Gary Marcus

AI Bait & Switch: bait: we’re gonna make an AI that can solve any problem experts could solve. it’s gonna transform the whole world. switch: what we have actually made is fun and amazing, but rarely reliable and often makes mistakes – but ordinary people makes mistakes, too. So … AGI solved!

When it comes to addressing the harms & risks of AI, should we be focusing on law (liability), regulation or business practices? @garymarcus.bsky.social says YES to all, and I couldn’t agree more!

"What we are facing near-term, anyway, is not extinction. And it’s not superintelligence… It’s wholesale deepfaked disinformation, and unreliable but persistent AI systems stealing credentials and launching cyberattacks, at scale." @garymarcus.bsky.social garymarcus.substack.com/p/big-news-a...

"The evidence that OpenAI is building something the company isn’t managing to control is vast...I stand by what I told CNBC yesterday (and wrote here earlier in the week) It’s time (past time, really) to (temporarily) shut down OpenAI" @garymarcus.bsky.social garymarcus.substack.com/p/breaking-o...

the us-china relationship is the single most important in geopolitics today. is collaboration > confrontation? (57% of americans think so)

@aoc.bsky.social just put out a great statement on AI. Refreshing to see a major politician on the left nail this. "Silicon Valley billionaires want us to believe that Al robots will destroy humanity. But the biggest threat to humanity are the tech oligarchs themselves." 🎯

A group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...

More Questions About AI Dangers, With Questions From Gary Marcus talkingpointsmemo.com/edblog/more-...

My thoughts on existential risk: * Humanity isn't going to be wiped out any time soon. * Negligible chance that AI by itself causes catastrophic damage to humanity (millions dead). * Some nontrivial chance that human beings will leverage AI to help them do something catastrophically harmful.

It is actually worth reading the different takes by Dwarkesh (easy to read but too anthropomorphising), @garymarcus.bsky.social (strong warning), and the one below by @eryk.bsky.social.

what Dwarkesh got wrong and why it matters, a detailed dissection: garymarcus.substack.com/p/dwarkesh-p...

“Draw a cartoon of Sam Altman talking out of both sides of his mouth”

Some important thoughts on the US and Canada that won’t fit in 300 characters.

‘The opening stages of OpenAI’s unraveling have begun.‘ garymarcus.substack.com/p/breaking-o... by @garymarcus.bsky.social

Breaking: Young adults are now more concerned about AI than enthusiastic. via @pewresearch.org

I'm keen on AI having a transformational impact on human health. But these cure projections are wildly off-base and impossible.

“.. what happens to the buyers of these chips when investors realize they’ve sunk billions into depreciating assets that they can’t even use?” - Ross Hendricks @theinformation.com @garymarcus.bsky.social www.theinformation.com/newsletters/...

‘A few weeks ago Zuckerberg was extolling the value of open-source. And then today he released a model that is NOT open-source, hoping you wouldn’t notice the difference.’ garymarcus.substack.com/p/open-sourc... by @garymarcus.bsky.social

Among other experts, the Guardian article also includes commentary from @garymarcus.bsky.social, whose recent blog - Richard Dawkins and The Claude Delusion - is well worth a read too. 👇

Crux of Musk-OpenAI trial so far is that Brockman keeps acting as if selling chatbots & API access for profit (which is what they do now) is exactly the same as the original mission (working on the AI safety problem for public benefit). They aren’t. If the jury figures that out, OpenAI loses.

Truth, c/o @garymarcus.bsky.social

Excellent reminder from @garymarcus.bsky.social of all the arguments about why chatbots(a far better term than the misleading “AI”) are not conscious, using Dawkins’ daft infatuation as the starting point.

Thank you again @garymarcus.bsky.social, this time for putting in words the disappointment I felt after reading how easily Richard Dawkins was deluded by a language model. garymarcus.substack.com/p/richard-da...

“The fundamental problem here is that #Dawkins doesn’t reflect on how these outputs have been generated. Claude’s outputs are the product of a form of #mimicry, rather than as a report of genuine #internalstates.” open.substack.com/pub/garymarc...

What is the one thing that most worries you about AI? In a survey of 3,700 AI researchers, only 3% replied existential risk—despite the prominence given to these risks in media. Far more experts are worried aout malicious use, misinformation, job losses, bias, etc. www.nature.com/articles/d41...

A deranged shooter racing through a hotel to shoot the President is a real wake-up call … on the need for g̵u̵n̵ ̵c̵o̵n̵t̵r̵o̵l̵ a gigantic ballroom. www.washingtonpost.com/politics/202...

What the F is The Onion supposed to do, when this actually happened?

Geoffrey Hinton still very wrong about radiologists futurism.com/artificial-i...

Should you worry about AI killing us all? I looked at the evidence so you don't have to 🧪 www.nature.com/articles/d41... @nature.com Ft. @garymarcus.bsky.social @sashamtl.bsky.social @gleave.me @andreamiotti.bsky.social @anthonyaguirre.bsky.social @ghadfield.bsky.social among others!

Regurgitating ≠ understanding c/o @garymarcus.bsky.social

OK, I'm obsessed with this study in @science.org It took Reddit "Am I The Asshole" posts & asked LLMs if the poster was the asshole. Aaand (surprise) AI was more likely to tell people they were NOT the asshole ... even when humans said yeah YTA 🧪 www.science.org/doi/10.1126/...

Excellent, heartfelt post by @garymarcus.bsky.social on the more-than-a-decade-old failure of AI/machine learning to create anti-cancer drugs, and why (hint: cancer is not a bug in code).

Dario is no saint either: open.substack.com/pub/garymarc...

This might be a REALLY good time to listen to @garymarcus.bsky.social 's advice...

Breaking: Altmans’ lack of character, which I have been warning about for years, has finally caught up with him: open.substack.com/pub/garymarc...

Just when you thought this tragedy couldn't get any stupider.

LLMs are an epistemic nightmare. open.substack.com/pub/garymarc...

I am seeing a lot of calls to boycott OpenAI — and I support them.

The whole thing was a scam open.substack.com/pub/garymarc...

It’s one thing for the government to reject @AnthropicAI’s terms—- and entirely another to banish them permanently (and, absurdly and punitively declare them a supply chain risk) in favor of someone else who took pretty similar terms and happened to have given more campaign contributions.

I will never not be angry at the way tech bros have told us that the danger is apocalyptic superintelligence when it was always obvious that the reason to worry is that we would put dumb AI in really stupid places.

Embedding unreliable AI deep into the source code of the military without human oversight is the single stupidest thing we could do as a species. And that is *exactly* what the United States is about to do.

The urgent plea of @garymarcus.bsky.social demands rapid action: the US Military, in the hands of fascists, seeks to bend Anthropic to the knee in the aims of incorporating AI into military weaponry. Gary asks us to call our political representatives right now. I just did. Now it's your turn.

open.substack.com/pub/garymarc... If I could, I would like your attention for a moment. This post from @garymarcus.bsky.social is exceedingly important-please read it, share it, and take action.

The most urgent plea I have ever written. Please read and share. open.substack.com/pub/garymarc...

open.substack.com/pub/garymarc... I don't normally share the news of people's loved ones passing. I have yet to become comfortable with how social media handles this. But this essay by @garymarcus.bsky.social in honor of his mother is too wonderful not to share widely.

Code red for humanity? I am genuinely frightened. open.substack.com/pub/garymarc...

Stick it up *your* ass, chatbot. www.404media.co/rfk-jrs-nutr...