The reason why LLMs fail to play a game of chess, even though they can cite the rules from their training data, is an awesome example of what's stopping autonomous AI agents from being reliable. Great discussion between @garrykasparov.bsky.social ♟️ and @garymarcus.bsky.social 🧠 on limits of GenAI.

"The chances of #AGI’s arrival by 2027 now seem remote. The government has let #AI companies lead a charmed life, with almost zero #regulation. It now ought to enact legislation that addresses costs and harms unfairly offloaded onto the public." ~ @garymarcus.bsky.social

“Did people promising to build potential future greenhouses for tulip-growers in 1636 ever have it so good?” garymarcus.substack.com/p/peak-bubbl... by @garymarcus.bsky.social

"Power has really gone to the tech companies, who have enormous influence over the government. And unless people get out of their apathy, that’s ... certainly where the U.S. is likely to stay." @garymarcus.bsky.social discussed what we need to do to prevent AI tools from undermining democracy:

Gary Kasparov (legendary Chess Player) & @garymarcus.bsky.social are having a sobering conversation about AI, LLMs and Deepblue in this @theatlantic.com piece. Many people believe generative or any AI show signs of "intelligence", but Marcus attributes it to brute forcing with pattern recognition.

Two trials will be repeated endlessly despite always producing the same results: 1) Does UBI work? (yes) 2) Does AI improve productivity? (no)

Finally! 🤩 Our position piece: Against the Uncritical Adoption of 'AI' Technologies in Academia: doi.org/10.5281/zeno... We unpick the tech industry’s marketing, hype, & harm; and we argue for safeguarding higher education, critical thinking, expertise, academic freedom, & scientific integrity. 1/n

When tech is built to replace and not empower people, the future of work & democracy is at risk. @garymarcus.bsky.social sounds the alarm about unchecked AI and discusses how we can reclaim tech for the public good. #GoodLife https://bit.ly/4nBRcPh

‘The current strategy of merely making A.I. bigger is deeply flawed — scientifically, economically and politically.’ www.nytimes.com/2025/09/03/o... by @garymarcus.bsky.social

"The government has let A.I. companies lead a charmed life with almost zero regulation. It now ought to enact legislation that addresses costs and harms unfairly offloaded onto the public," argues @garymarcus.bsky.social.

In the morning @nytimes: Where I think AI has gone wrong — and what we should do about it.

Anybody remember AI a month ago? • Everybody in the Bay Area seemed to think that AGI was imminent • Expectations for GPT-5 were through the roof • Zuckerberg was spending bajillions of dollars on Alexandr Wang, staff and GPUs, and seemed to have a coherent plan

What scaling does and doesn’t buy you: peeling back the hype surrounding Google‘s trendy Nano Banana "But it’s all still just an extended form of mimicry, not something deeper, as the persistent failures with parts and wholes keep showing us." ~ @garymarcus.bsky.social garymarcus.substack.... #AI

not an April Fool’s joke, not the Onion

possibly among the most dangerous PACs in history.

“This entire market has been based on people not understanding that these machines don’t actually work like you, imagining that scaling was going to solve all of this, because they don’t really understand the problem. I mean, it’s almost tragic.” - @garymarcus.bsky.social fortune.com/2025/08/24/i...

Sam Altman Places Gun To Head After New GPT Claims Dogs Are Crustaceans For 60th Time theonion.com/sam-alt...

that would fuck with their narrative

Breaking: Sam Altman admits they “totally screwed up” the GPT-5 launch - and shamelessly asks for a LOT more good money to chase after bad.

“.. Fears of a bubble are mounting, as tech stocks follow a pattern that is ‘surprisingly similar’ to the dot-com bubble of the late 1990s ..” @bloomberg.com $QQQ www.bloomberg.com/news/article...

If something is incorrect on Wikipedia, it can be sourced, traced, disputed, fixed. If it's wrong in the LLM, it's just...wrong. It's not a fact explicitly stored somewhere, it's just a string of words generated by a probability map

sam’s really getting desperate now: open.substack.com/pub/garymarc...

This is how the AI apocalypse will arrive - not as malicious super intelligent AGI, but by well practiced malicious hackers taking advantage of the myth of “coding is easy, so let an LLM do it” Excellent piece by @garymarcus.bsky.social and Nathan Hamiel: garymarcus.substack.com/p/llms-codin...

LLMs + Coding Agents = Security Nightmare explainer of some scary new research, co-written with security researcher Nathan Hamiel open.substack.com/pub/garymarc...

My subscription to Gary Marcus’s blog has been some of the best money I’ve spent. All my AI anxieties evaporate after reading a few posts and they’re fun and accessible even to technological doofuses (technical term of art), like myself. Tune in!

LIVE at 1140 AM ET on CNBC Squawk on the Street Was GPT-5 OpenAI’s Waterloo? What does it mean for Sam Altman’s credibility? For the future of A(G)I? So much to talk about!

@garymarcus.bsky.social called it, way back before the hype cycle: “Speaking to The New Yorker, Gary Marcus, a neural scientist and longtime critic of OpenAI, said what many have been coming to suspect: despite years of development at a staggering cost, AI doesn't seem to be getting much better.”

Continuing to talk about why AGI is not near is, at this point, continuing to beat a two-headed horse. Oh, wait.

Wise words from @garymarcus.bsky.social Science is not a popularity contest; you can’t bully your way to truth. And you can’t make AI better if you drown out critics & keep pouring good money after bad. Science simply cannot advance without people sticking to the truth even in face of opposition.

🤣 Check all that apply.

@garymarcus.bsky.social, again, writes a very interesting article about LLMs and how the lack of a decent world model will never let them be more than a glorified autocomplete engine.

How can your AI be superhuman if it can't even beat a mouse?Lots of updates for this 2025 NeurIPS competition. Get your submission in! robustforaging.github.io Nov. 1 is the deadline. Please share widely. @garymarcus.bsky.social @dcower.bsky.social @tyrellturing.bsky.social @rowantmc.bsky.social

For everyone saying that "it's just round the corner", it's a must read. Do you repost a used cars salesman's pitches? Why would you repost Sam's then? Please read Professor @garymarcus.bsky.social 's article, it's both informative and morbidly funny.

"I love AI. (Or at least what I optimistically imagine it could be.) But I hate this bullshit." Exactly. I've always been fascinated by AI - from the simple kinds in old computer games to the fantastic ones in sci-fi. But the hype we've been seeing for the last three years has become insufferable.

“The essential read” on GPT-5 and Sam Altman’s first major blunder. Well over 100,000 people have read it. Check it out!

Too funny

Remember the time ChatGPT effectively breadcrumbed me in the grand tradition of many a grammatically proficient gaslighting narcissist? I had some more thoughts! open.substack.com/pub/amandagu...

AI Agents have, so far, mostly been a dud, by Gary Marcus @garymarcus.bsky.social / Marcus on AI - Substack / garymarcus.substack.com/p/ai-agents-...

Where to even begin with this one? ChatGPT gave The Atlantic advice and encouragement on murdering someone—also self-mutilation, devil worship, and ritual sacrifice of large animals...

Absent any information of that kind, there is little reason to take on faith that whatever unknown techniques allowed them to do so well on IMO problems will generalize to any other kind of problem." ~ @garymarcus.bsky.social and Ernest Davis garymarcus.substack.... #GenerativeAI #deepmind 2/2

Thought y’all needed to know that Grok has pencilled in civil war in Britain for a week on Tuesday

My thoughts on the crucial importance of methodology on self-reported AI performance on mathematics competitions, and my policy on commenting on such reports going forward: mathstodon.xyz/@tao/1148814...

The best way to foster them is to have an intellectually open attitude. It would be great if we could see more of that." ~ @garymarcus.bsky.social garymarcus.substack.... #GenerativeAI 2/2

Proud to see our research with @garymarcus.bsky.social cited in the @financialtimes.com ☀️

My latest research on LLM grooming, co-authored with @ninajankowicz.com and @garymarcus.bsky.social, found that OpenAI's so-called "reasoning" o3 model cited Pravda network content 26.8% of the time in response to specific prompts. This research got cited today in a column by the Financial Times:

Disinformation warriors are ‘grooming’ chatbots - on.ft.com/46hV6qW via @FT

Even against all of Trump's cascading attacks over the past seven months, it had seemed that up until now, our culture at large—the literal shows, movies, art, music, literature we consume—seemed somewhat siloed from Trump's fury. But the end of Colbert marks a dark new stage.

This from @garymarcus.bsky.social is exactly where I land: the computers aren’t the real threat, it’s the humans who control the computers that we really need to worry about. Serious talk of rogue AI is mostly a gambit to distract us from seeing who the real rogues are

"It turns out that building it on black boxes..., and it’s all a crapshoot based on how similar your query is to what it happened to have been trained on. It’s just the wrong paradigm. That doesn’t mean we can’t invent a better one..." ~ @garymarcus.bsky.social www.politico.com/new...