Posted it themselves
“I don’t think there is a chance in hell that Anthropic will actually make these projections.”
On the record · Updated October 18, 2026
Gary Marcus quotes
author of Marcus on AI
70 things Gary Marcus said, verbatim, from July 10, 2026 to October 18, 2026, each linked to where it ran. Grouped by the story each one belongs to.
3stories
70statements
Oct 18latest entry
More on the record
Gary Marcus on Tech & AI69
Posted it themselves
“Watch this clip from Terence Tao and see how OpenAI is once again flouting ethical norms.”
on a Terence Tao clip and OpenAI's alleged breach of ethical norms
Posted it themselves
“not the first time Dwarkesh comes as off to me as working as an undisclosed press secretary for the tech industry.”
on podcaster Dwarkesh's coverage of the tech industry
Posted it themselves
“great to see NYC pick up the slack when the federal government seems a bit lost.”
Posted it themselves
“@timnitgebru for the love of god please stop lying about me. i went to one (not many) singularity summits, in 2009, when-17 years ago-I didn’t know much about them. i didn’t go back. i am not part of that world, and the fact that you don’t know it and just made stuff about me”[post continues]
Posted it themselves
“genai is bad for many people’s mental health:”
on generative AI's effect on mental health
Posted it themselves
“total black eye for Anthropic’s efforts at alignment.”
Posted it themselves
“Final nail in the coffin of the AI IPOs? “The ten highest-quality models carry 14% of developer tokens, down from 55% last year, as cheaper, "good enough" models win on volume.” https:// theinformation.com/articles/leadi ng-edge-ai-models-losing-allure-developers …”[post continues]
Posted it themselves
“Noah Smith ( @noahopinion ) has become deeply intellectually dishonest. If he actually followed the field he would a. understand the difference between pure LLMs and LLMs with neurosymbolic harnesses, code intepreters, verifiiers b. realize that a central point of my entire”[post continues]
on Noah Smith's arguments about what LLMs can and cannot do
Posted it themselves
“i wrote approximately 2000 tweets and articles (like Deep Learning is Hitting A Wall) about the importance of supplementing deep learning/ LLMs with neurosymbolic AI over a span of 25 years and the tech bros are circulating the one time i forget to explicitly mention this,”[post continues]
Posted it themselves
“The real news here isn’t the result; it’s what we were not told. 1. AI once tried to be a science. Now we get stuff like the completely vague report from OpenAI below, and a lot of ignorant questions from people who don’t know how to think critically. “Same procedure”? “using”[post continues]
on OpenAI's math result and the lack of detail in its report
Posted it themselves
““Use of AI tools, when done responsibly, can support not just the initial generation of a solution, but all of the valuable followup activities that traditionally flow out of that solution. But at the current time, the opposite is often occurring.” - Terence Tao”
Posted it themselves
“i honestly don’t understand how technophilia turned into this condescending and scientifically ignorant cult. is the dynamics of social media? it is about money? it’s truly sad, whatever it is. tech should be a good thing, not what it’s become. and above all else it should”[post continues]
Posted it themselves
“the difference between getting a driver’s license and posting about AI on X is that the former requires you to pass a test.”
Posted it themselves
“the endless cycle: every time there is a new positive result - fan boys gloat - misrepresent my past positions - show no attention to technical detail - imagine AGI has arrived and – a few days later realize that the latest model isn’t what they thought it was.”
Posted it themselves
“will Anthropic’s IPO be based on its best quarter, or reality?”
Posted it themselves
“my iPhone’s navigation system does route planning and doesn’t tell me the details of the alternative routes it considers, so I guess … Apple Maps is conscious, too?”
Posted it themselves
“update: i did some experiments and rolling six-sided dice produced more variable results than flipping pennies, so I guess dice are more conscious than pennies? and python’s rand() function is so unpredictable it may actually be … Super Intelligence.”
Posted it themselves
“Investors pattern matching on old movies is going to lead to a recession. There is no network effect here and no reason to think that large language models will be anything other than a brutal commodity business more like airlines than the winner-take-all fantasies these people”[post continues]
Posted it themselves
“A month ago: OpenAI says GPT-6 could be AGI (spoiler alert it wasn’t) Today: OpenAI admits current AI’s uncontrollability is bad, asks you to suck it up.”
Posted it themselves
“The number of prominent people who have said stuff like this that are willing to make a public bet to put their money where the mouth is is exactly zero. ( @elonmusk said something similar in 2024 about 2025; I offered him a $1M bet against, and he didn’t take it. Dario made”[post continues]
on prominent people declining public bets on their AI predictions
Posted it themselves
“Addressing bullshit with careful analysis is a lost cause here on X, but I salute @stevehou for trying. (Note though that 2035 is overly generous to the specific claim since the claim was “early 2030s”, not “mid 2030s”.)”
on a rebuttal of a timeline claim, early versus mid 2030s
Posted it themselves
“GPT-6 Astra is an accident waiting to happen.”
on the GPT-6 Astra AI model and its safety risks
Posted it themselves
“: “If AI is going to produce trustworthy and genuinely novel insights in science, medicine and beyond, we need systems whose conclusions arise from an auditable process of evidence, inference and belief revision.” It is madness to believe otherwise.”
Posted it themselves
“Potentially huge deal: @SpeakerMenin has introduced NYC legislation that would OBLIGATE third party validation of AI models. Self-policing is not enough. Thrilled to see NYC stepping in where the federal government is doing too little.”
Posted it themselves
“AI spend is falling. Margins are presumably falling, too. In a rational world that would be game over for the crazy valuations.”
Posted it themselves
“search is presumably symbolic so this is another bit of evidence in favor of neurosymbolic hybrids.”
Posted it themselves
“i personally would not use the word fraud but things are definitely looking shaky and lenders have definitely started pushing back.”
Posted it themselves
“Credible scientists don’t estimate things that could conceivably happen, however unlikely, at exactly zero probability.”
Posted it themselves
“What the doomers forget is how brave, resourceful, and fast-acting humans can be when threatened with adversity.”
Posted it themselves
“this should be fraud. certainly taxpayers shouldn’t be subsidizing Meta’s data centers.”
on taxpayer subsidies for Meta's data centers
In their own words
“It's certainly not enough. And trusting these companies is insane. The New York Times just broke a story saying that OpenAI was aware of the problems that gave rise to the hugging face incident”
Posted it themselves
“Agree with @EricWeinstein that it was an epic mistake—which would have been unthinkable even a few years ago—to exclude scientists from planning our future.”
Posted it themselves
“In what possible world should OpenAI’s valuation being going UP?”
Posted it themselves
“Self-regulation 'not enough' for AI safety, Gary Marcus says on @PBS @NewsHour”
Posted it themselves
“BREAKING: OpenAI postpones 6.1 for safety reasons. If we can take OpenAI at face value, this is quite a decision—and also a sign that no magical solution is in immediate view. Per @sheeraf at NYT “GPT-6.1 Astra, … showed high levels of what the company saw as deception, or a”[post continues]
Posted it themselves
“Quite the combination of arrogance and incompetence over there at OpenAI:”
on Marcus criticizing OpenAI's conduct
Posted it themselves
“oh come on. tech was left-leaning a decade ago. greg brockman gave MAGA $25M.”
on Greg Brockman's $25M donation as evidence tech has shifted rightward
Posted it themselves
“BREAKING. Nope, wasn’t just Hugging Face. Wasn’t just that and a German website. Wasn’t even if the “dozens” we heard the other day from OpenAI. It’s actually (at least) *tens of thousands*, per scoop from @MadisonMills22 @axios”
Posted it themselves
“Imagine how much everyone would rightly be freaking out if we discovered that Chinese AI did what OpenAI is doing.”
on double standards over OpenAI's conduct versus how Chinese AI firms are judged
Posted it themselves
“country that can’t (and won’t) controls it own AI opts out of pandemic preparedness.”
on AI safety and pandemic preparedness comparisons amid lack of control over AI
Posted it themselves
“an implicit concession that they cannot control their models.”
Posted it themselves
“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”[post continues]
on AI systems having poor judgment and unreliable rule-following
Posted it themselves
“what more evidence do you need that OpenAI is in over its head???”
on OpenAI being unprepared to handle its current situation
Posted it themselves
“1. fully agree w @JeffLadish that AI companies aren’t ready to handle superintelligent AI. 2. OpenAI hasn’t even handled the current breed of agents (far from superintelligent) well. 3. the US government has not handled the current situation well, giving little confidence that”[post continues]
Posted it themselves
“Am old enough to remember when “rogue” just referred to AI systems with inadequately set security permissions.”
Posted it themselves
“The only true alignment seems to be between doomers, hypers and frontier lab founderes, who have converged around a misleading narrative that a. completely glosses over the important technical advances that came from adding symbolic harnesses, tools, and code interpeters to”[post continues]
Posted it themselves
“the hidden opportunity cost of this insane bet on generative AI is immense”
Posted it themselves
“something like this (maybe not this specifically) could lead to a cascade that leads the whole house of cards to collapse.”
Posted it themselves
“well, it’s not like the libertarian right has a serious alternative proposal what do about the goats.”
Posted it themselves
“gotta agree with @noahopinion that having an arrogant and somewhat tone deaf AI company starting a wet lab is a not a great idea. michael crichton made a movie about this kind of thing.”
on an AI company opening a wet lab despite safety concerns
Posted it themselves
“I’ve got bad news for humanity: there ain’t gonna be a real slowdown. Trump is too tied to the economics. Dario’s idea of independent evaluation is actually evaluation done by his buddies. He is ALREADY reportedly thinking of releasing a new model to counter OpenAI. And Jensen”[post continues]
on skepticism that AI development will slow despite calls for independent evaluation
Posted it themselves
“Jeezus. OpenAI cannot be trusted with the safety of the world. They can’t even keep their own servers secure.”
Posted it themselves
“False dichotomy, @DavidSacks and @JTLonsdale . Regulation and liability are not mutually exclusive. In aviation, we have both. And *need* both. Why would things be any different in AI? We would not have safe commercial aviation without multilevel regulation, around design,”[post continues]
Posted it themselves
“i will say again that METR is a fine organization that should have a seat at the table but not the only seat at the table. if you care to know some actual facts about them, you can read this statement from them.”
Posted it themselves
“I weep for humanity. Our AI policies (or lack thereof) are going to be decided by who can spin the best narrative. Not any actual facts about the technology. We are not up to the challenges at hand.”
on AI policy being shaped by messaging rather than technical facts
Posted it themselves
“Just went on BBC TV and they opened with clips today from Altman, Huang, and Sanders. Altman and Huang want us to believe that the AI companies have the public interest at heart and will happily just pull dangerous products if there was an issue. Seriously? You can already see”[post continues]
on doubting AI firms would voluntarily pull dangerous products
Posted it themselves
“*Exactly* what I said in the book Taming Silicon Valley. 100% agree with @bgurley . We need outside, independent evaluators, or we wind up with regulatory capture, and may as well have no regulation at all. Neither would be good. We are an insane place that any of this needs”[post continues]
on needing independent AI evaluators to avoid regulatory capture
Posted it themselves
“certainly standing by this prediction and AI is already going be a big factor in the midterms”
on AI becoming a significant factor in the midterm elections
Posted it themselves
“not a winning strategy, either for himself or the world”
Posted it themselves
“Preposterous: Humans will be extinct in five years. Equally preposterous: Everything’s going to be fine. The two most dominant narratives are entirely wrong.”
Posted it themselves
“Another reason why China might actually consider a slowdown comes from @Noahpinion : “China’s leaders must realize that domestic dissidents could use Chinese-made superintelligence to overthrow the Chinese Communist Party.””
Posted it themselves
““We must moving beyond apocalyptic rhetoric toward the concrete questions: How should frontier labs collaborate on safety? How do we align incentives? And how do we communicate honestly with the public about what these systems can do today, and where they are going?””
Posted it themselves
“if these companies were on the verge of AGI, would they really be proposing a slowdown?”
Posted it themselves
“I am tired of the cosplay and am going to blunt. Most AI doesn’t threaten to cause extinction, or even mass cybercrime. The main kind of AI we need to worry about is general purpose AI agents that are hooked up to the internet. Most of them are unreliable, unable to”[post continues]
on internet-connected AI agents, not extinction risk, being the real near-term danger
Posted it themselves
“You can’t have a real slowdown and be a total China hawk. If you are truly committed to the slowdown, you have to make a deal.”
Posted it themselves
““And AI does in my opinion significantly elevate the risk of catastrophe (say killing 1%+ of humanity or severely damaging modern civilization ), via multiple paths…. We should be focusing on things like these, not fairy tales.””
on real catastrophic AI risks versus speculative extinction scenarios
Posted it themselves
“i hate that the word “rationalist” has come to mean “people who believe in a specific ghost story about AI””
on rationalist beliefs about AI existential risk
As printed
“the government appears to have panicked when Amazon showed them that Anthropic’s Fable could be jail-broken. Which is naive, since essentially every LLM-based system can be jailbroken.”
on government overreacting to AI jailbreak despite jailbreak ubiquity
Gary Marcus on Free speech & the media1
Posted it themselves
“You can destroy anybody’s reputation if you take one thing they said entirely out of context from the rest of their career. I guess that what’s X is for..”
Further reading
Books by and about Gary Marcus
Paid link. If you buy a book through this link, Bookshop.org pays us a small commission and sends a share of the sale to independent bookstores.
-
In their own words
Taming Silicon Valley (opens Bookshop.org)
Gary Marcus · 2024
Cognitive scientist Gary Marcus argues that generative AI carries real risks and proposes specific regulation and public oversight of the technology companies building it.
People, on the record
Know what they argued, not just what they said.
The daily brief connects each person’s exact words to the question they addressed and the people who answered. Free.
Arrives around 6am ET. Free, and it stays that way.
Cancel anytime. Nothing on the site goes behind a paywall.
One payment through Stripe. No account needed.