America, and probably the world, stands on a precipice
garymarcus.substack.com
garymarcus.substack.com
We already did it with companies, buddy!
I guess your point could be: LLMs are just another level of capitalistic opacity to maximize opacity and dilution of accountability.
A computer that values the life of a Israeli the same as that of a Palestinian... Ah a man can dream.
The IDF used tools like Lavender and "Is Daddy Home" to analyze communications, identify members of Hamas, and learn when they were home so they could be killed with bomb strikes.
This has long been possible for humans to do, but it's a laborious process. In the past, only people high up in chains of command received such bespoke treatment. AI tools permitted the IDF to grant the same treatment to raw recruits who had been given the sum total of a pep-talk and a pistol.
The result was widespread destruction and indiscriminate killing of civilians. The IDF didn't spend much time scrutinizing AI recommendations and were willing to act on false positives. Every bomb strike, by design (i.e. "Is Daddy home"'s purpose was to determine when targets were in their family homes), took out civilians. Just taking a pizza order from a Hamas member years before the war might have been enough to get entire families and their neighbours killed.
If humans hate another group of humans enough and an AI says "Kill", they'll kill. Without thought or remorse. We don't merely need to be worried about murderous robots on battlefields, we also need to worry about humans implementing the recommendations of AI without thinking for themselves.
Do you have the same sentiment about self-driving cars?
SEC. 1638. Sense of Congress with respect to use of artificial intelligence to support strategic deterrence.
(a) Sense of Congress.--It is the sense of Congress that--
(1) the considered use of artificial intelligence and machine learning tools presents opportunities to strengthen the security of critical strategic communications and early warning networks, improve the efficiency of planning processes to reduce the risk of collateral damage, and enhance U.S. capabilities for modeling weapons functionality in support of stockpile stewardship; and
(2) even with such applications, particular care must be taken to ensure that the incorporation of artificial intelligence and machine learning tools does not increase the risk that our Nation's most critical strategic assets can be compromised.
(b) Statement of policy.--It is the policy of the United States that the use of artificial intelligence efforts should not compromise the integrity of nuclear safeguards, whether through the functionality of weapons systems, the validation of communications from command authorities, or the principle of requiring positive human actions in execution of decisions by the President with respect to the employment of nuclear weapons.
It’s interesting that many of us myself included once thought that the butlerian jihad was silly until now. Frank Herbert wrote something that is particularly prescient.
(Usually writers are just a decade ahead of their time. Whatever Podcasters are talking about today, has usually already been discussed in literature a decade ago. Prediction markets come to mind. Socially, over vs under population as discussed in popular books like the rationale optimist or the accidental superpower.)
Curiously enough the Dune books are a strong rebuke against prescience (read: humanity coveting precog abilities.) I'll leave it at that to avoid spoilering.
Keep an eye on the next movie this year!
In fact they are somewhat similar to the reasons why the Melnibonéan empire fell in the Michael Moorcock Elric universe: people got lazy, spent their time drugged out of their minds, and cruelty seemed to be one of the few things to get a rise out of them. In Dune labour was delegated to (thinking) machines, in Elric it was delegated to slaves. Eventually such a society will collapse or be conquered.
So I don't really think this spat proves anything at all.
Honestly trying to throw in the article that AI may be responsible for nuclear weapons also throws this whole article into junk opinion tier. Just trying to riot up support.
Though it is funny another commenter already found they passed a law where AI can't launch nuclear weapons last year.
After they use violence to discipline their own side, everything gets much more serious. But it's also a point of risk for them, because it might break their coalition. So it's important to defend the rights even of those who are on the side of autocracy . Because if they are successful in establishing violence as their right over even their own allies and followers, there may be no way back without civil war.
At first jail, where they release them after a day or two, then in warehouses converted into jail-hospitals, where people stay weeks, get treated, sometime lose teeth. Almost always syndicalists and KPD supporters, sometimes SPD local leaders, it was often political?
Oh sorry, you talked about the US?
I would say, when the administration openly refuses to follow a court decision. (And in practice, it's going to have to be a Supreme Court decision, because the administration seems to appeal every lower court decision that is against them.) Or, when they openly and blatantly interfere in the 2026 election (either by cancelling it via martial law or state of emergency, or interfere in it via troops or ICE in the streets or state of emergency, or after the fact refusing to accept the results because of some claim of invalidity).
If someone like that wants to be in a position to warn society of actual harms, they'd have to behave differently.
And the corporations won't fight this. They're in it for the money and they're willing to bring actual gold bars to White House to ensure it keeps rolling in. They know what they're doing is corrosive and debasing, the more conscientious of them probably want to vomit on the inside. But they mostly suck it up and do it anyway, for their investors will discipline them if they don't.
Either people run candidates and vote for the ones that campaign on stopping this, or it happens.
Worse: they'll be sacked and replaced with someone who will.
Like Trump's FCC chair was saying he'll revoke the license of stations that make republicans look bad. Those stations will then be replaced with more copies of Newsmax. CBS either toes the line or it gets shut down and replaced by a station that will.
The idea that somehow the current actions are 'real' history and what people were doing before is fake just feeds the claim of inevitabiility, a basic psyops maneuver - you can't win; our victory is inevitable.
People have made history for centuries of Enlightenment - the whole idea is that we can control our fates as individuals through reason and compassion (humanism), and we have done it. We have transformed the world. The only problem is people giving up - despite the incredible success of this idea over centuries - and accepting that they can't control their fate. Certainly MAGA-ish conservatives believe they can make history.
The "return of history" is snarkily pointing out how historical trends have been reasserting themselves, and Fukuyama (the author) was, at best, overly optimistic
> historical trends have been reasserting themselves
Western liberalism is an historical trend, just like the others.
I mean, I agree with you, but if you've only realized that now, you've missed out on some really weird stuff going on for the past couple of years.
I'm really surprised Substack thinks Australia's social media laws apply to them.
(And no, I'm not willing to do that just to read an article.)
Marcus on AI
America, and probably the world, stands on a precipice. Call your Senators and Representatives, right now. Gary Marcus Feb 26, 2026 As I wrote here yesterday, Anthropic’s showdown with The US Department of War may literally be life or death for all of us. If Pete Hegseth forces Dario Amodei to fold, not one but two monstrous precedents will be set.
The first is obvious, and terrible in itself. The second is subtle but no less important.
The first is that what Secretary Hegseth is demanding, backed by heavy threats, is that the US military have full, unrestricted access to Anthropic’s AI software, for applications such as military surveillance and autonomous weapons without humans in the loop. This could well extend to nuclear weapons.1
Nothing that I have read convinces me that Secretary Hegseth has a nuanced understanding of the strengths and limits of current AI, or that he will show restraint in how he applies it. Rather he is trying to define his career in part around deploying AI as broadly and as quickly as possible.
The second is that Hegseth’s maneuver is an audacious power grabs that aims to circumvent Congress. By setting a deadline of 5:01 PM eastern tomorrow, Hegseth aims to cut everybody else – even Congress — out of the loop.
A reader of this newsletter, a tech writer who describes himself as a political independent just wrote to me, rightfully panicked:
Today the Pentagon will force Anthropic to change their corporate goal of responsible AI. This is not something to be decided in the marketplace by a bully with deep pockets; it must be decided in Congress. Senators and Congressmen must take a position and deliberate in public about whether it is OK to use AI for surveillance of Americans and to launch lethal strikes controlled by AI with no “human in the loop”. Please say something today, before Amodei has to surrender.
He is right.
Please call or write your Senators and Representatives right now.
§
AI policy, especially of this magnitude, is something that American people should have a say in. Congress should deliberate. Mass surveillance and AI-fueled weapons, possibly nuclear, without humans in the loop are categorically not things that one individual, even one in the Cabinet, should be allowed to decide at a gunpoint.
But that is exactly where we are headed.
1 Hegsseth’s demand would in principle extend to apply Anthropic’s software to nuclear weapons without humans in the loop. In that connection, people should probably be aware of the fact that S. 1394 - Block Nuclear Launch by Autonomous Artificial Intelligence Act of 2023 failed to pass. Which means we might not have a Stanislav Petrov next time around.
Discussion about this post Write a comment... Richard Self 7h
Given the unreliability of GenAI in everything that it do, the use in unsupervised warfare will be catastrophic.
Reply Share 1 reply Roman's Attic 7h
From what I’ve heard, these calls are especially valuable if you emphasize that this is an issue that will determine how you vote in the future
Reply Share 1 reply 58 more comments...
No posts
Ready for more? Type your email... Subscribe © 2026 Gary Marcus · Privacy ∙ Terms ∙ Collection notice Start your Substack Get the app Substack is the home for great culture This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts
https://web.archive.org/web/20260226214404/https://garymarcu...
I think this should just go work. Hope this helps.
Edit: this is my first time using archive.org for susbtack and looks like it doesn't work/ shows page not found after some time.
I have now used singlefile to get the index.html and put it to github pages for anyone to see.
https://serjaimelannister.github.io/american-article/
And I have also created an archive of the GH page itself, here's the link to archive.org itself as well which should work
https://web.archive.org/web/20260226235745/https://serjaimel...
Why would they not?
The law restricted accounts. Not reading. But this was not what you said surprised you.
This administration is criminal and has repeatedly shown disdain for the fundamental rights enshrined in the constitution. Our democratic system is only as strong as the will of the people to protect it.
The call to action at the end of the article is good. I think it's worth calling your senators and house reps to speak up about this. Politicians in general are slimy and spineless, those seem to be traits required to succeed in politics, but as a consequence they do tend to notice when people really speak up and care about issues. As a survival instinct if nothing else.
If you happen to be part of this cohort, you should absolutely always question the moral implications of any project you work on, and you're in a better position than the average person to fully explore the possible ways tech can be used and abused. I'd argue that should be considered one of the primary purposes of Hacker News.
And I assume the only reason this thread hasn't been flagged like every other "political" thread is that the current top comment references Dune.
2. The hero takes action to avoid it.
3. Those very actions are what make the prophecy come true.
Peter Thiel has all these paranoid delusions about the future. (Dude, I know you are listening, if you are scraping my Hacker News posts we seriously need to have a conversation about this!) Yet, his actions trying to prevent this future he is terrified of are the very things that will bring it to fruition. Hamlet is correct -- we defy augury.
Have a nice day.
It's capitalism.
Capitalism is a functioning AI that controls the world, that has had humans serving it. AI is the tool that capitalism will use to remove humans from the equation.
The end state of capitalism is slave labor. The end state of technocapitalism removes humans from the labor equation.
Don't get suckered by his takes on things; he's nothing but a credentialed blowhard pimping his pedigree for clout and cash. He's been furiously slinging arrows from the peanut gallery for years and it's gross.