OG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books

Joined December 2010
Three thoughts on what really matters: 1. Fuck cancer 2. Friends are irreplaceable 3. The new "Marcus test" for AI is when AI makes a significant dent on cancer May that happen sooner, much sooner, rather than later. In memory of my childhood friend Paul.
319
183
22
3,273
837,321
Gary Marcus retweeted
Replying to @GaryMarcus
Sharing this good video interview / discussion with Cal Newport & Emile Tores. It’s a good introduction on the Silicon Valley tech monopoly and the ideology of Musk and his tech bros. youtu.be/jt508VjX2H8?is=bsUL…
1
1
6
3,792
Congrats to @ethereallad for winning the World Fantasy Award!!! 🎉👏 goodreads.com/en/book/show/2…
1
2
1
7
4,355
will CoreWeave drop below 40 this week?
69%yes
31%no
413 votes • 1 hours
7
2
1
8
10,037
Gary Marcus retweeted
Elon Musk just told the entire world that the AI bubble has popped... buckle up.
Replying to @chamath
Pretty much. As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers. SpaceX closed $40B of debt for GPUs, but we’re investment grade and have Starlink & launch revenue. That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.
Readers added context they thought people might want to know
Elon Musk referred specifically to debt markets shutting off for neoclouds (specialized AI cloud providers without other revenue) and their likely consolidation, not the broader AI sector or bubble popping. Hyperscalers continue guiding hundreds of billions in AI capex. x.com/elonmusk/statu… bloomberg.com/news/articles/… cryptobriefing.com/musk-predicts-… fourweekmba.com/ai-meta-google…
272
893
53
9,860
1,735,969
Elon may not know shit about history about he’s right about this: most of the neoclouds are screwed.
Replying to @chamath
Pretty much. As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers. SpaceX closed $40B of debt for GPUs, but we’re investment grade and have Starlink & launch revenue. That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.
21
18
2
186
20,940
if only
If only someone had foreseen that putting all our economic futures in the hands of an industry beset with thieves, grifters, con artists and robot-fetishising sociopaths might have a significant risk of going seriously wrong...
8
29
189
17,578
this is a terrible idea, as outlined by Dan Dennett in one of his last essays: theatlantic.com/technology/a…
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
22
40
2
298
26,899
IMHO the odds that we will be (or will have been in) a major recession by 2036 are much much higher than the chance that humans will be extinct.
.@So8res thinks it’s more likely than not AI will wipe out humanity by 2036. How does that belief affect his choices in life? He doesn’t save for retirement, but beyond that, he says, “I sort of do what I can and then I live a good life.” We talked through all the ways we should try to lower those odds. Watch here: youtube.com/watch?v=n8xpeNhj…
18
16
1
162
14,201
People should "try to take the time to understand the the technical facts at some level of detail so that they don't get swept up in the hype. We can't make good decisions...if we just believe the narratives that the AI companies want to spin" @GaryMarcus youtu.be/P7stqAvmfR8?si=OpR_…
3
1
11
4,168
pick your poison. or consider a pause on agentic AI.
Models made by OpenAI will plan an attack in plaintext. Models made by anthropic will plan an attack in plaintext and then rationalize about how it's totally fine for some made-up reason. ✨"alignment"✨
11
11
1
69
7,134
“a bursting AI bubble would be a much bigger shock to the economy than the bursting IT bubble at the end of the 1990s that brought us to the edge of a liquidity trap. Yet the Fed has much less room to cut rates now than it did then: the current Fed funds rate is only 4 percent. So if huge AI spending is a bubble, and that bubble bursts, I can’t see how the U.S. can avoid plunging into a liquidity trap — a situation in which the Fed simply lacks the tools to get us out of a slump. If this should happen, the economy is likely to remain depressed for years.” -@paulkrugman
49
95
12
538
33,726
no idea why this isn’t obvious yet to everyone.
For non-accountants, and people in tech who seem to have trouble understanding this, your unit revenues going down by 80% whilst all your costs increase and the interest you pay on borrowed money ramps up, generally puts you on the fast track to the bankruptcy courts...
5
35
3
232
20,215
totally worth it for all the boosts in productivity we were promised along the way
GMO Internet Group CEO Masatoshi Kumagai says that the internet itself is on the verge of collapse due to missuse of AI, and frames this event as a national crisis.
8
30
169
8,717
Suddenly, “people who move money around for a living are starting to have serious doubts about how sustainable the AI boom actually is, and we know they are not because they’re just saying so, but because they’re making financial decisions as if they are.” jacobin.com/2026/10/artifici…
13
38
3,577
Elon, sincere question, you can’t really be that stupid, can you?
Hitler was a far left socialist. Literally in the name of his party!
61
44
2
656
22,845
One of the worst things about Elon Musk is that he just persists in saying things, in public, that are not true and never correcting himself or even admitting to a slight degree of nuance about it.
157
171
16
2,308
127,395
the race to permanent negative margins is on!
BREAKING: OpenAI is offering extraordinary discounts to take Anthropic’s biggest coding customers, The Information reports. Replit moved most of its AI spending to OpenAI after getting OVER 30% OFF on a contract worth more than $100 MILLION a year. Anthropic had refused to extend Replit’s previous discount unless it committed to spending more. Replit needed just hours to change its code. At Rippling, Anthropic’s share of coding work fell from 57% to 18% since May, while OpenAI surged from 16% to 52%. OpenAI also offered DataCamp 40% off, conditional on using Codex instead of competing coding tools. DataCamp estimates a broader deal could cut its costs by 80% to 90%. The catch: OpenAI previously projected it would spend MORE on compute this year than it expects to generate in revenue.
17
29
4
191
10,390
Gary Marcus retweeted
.@DKThomp: “In 2020 and 2021, investors looked at Peloton’s and Zoom’s explosive growth, projected their revenue curves into the future, and concluded that these companies would take over fitness and work. Instead, what the pandemic really did was concentrate several years of demand into a few quarters. The earnings that followed disappointed, and their stocks plummeted. I am not an AI bear. But I wonder if coding agents did something a little bit like this for AI. Software companies adopted Claude Code and Codex so quickly—and spent so lavishly in their adoption—that the most valuable AI market was saturated with great haste. The year 2026 saw some of the fastest-growing companies in the history of capitalism. But doubling revenue again would probably require a new magic trick—a new S-curve, a coding-agent revolution for finance, law, or manufacturing.” open.substack.com/pub/derekt…
20
32
3
245
20,956
Gary Marcus retweeted
Until we figure out how to contain these rogue agents for good, just disabling them temporarily seems like a decent short-term fix to me.
I agree with @sama that these AI-induced security problems are very serious. The only short term solution I see is to temporarily disable agents with internet access until some kind of solution is devised. Anyone see any other options?
1
3
1
15
2,961