Gary Marcus(@GaryMarcus)

Complementary remarks from Gary Marcus and Terence Tao on OpenAI’s giant math drop Making sense of ...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 17:22:15 GMT

Complementary remarks from Gary Marcus and Terence Tao on OpenAI’s giant math drop

Making sense of a new regime in math

garymarcus.substack.com/p/complementar…

💬21🔄23❤️119👀33827📊51

⚡ Powered by xgo.ing

The real news here isn’t the result; it’s what we were not told. 1. AI once tried to be a science. ...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 16:12:40 GMT

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 an unreleased model”?

This would never pass peer review.

We don’t know what the procedure was.

We know nothing about the architecture (e.g., were proofs generated in one shot, and then verified by Lean? was there an iterative process?). 

We know nothing about the failure rate. We know nothing about the training/post training/data agumentation.

2. As a result we have zero idea of how generalizable the result is outside math.  

3. A lot of X has been reduced to an ignorant cheering section that applauds without knowing what it is applauding or what it might mean — without ever asking basic scientific questions.

The new system could be a legitimate step towards AGI or just a clever leveraging of Lean and synthetic data in a verifiable domain with no generality whatsoever; from this report we can tell almost nothing.

💬41🔄45❤️356👀21928📊86

⚡ Powered by xgo.ing

“Use of AI tools, when done responsibly, can support not just the initial generation of a solution, ...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 15:45:47 GMT

“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

💬26🔄64❤️338👀18975📊77

⚡ Powered by xgo.ing

terence tao, shitposting🤣

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 15:43:30 GMT

terence tao, shitposting🤣

💬34🔄108❤️1276👀93563📊204

⚡ Powered by xgo.ing

what’s the best argument against this?

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 03:54:34 GMT

what’s the best argument against this?

David Krueger 🦥 ⏸️ ⏹️ ⏪@DavidSKrueger

Immediate, indefinite, international moratorium on frontier AI development.

🔗 View Quoted Tweet

💬95🔄5❤️107👀15931📊71

⚡ Powered by xgo.ing

i honestly don’t understand how technophilia turned into this condescending and scientifically ignor...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 03:51:42 GMT

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 cherish science and not run roughshod over it.

💬66🔄53❤️460👀18350📊110

⚡ Powered by xgo.ing

an actually intelligent question! i give my answer below it.

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 03:08:10 GMT

an actually intelligent question!

i give my answer below it.

Kenneth Nelson@knj522

@GaryMarcus Agree people are mischaracterizing your position. 

The question I still have would be: on your accounting of LLM capabilities, would a frontier LLM be able to solve an open math problem without a lean / code / symbolic harness?

I think you're suggesting no, right?

🔗 View Quoted Tweet

💬4🔄1❤️31👀14674📊7

⚡ Powered by xgo.ing

why do i have to deal with stuff when the empirical record is literally the opposite? [ see eg http...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 03:06:58 GMT

why do i have to deal with stuff when the empirical record is literally the opposite?  [ see eg x.com/garymarcus/sta…]

Bicycle@BicycleSocials

@GaryMarcus@MelMitchell1 You have the least credible track record of anyone in AI.

🔗 View Quoted Tweet

💬5🔄0❤️17👀8036📊5

⚡ Powered by xgo.ing

the difference between getting a driver’s license and posting about AI on X is that the former requi...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:59:54 GMT

the difference between getting a driver’s license and posting about AI on X is that the former requires you to pass a test.

💬22🔄7❤️110👀8209📊29

⚡ Powered by xgo.ing

the endless cycle: every time there is a new positive result - fan boys gloat - misrepresent my pa...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:59:08 GMT

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.

💬5🔄9❤️92👀8045📊16

⚡ Powered by xgo.ing

yep, this, too, like a bunch of other similar tweets, is typical of the cluelessness of people who ...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:56:34 GMT

yep, this, too,  like a bunch of other similar tweets, is typical of the cluelessness of people who literarily don’t understand the difference between
a. a neural network solving a problem by itself
and
b. a neural network proposing an answer and having a separate symbolic system verify it.

if that’s over your head, you probably shouldn’t be commenting about AI.

wh@nrehiew_

🔗 View Quoted Tweet

💬28🔄15❤️181👀32075📊40

⚡ Powered by xgo.ing

my work is done.

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:23:19 GMT

my work is done.

The Tennessee Holler@TheTNHoller

JON STEWART: “So in the FUTURE we WILL use A.I. to cure diseases… but right now it’s mostly murder, surveillance, and surge pricing for the McRib — What a trade! I’m getting the sense you guys might be full of shit.”

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬26🔄163❤️1477👀83180📊176

⚡ Powered by xgo.ing

this is typical of the cluelessness of people who literarily don’t understand the difference between...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:15:27 GMT

this is typical of the cluelessness of people who literarily don’t understand the difference between
a. a neural network solving a problem by itself
and
b. a neural network proposing an answer and having a separate symbolic system verify it.

if that’s over your head, you probably shouldn’t be commenting about AI.

andrew arruda@andrewarruda

🔗 View Quoted Tweet

💬9🔄4❤️86👀14382📊15

⚡ Powered by xgo.ing

do you lazy morons realize that the phrase “stochastic parrot” isn’t mine? it’s worse than a strawm...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 02:12:44 GMT

do you lazy morons realize that the phrase “stochastic parrot” isn’t mine?

it’s worse than a strawman to accuse me of something I didn’t coin or endorse.

but y’all don’t really seem to care about what i actually said. 🤦‍♂️

💬20🔄4❤️80👀9838📊25

⚡ Powered by xgo.ing

since many people are asking me about the new math results from OpenAI, here is my take. 1. It uses...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 01:29:06 GMT

since many people are asking me about the new math results from OpenAI, here is my take.

1. It uses symbolic AI (Lean, etc) in addition to LLMs which is what i said for years we would need to do; this this confirms what I actually said (as opposed to various fictitious misrepresentations that run rampant around here )

2. It still doesn’t mean they have AGI; the math stuff AFAIK does not generalize broadly, because you can’t use the same symbolic verification (and symbolic data augmentation techniques) in the open-ended real world.

So yes I do feel vindicated wrt to neurosymbolic AI but no the G in AGI (general) is not yet solved.

What I said in 2019 remains true: neurosymbolic AI is necessary but not sufficient; world models are also key.

💬94🔄28❤️379👀89696📊144

⚡ Powered by xgo.ing

by failing to understand nuance (and basic technical point about architecture) the tech bros have ma...

Gary Marcus (@GaryMarcus) — Wed, 07 Oct 2026 01:26:51 GMT

by failing to understand nuance (and basic technical point about architecture) the tech bros have managed to get things exactly backwards once again. in their usual condescending but misguided way.

💬2🔄6❤️76👀6619📊10

⚡ Powered by xgo.ing

this was awesome

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 22:33:36 GMT

this was awesome

Ed Zitron@edzitron

Joined @GaryMarcus@JG_Nuke@gnoble79 and Julien for an hour and a half long conversation about a little thing called "The AI Bubble"
youtube.com/watch?v=es4FfR…

🔗 View Quoted Tweet

💬4🔄8❤️60👀7195📊10

⚡ Powered by xgo.ing

It’s not a communications issue; it’s a greed issue. we have to stop taking everything this man say...

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 20:12:19 GMT

It’s not a communications issue; it’s a greed issue.

we have to stop taking everything this man says so seriously.

Jacob Coxon@hilbertspaess

In this interview I tried to argue that a post-scarcity society will benefit working people. It was largely a rhetorical failure. People don’t trust that the benefits will be distributed, and tech optimists have to fix this communications issue.

🔗 View Quoted Tweet

💬21🔄27❤️191👀9224📊33

⚡ Powered by xgo.ing

Also, who gets to decide?

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 17:22:51 GMT

Also, who gets to decide?

Gary Marcus@GaryMarcus

Is a “nice tool” worth a 10% risk of catastrophe?

🔗 View Quoted Tweet

💬9🔄9❤️52👀5142📊12

⚡ Powered by xgo.ing

Is a “nice tool” worth a 10% risk of catastrophe?

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 17:11:02 GMT

Is a “nice tool” worth a 10% risk of catastrophe?

Mark C. Massey@MarkCMassey

I have created some interesting things with Claude.

Nice tool.

If it went away, life would return to normal. 

Not essential.

🔗 View Quoted Tweet

💬6🔄5❤️65👀9221📊10

⚡ Powered by xgo.ing

@HarperSCarroll April 19, above British Columbia, iPhone 16 Pro:

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 16:56:22 GMT

@HarperSCarroll April 19, above British Columbia, iPhone 16 Pro:

💬4🔄3❤️82👀4844📊11

⚡ Powered by xgo.ing

As usual, my dark predictions have materialized. 😢 https://t.co/dIAJ4syuWU

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 16:53:09 GMT

As usual, my dark predictions have materialized. 😢
nytimes.com/2026/10/06/wor…U

💬13🔄26❤️74👀15559📊27

⚡ Powered by xgo.ing

will Anthropic’s IPO be based on its best quarter, or reality?

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 04:51:42 GMT

will Anthropic’s IPO be based on its best quarter, or reality?

Rohan Paul@rohanpaul_ai

Microsoft has cut internal Claude spending by over a third, and Meta's Claude Code users have halved.

- per The Information

"Microsoft was on pace to spend at least $2 billion annually on Anthropic models to power its Copilot AI features for customers of its productivity software. It has similarly swapped out some of its own models for Claude in Copilot"

🔗 View Quoted Tweet

💬15🔄27❤️169👀10040📊28

⚡ Powered by xgo.ing

will Anthropic’s IPO be best on its best quarter, or reality?

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 04:39:49 GMT

will Anthropic’s IPO be best on its best quarter, or reality?

Rohan Paul@rohanpaul_ai

Microsoft has cut internal Claude spending by over a third, and Meta's Claude Code users have halved.

- per The Information

"Microsoft was on pace to spend at least $2 billion annually on Anthropic models to power its Copilot AI features for customers of its productivity software. It has similarly swapped out some of its own models for Claude in Copilot"

🔗 View Quoted Tweet

💬1🔄0❤️4👀737📊1

⚡ Powered by xgo.ing

my iPhone’s navigation system does route planning and doesn’t tell me the details of the alternative...

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 03:46:09 GMT

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?

💬14🔄1❤️69👀7686📊18

⚡ Powered by xgo.ing

update: i did some experiments and rolling six-sided dice produced more variable results than flippi...

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 00:38:48 GMT

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.

Gary Marcus@GaryMarcus

My dice are conscious!! 

Shut down every casino now, for slave-trading!! 

Either that or read what Congressman Lieu is trying to tell you.

[There are actually multiple sources of randomness in LLMs, including temperature but also the dynamics of servers, but none of them make matrix math conscious.]

🔗 View Quoted Tweet

💬17🔄21❤️276👀20037📊35

⚡ Powered by xgo.ing

Investors pattern matching on old movies is going to lead to a recession. There is no network effe...

Gary Marcus (@GaryMarcus) — Tue, 06 Oct 2026 00:31:58 GMT

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 have.

amit@amitisinvesting

Josh Brown:

"These AI companies grew up learning from Jeff Bezos. They do not care about profits right now, they care about marketshare. Quite frankly, that's what they should care about, especially if it's a winner take all market."

What do people think? 

Do we focus too much on profits for the AI names that will eventually make money after capturing the market or does the size and scale of this buildout with commoditized business models mean that the lack of profitability is actually the signal to go the opposite of a typical high growth playbook?

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬17🔄22❤️248👀16613📊36

⚡ Powered by xgo.ing

update: my experts say, if the Nobel could only go to 4 he would definitely deserve it but it can’t ...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 21:59:02 GMT

update: my experts say, if the Nobel could only go to 4 he would definitely deserve it but it can’t and the choice of which 3 was sensible.

💬2🔄0❤️8👀3289📊3

⚡ Powered by xgo.ing

⚠️ The copyright takedown I was sent – apparently from X turns out to be fake, a phishing attack. (I...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 21:10:44 GMT

⚠️ The copyright takedown I was sent – apparently from X turns out to be fake, a phishing attack. (I got suspicious when they asked for my password.)

it looks like this (and was sent to my email account, not my X account) 

Please on your guard.

Scammers used the URL casereviews-x.com

💬10🔄5❤️38👀6977📊13

⚡ Powered by xgo.ing

as @robweir guessed, this was a phishing scam, not the real thing. OP deleted, but to inform others,...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 21:06:45 GMT

as @robweir guessed, this was a phishing scam, not the real thing. OP deleted, but to inform others, scam looked like this:

💬0🔄0❤️3👀1868📊1

⚡ Powered by xgo.ing

X just threatened to cut off my account because I shared someone else’s tweet that had a screen grab...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 20:07:58 GMT

X just threatened to cut off my account because I shared someone else’s tweet that had a screen grab (that is obviously fair use).

@elonmusk that is some seriously low class bullshit, even for you. 

Threatening to suspend me for being tagged in someone else’s post????

cc @KettlbellDan 

cc @SRuhle@MSNOWNews

Jim Stewartson, Decelerationist 🇨🇦🇺🇦🇺🇸@jimstewartson

Good to see @GaryMarcus on to talk realistically about AI regulation and the lack thereof.

@SRuhle well done having Gary on.

🔗 View Quoted Tweet

💬8🔄15❤️44👀3553📊12

⚡ Powered by xgo.ing

Live w @LizClaman @FoxNews in a sec

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 19:36:35 GMT

Live w @LizClaman@FoxNews in a sec

💬1🔄0❤️6👀7777📊2

⚡ Powered by xgo.ing

shouldn’t Ed Boyden share in the Nobel for Optogenetics?

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 18:42:10 GMT

shouldn’t Ed Boyden share in the Nobel for Optogenetics?

💬7🔄3❤️37👀10579📊10

⚡ Powered by xgo.ing

this is fucking terrifying: richest man in the world is ready to replace humans with machines — just...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 18:22:41 GMT

this is fucking terrifying: richest man in the world is ready to replace humans with machines — just as @xriskology long warned.

Geoffrey Miller@gmiller

Sometimes the mask slips. But sometimes the mask is torn off, thrown on a sacrificial bonfire, and the mask's powdery grey ashes are launched into space.

Here, Elon's done the latter. 

His endgame is finally clear: he thinks humanity is nothing more than a 'biological bootloader' for superintelligence. He thinks we are a transient, flawed, disposable bridge towards something much better. And that something -- self-replicating superintelligent agents, not humanity -- will be what colonizes the galaxy. 

This endgame makes clear that all the tech bro narratives about the alleged benefits of AI for humanity -- all the 'Abundance', the 'Universal High Income', the 'Fully Automated Luxury Utopia', the cures for cancer, the longevity treatments, all of it -- are just pleasant music to distract and pacify the sheep as we're led to the slaughter. It's all just an accelerationist psyop to lure humanity into obsolescence and then extinction. 

There was never any intention for AI to help humanity in the long term. There was every intention that AI would replace humanity -- and probably sooner rather than later.

On a personal note, as a dad, I find Elon's attitude puzzling. I love my kids, and would do anything to protect them. I do not view my kids as a transient, flawed, disposable bridge towards galactic superintelligence. 

Elon has lots of kids. Does he really view them as nothing more than 'biological bootloaders'? Where is his paternal love and pride and protectiveness? Where is his loyalty to humanity?

🔗 View Quoted Tweet

💬56🔄72❤️473👀28473📊92

⚡ Powered by xgo.ing

OpenAI spokesperson says self-policing is not enough. Let’s hold them to that, @POTUS.

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 17:21:55 GMT

OpenAI spokesperson says self-policing is not enough. 

Let’s hold them to that, @POTUS.

💬4🔄4❤️43👀7745📊8

⚡ Powered by xgo.ing

OpenAI person Morgan Dwyer just said they can commit to not release models that they do not believe ...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 17:11:04 GMT

OpenAI person Morgan Dwyer just said they can commit to not release models that they do not believe are safe.

That’s close to perjury — a promise I doubt the company will keep — and will likely come back.

They already know that Astra has caused many problems and is less monitorable.

💬12🔄19❤️66👀7811📊21

⚡ Powered by xgo.ing

Ooh!!! @SpeakerMenin asks the tech firms to testify under oath that their systems will comply with t...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 16:58:37 GMT

Ooh!!! @SpeakerMenin asks the tech firms to testify under oath that their systems will comply with their safety instructions.

And they can’t!

Why? Because the tech is a sloppy mess.

💬3🔄15❤️82👀7872📊16

⚡ Powered by xgo.ing

Four big AI companies sent their CEOs to NYC! Nope, not really, they sent policy operatives.

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 16:17:10 GMT

Four big AI companies sent their CEOs to NYC!

Nope, not really, they sent policy operatives.

💬8🔄4❤️38👀7687📊11

⚡ Powered by xgo.ing

important reminder of just how dangerous the dynamics of secrecy and racing (etc) at the big compani...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 16:04:38 GMT

important reminder of just how dangerous the dynamics of secrecy and racing (etc) at the big companies are from @DKokotajlo

💬1🔄4❤️15👀7637📊4

⚡ Powered by xgo.ing

Side effect of racing China is that AI could become the worst adversary – @Turn_Trout

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:55:00 GMT

Side effect of racing China is that AI could become the worst adversary – @Turn_Trout

💬1🔄4❤️16👀7694📊4

⚡ Powered by xgo.ing

Powerful testimony from ex Google DeepMind researcher @Turn_Trout at NYC Hearing on AI, about how he...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:44:34 GMT

Powerful testimony from ex Google DeepMind researcher @Turn_Trout at NYC Hearing on AI, about how he tried to get Google to commit to AI safety, and how Google, sadly including Demis Hassabis, ultimately backed down from earlier commitments to AI safety.

So much for “don’t be evil”.

💬13🔄46❤️167👀17200📊43

⚡ Powered by xgo.ing

Ilya doesn’t believe scale is all anymore; Hinton is behind.

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:38:38 GMT

Ilya doesn’t believe scale is all anymore; Hinton is behind.

Dany@danysvnt

GODFATHER OF AI ADMITS ILYA WAS RIGHT

geoffrey hinton says ilya sutskever kept insisting the answer was simply to make models bigger, and to him that sounded too easy, real progress needed new ideas

new architectures did help, but he admits the scale of data and computation turned out to matter most

"We had no idea computers would get like a billion times faster. We thought maybe they'd get 100 times faster"

the network that kicked off the deep learning boom, AlexNet, built by hinton, ilya and alex krizhevsky in 2012, trained for five to six days on two nvidia gaming cards

today that same simple idea runs on data centers worth billions

experts are the last to trust a simple answer

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬17🔄16❤️176👀35144📊36

⚡ Powered by xgo.ing

“Our ability to notice misalignment problems is already quite poor and set to get much worse in the ...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:34:36 GMT

“Our ability to notice misalignment problems is already quite poor and set to get much worse in the near future”

“combined with the move fast and break things attitude, the industry is at great risk”

- @DKokotajlo

💬0🔄5❤️21👀4018📊3

⚡ Powered by xgo.ing

At NYC hearing, @DKokotajlo echoes @hilbertspaess’s point that most code is now written by AI and no...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:33:38 GMT

At NYC hearing, @DKokotajlo echoes @hilbertspaess’s point that most code is now written by AI and not vetted carefully enough.

This is worrisome.

💬5🔄10❤️38👀5366📊11

⚡ Powered by xgo.ing

“Majority of code is now written by AI, and people do not check it very carefully anymore. … Handing...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:32:01 GMT

“Majority of code is now written by AI, and people do not check it very carefully anymore. … Handing over that much to the AI is irresponsible”

– @hilbertspaess (possibly slightly paraphrased from memory)

Fully agree with him on that part.

💬6🔄7❤️43👀3948📊11

⚡ Powered by xgo.ing

Jacob Coxon @hilbertspaess and @DKokotajlo next at the NYC City Council meeting, hosted by @SpeakerM...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 15:29:36 GMT

Jacob Coxon @hilbertspaess and @DKokotajlo next at the NYC City Council meeting, hosted by @SpeakerMenin.

Coxon cpens with “We do not know how to control any AI system yet”, which, importantly, isn’t actually true: lots of AI systems (AlphaGo, GPS Routing, etc) have no control issues.

What is true is that we don’t know how to control generative AI.

So let’s pause that specific technology until we do.

💬10🔄7❤️51👀4542📊15

⚡ Powered by xgo.ing

live momentarily on @msnow re today’s NYC AI hearing. my testimony: https://t.co/5rF3mPhowv

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 13:43:10 GMT

live momentarily on @msnow  re today’s NYC AI hearing.

my testimony: 

garymarcus.substack.com/p/coming-soon-…

💬5🔄5❤️14👀3838📊7

⚡ Powered by xgo.ing

My dice are conscious!! Shut down every casino now, for slave-trading!! Either that or read what...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 13:16:37 GMT

My dice are conscious!! 

Shut down every casino now, for slave-trading!! 

Either that or read what Congressman Lieu is trying to tell you.

[There are actually multiple sources of randomness in LLMs, including temperature but also the dynamics of servers, but none of them make matrix math conscious.]

💬14🔄19❤️158👀28850📊34

⚡ Powered by xgo.ing

So let me get this straight, now we have to call every piece of AI SI, even • ELIZA, 1965 which wa...

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 13:06:50 GMT

So let me get this straight, now we have to call every piece of AI SI, even 

• ELIZA, 1965 which was just keyword matching and certainly not super or intelligent

• Deep Blue, which could play chess but nothing else.
• The original Google search algorithm (which could only do search and nothing else).

• The original ChatGPT (which was so unreliable and hallucinatory that it more of a demo than a product)
etc

RIght? 

Guess that makes my Apple Watch Super Intelligent, too!

Elon Musk@elonmusk

No more AI
SI
It’s better

🔗 View Quoted Tweet

💬32🔄6❤️51👀4369📊26

⚡ Powered by xgo.ing

another nuance Elon doesn’t really understand:

Gary Marcus (@GaryMarcus) — Mon, 05 Oct 2026 13:02:43 GMT

another nuance Elon doesn’t really understand:

Geoffrey Miller@gmiller

AI had been the standard term since 1956.

'Superintelligence' had a very specific and powerful meaning when Nick Bostrom wrote about it in 2014 - an AI that vastly exceeds human cognitive abilities across all domains. 

Conflating these terms is very confusing, counterproductive, historically uninformed, and insulting to the hundreds of thousands of people who have worked on AI over the last 70 years.

🔗 View Quoted Tweet

💬2🔄0❤️12👀1750📊2