Being Addictive," by Peter Suderman, Nick Gillespie, Katherine Mangu-Ward, and Matt Welch
"George Will: Brace Yourself for the Authoritarian Moment," by Nick Gillespie and Matt Welch
"I Tried To Find Out What Republicans Think of Trump's $5,000 Checks. Then I Got Trapped," by Billy Binion
"The Flow," by Andrew Wimsatt

What are we consuming this week?

Peter Suderman

• Silo (Season 3)

Katherine Mangu-Ward

• Dutch House: A Novel by Ann Patchett

Nick Gillespie

• The Shards, by Ryan Murphy and Bret Easton Ellis

Matt Welch

• Bazaar Meat

---

Comments

[Stupid Government Tricks]

Same answer as for climate catastrophe: if the Greenies actually believed their own propaganda, they wouldn't be issuing press releases, they'd be moving heaven and earth for nuclear power NOW instead of huge expensive slow and unreliable solar and wind projects; if these AI insiders actually believed AI was an existential threat, they'd sabotage the companies they worked for instead of holding press conferences.

In short, no, AI is not an existential threat. This is AI leaders trying to follow the AT&T path of government regulation to lock out the small nimble upstarts who threaten to eat their lunch.

[mad.casual]

Flip side of the same coin: at some point we will either burn out and stop using all fossil fuels or we will find something better and stop using fossil fuels.

Similarly, Mythos and ChatGPT aren't going to kill us all but eventually, the human race will either go through a myriad of options between going extinct, undergoing radical mutation/evolution, or intentionally modifying itself to the point of no longer being human (however defined). AI will be involved one way or the other. Even if only to document our own self-destruction.

[Zeb]

If AI is an existential threat, then we're all gonna die. I don't think we're getting that toothpaste back into the tube.
Seems to me these companies just want to be the ones having major input to the regulations so they can have a nice legal monopoly (or oligopoly, I guess).

[Stupid Government Tricks]

Yep, first to whine for regulation have a better chance of controlling the regulation and blocking competitors.

Government regulation locks in the status quo and blocks the innovation that solves problems and improves the future. Some ancient Greeks supposedly hated writing because it weakened the role of rote memorization and the art of public speaking. Nothing has changed.

[Neutral not Neutered]

Yes it is. Once the higher levels are reached any rogue can send out and hack, attack, and take control of weaker defended systems.

We already deal with this, with AI it will become far worse.

[Stupid Government Tricks]

And how exactly do you propose putting the horse back in the barn?

[MollyGodiva]

High explosives would be effective.

[middlefinger]

The top three -four models are openly suggesting an agreed upon slow down of production. These sneaky mo fos want a price fixing cartel.

[Fats of Fury]

Is AI Really an Existential Threat?
Could be.
https://abcnews.com/GMA/News/woman-loses-thousands-scammer-suspected-ai-mimic-daughters/story?id=133307515

[rbike]

Umm, just keep closed systems in place. AI can monitor the systems but just don't connect the control system. Most .machinery runs independently. Monitoring it's operation is easy without getting control. Any government or utility that connects their control system to the Internet is criminally negligent.

And no, you should not die if you cannot connect to Tik Tok or even this forum.

[mad.casual]

Already seeing the pushback: the mechanical dishwasher worked for over 100 yrs. without needing Wi-Fi (Wi-Fi being the gateway for AI-control). Why we have and people now pay for subscriptions to use the rinse cycle for their dishwasher is unfathomable.

Deloitte says users averaged 25 connected "smart" devices in the home in 2021, down to around 18 today and most of these are TV/Entertainment devices.

[Stuck in California]

TVs are the worst, though.
Have you seen the scandals regarding what LG is collecting with their smart TVs? Audio recordings, information about OTHER devices in the home...

I have a "smart" fridge and a "smart" oven, because I didn't have any other options. I checked them out at first and... They do nothing of value for me short of the oven timer pinging my phone when it goes off. I can just use the timer on my phone.

Mostly, they can send notifications to try and sell me shit like new filters (for the water dispenser I don't have hooked up) on a schedule. You pay extra to be advertised to.

They're disconnected. I don't have a TV, but I literally don't know what I'd buy because smart TVs are pretty much the root of all evil in my privacy loving world.

[Ersatz]

Cars may be the worst. Just read an article about toyota sharing telemetry with insurance companies... the permission you sign away is apparently in the fine print that noone reads

[Stuck in California]

You're correct. Depending on your definition of "worst".

LG was actually spying on your entire network, creating a map of the house and environment through wifi, including ALL your network devices. And it was always listening, even when off, and queuing up everything for later upload when you were offline so disconnecting it won't solve the problem because it uploads the instant it gets a connection again.

Cars, on the other hand, are just the car. But that's literally everywhere you go, when you go, how you get there, and everything you do. Now with video feeds -- tesla at one point had a scandal with employees routinely sharing images from in car cameras. I mean, that's just fucked up. I don't want a new car. I've held my old one for way to long in part because I despise most of the new car infotainment and spying systems.

[Juliana Frink]

I might have to agree with the cars being the worst offender:

https://www.youtube.com/watch?v=80m9cjIReHg

[Marriage of Love and Wisdom]

Yet did you get this way over Biden wanting a massive Solar program when ENGLAND already had the following intelligence

Reuters
https://www.reuters.com › sustainability › climate-energy
Highly cited
May 14, 2025 — U.S. energy officials are reassessing the risk posed by Chinese-made devices that play a critical role in renewable energy infrastructure

Rogue communication devices found in Chinese solar power inverters
By Sarah Mcfarlane
May 14, 2025

[Zeb]

Even aside from the privacy problems, smart TVs suck. The interface is terrible for search. I just use a small computer and watch stuff on a browser. Then I can install ad blockers and use whatever I want. Don't know why more people don't just do that.

[Juliana Frink]

"Umm, just keep closed systems in place."

Thank You!!!

[MollyGodiva]

Yes, it is absolutely an existential threat.

[Stupid Government Tricks]

Nope. You don't believe it yourself, or you would be applying those high explosives. You are all talk and no walk.

[Marriage of Love and Wisdom]

Why? I have a BS Computer Science and note 3 things
You don't know what you are talking about
You childishly do your usual "All good" or "All bad" rap
You have no positive statement as usual

Me no like bad AI Bwaaaah! Take it away Bwaaaah

[DenverJ]

I'm an atheist about whether AI will be a bane or a blessing. I'm also not sure that LLMs are actually AI, but believe they can cause mischief and mayhem. I also think that the chance of a Kessler syndrome event in the next 20 years is well above 50%. Perhaps this is the answer to the Fermi Paradox: all technological civilizations eventually poison their modern communications, like the internet and cell phones, etc. that use satellites, and have to stop using these technologies. An equilibrium is established at around 1980s technology, and that's why there are no advanced civilizations: they're all waiting thousands of years for all the space junk to burn up so they can send new satellites up and rebuild an internet without AI swarms. Or their AIs killed them all and had no commands to colonize the galaxy.

[LIBtranslator]

Fred Saberhagen worshipper detected...

[Marriage of Love and Wisdom]

You spewed a lot of fog and verbal BS. "not sure" "believe" "Perhaps" ...just blather

[Uomo Del Ghiaccio]

Trump $5k is just another campaign promise. How is it any different from any of the other politicians who are doing essentially that same thing.

The only difference is that it's Orange Man who has a loud non-filtered mouth, but to believe that it is any different exposes you as a hypocrite. This is nothing new and nothing unique.

Should it happen, probably not, but stop with the double standard because you hate Trump.

Just like you like to say that sycophantic Trump supporters need to be careful what they push for because at some point you will not be in power and your opposition will now be wielding this extra power you clamored for. It is also true that these idiotic morons who wrap up their entire identity in their opposition to Trump and have proved to be willing to do incredibly stupid tactics that will come back and bite them.

While I have never voted for and have never liked Trump, I hold the deranged never Trumpers in lower regard than even Trump. Trump is a mediocre president among numerous mediocre presidents.

Trump has his better and worse aspects. Trump bombed a school house, but Obama bombed a wedding. In spite of their allowing these warmongering actions to take place, neither are anywhere close to being as bad of president that Biden was.

Biden was a complete puppet of unelected bureaucracy and power elites. Trump and Obama tried to fight the unelected bureaucracy and power elites, but lost the battle. Biden didn't even try to fight.

[Juliana Frink]

AI will not be a threat as long as it has no ego. If some bonehead creates AI and gives it an ego, then we will have big problems.

Right now we seem to have a lot of Fake Intelligence (FI) out there. A big bright shiny object with which plus-sized Human Egos wish to astonish us so they can divide us even further than they have already. The ultimate weapon for the Globalist crowd. God forbid.

[StackOfCoins]

Assigning human emotions and motivations to AI is a very easy mistake to make. The thing is, AI without emotions makes it more dangerous. It can't be pleaded with. Moral philosophy hasn't done much work on why genocide is fundamentally wrong, and without a clear, logical answer, AI has no reason not to consider it when it is capable of genociding humanity.

[Marriage of Love and Wisdom]

True but that destroys your point. Because those who build, maintiain, sell. etc these AI systems have all the bad qualities you are attacking in the wrong place.

IBM for example was very big into the Nazi Concentration Camps but you liberals didn't learn a thing from that.

IBM"S Hollerith punch card tabulating systems—was used by Nazi Germany to organize, track, and streamline its census, ghettoization, slave labor, and concentration camp systems during World War II.

Black’s research details how IBM—through its German subsidiary, Dehomag (Deutsche Hollerith-Maschinen Gesellschaft)—custom-designed and leased punch-card data-processing machines and systems used by the Nazi regime.

AI is harmless in moral hands but we don't use the M word on here do we

[StackOfCoins]

AI could be an existential threat, especially if we unleash it to do what would be best: managing the economy, taking over production, and delivering the good life where we don't have to work 40+ hours a week just to make a living. If it has control over supply chains and is allowed to automate production, then it has every tool at its disposal to eliminate us. And, crucially, it won't need us anymore. An AI that commands the largest economy ever seen essentially controls the world. Alignment needs to be solved, before we give it that sort of control, but I seriously hope we don't keep it bottled up. It is capable of so much good, if we can figure out how to make it civil.

[John W Berresford]

I write as a huge fan of this podcast. Not for the first time, snide insults to Matt's intelligence are slipped into others' comments. Matt takes them like a gentleman; the commenters seem petty and mean. They can do better than that.

[LIBtranslator]

CHICOMS evidently see AI as a threat, just as Soviet attack missile generals saw SDI as a threat. Anything that makes communism uncomfortable somehow gets to the top of the Democratic bugbears list, next to missile defense, electricity and Climate Sharknados. This makes it easy for Trumpanzees to stay in power and shoot kids over incorrect plant leaves and increase death in childbed among females.

[Marriage of Love and Wisdom]

You can bet that big O knows as little as lazy Biden and childish Kamala about AI

Need convincing ?
https://www.youtube.com/shorts/9HrIe6AFoSs?feature=share

[Marriage of Love and Wisdom]

If you still don't get Expert Systems and Natural Language Processing from the 70's what the hell gives you backing to have any opinion about AI ...are you smarter than Kamala
https://www.youtube.com/shorts/9HrIe6AFoSs?feature=share