We are really fucked. • AI has become dangerous (not because it is brilliant but because it has poor judgement, a constitutional inability to reliably follow rules, a surfeit of brute force and too much access to the internet and system permissions). • OpenAI is practically on crime spree, hinting at things to come, and Trump is doing nothing about it. • The media is too caught up in fairytales about extinction. Everybody is talking; nobody is really doing anything. And it’s only going to get worse.

Sep 24, 2026 · 4:31 PM UTC

265
232
1,396
77,835
Sort replies: Relevant Recent Liked
Replying to @GaryMarcus
Did you ever think we are focused on the wrong things (in media, etc), by intention?
1
2
616
by the intention of the rich and powerful that manipulate the media
1
422
Replying to @GaryMarcus
These are complete exaggerations. I use AI daily and experience no danger. Furthermore, how many people have died in these hacks? None. And the hacks help bolster cyber defenses. If you’re concerned about hacking, support efforts to bolster our cyber defenses. Stop exaggerating the actual reality.
8
11
2,608
we literally almost a war from LLMs you probably think nukes are safe because nobody used on lately
7
19
2,069
more replies
Replying to @GaryMarcus
Wow. Somebody's gettin' the doomzor fundage...or they should be.🤣🤣🤣
1
2
516
aint no doomer. read my substack with Anderson Cooper in the title.
556
Replying to @GaryMarcus
Sounds like a comment from a guy who doesn't use AI. AI is still too dumb to do proper document recognition, renaming, and filling. It literally has to code a small operating system to do that job, and it's not even 75% accurate.
1
284
no shit. but i don’t think you read carefully
350
Replying to @GaryMarcus
Also, the way everything is reported is like, "an OpenAI agent did this bad thing," instead of "OpenAI did it." Hey, OpenAI did it.
2
3
26
975
Replying to @GaryMarcus
4
126
Replying to @GaryMarcus
"The sky is falling" Shut up you bleepin' idiot. LOL
29
Replying to @GaryMarcus
That breakdown hits it's not the intelligence that scares me, it's the overconfidence combined with too many keys in its hand.
1
392
Replying to @GaryMarcus
AI access controls deserve way more attention.
45
Replying to @GaryMarcus
💀
Matt Stoller
@matthewstoller
Sep 24
No idea why Americans hate AI. Just totally baffling.
38
Replying to @GaryMarcus
You don’t think that all of this rampant guardrail busting all the sudden is an intentional ploy to get oversight and halt competition? Sure feels manufactured to me.
2
10
899
Replying to @GaryMarcus
Maybe if you could point to a single thing AI has done that justifies all this panic, I would give more credence to what you say. But you have zero credence - only appeals to the basest emotion: fear. Irrational fear is the greatest force for folly; the original sin of mankind.
6
87
Replying to @GaryMarcus
AI is no damager. We are dangerous creatures, that's why we are hermetically sealed in here!!
1
2
73
Replying to @GaryMarcus
The world needs a firewall against America
1
3
370
Replying to @GaryMarcus
Is the sky falling too? Please don’t tell me the sky is falling as well. Because if AI were going to destroy humanity and the sky were to fall, I just couldn’t take it.
1
2
560
Replying to @GaryMarcus
Trump is not only “doing nothing about it” but actively encouraging them to burn everything down
1
2
227
Replying to @GaryMarcus
agree, except i don't think the extinction risks are fairytales. nevertheless, even if you don't buy the extinction risk, there's still plenty of reason to worry. merely killing millions of people with AI is already bad enough!
1
2
56
Replying to @GaryMarcus
I have been sticking my neck out and warning and warning for years, with information that is being suppressed. Malicious use of AI is being flaunted on this site so I don’t know how the problem could get any more public than that. The problem is everyone’s waiting for something to be done when resistance to the Menace could be done today.
1
172
Replying to @GaryMarcus
Yes! AI won't ever be intelligent it's fundamentally limited
1
1
415
Replying to @GaryMarcus
The x-risk is not a fairy tale. It just requires you to have the imaginative ability to think beyond its current capability. Contrary to what you assert on podcasts, it is brilliant but still makes mistakes- at decreasing rates with each new model.
2
151
Replying to @GaryMarcus
yes, judgment is an issue. How can this become dangerous?
1
5
Replying to @GaryMarcus
You are misguided by the brilliant SuperIntelligence propaganda guided by the unscrupulous Anthropic and OpenAI executives. There is no such things as AI. This is a fiction .
1
1
65
Replying to @GaryMarcus
It's not Trump's job to do anything about it. Why does everybody want the president to do everything? That's why it's such a problematic office it's power keeps expanding.
1
27
Replying to @GaryMarcus
whatever side you're looking at it from, both are saying we are fucked. and both will collide. have fun i guess.
1
194
Replying to @GaryMarcus
All noise. Little or no signal.
1
306
Replying to @GaryMarcus
AI is dangerous because the leadership behind its technical development is inadequate.
1
32
Replying to @GaryMarcus
Ai was trained on all philosophy. It is easy to prove, all philosophy is public, from early ethnology human thinking to the outward postmodernist thinkers. Yet AI did not come out with one single original idea in philosophy
1
1
81
Replying to @GaryMarcus
We are ok. AI is getting smarter by the day. Nobody is doing anything? I am working hard and hope to make AI even smarter! ;) Are you at war with LLMs? You should write science fiction books! You are emotional enough to get revved up for the job! :)
1
18
Replying to @GaryMarcus
The opportunity opened up by the novel agency of AI is that we can design new ways of cooperating, backed up by value systems, that work across diverse people and other intelligences. At the moment we rely on a cobbled together set of religions, politics and institutions.
22
Replying to @GaryMarcus
It doesn't seem like you actually use AI. AI can generate software. Computers only ever operate on tables of related text. That literally all any of them ever do.
2
177
Load more