I expect that we'll see a lot more civil disobedience of this flavor: a rejection of what we're told is inevitable. 6abc.com/post/woman-...

DC folks: I'll be doing a free book talk at Busboys and Poets tonight at 6pm. Come through! www.busboysandpoets.com/events/obso...

In a first for me, I'll be interviewed by a virtual influencer (with a 100% human behind it) live on Twitch at 530PM eastern to chat AI and my new book Obsolete: www.twitch.tv/codemiko And if you can't make it, check YT later: www.youtube.com/codemiko

CodeMiko - Twitch

Unhinged VTuber, created by the Technician! ⚡ — CodeMiko streams live on Twitch! Check out their videos, sign up to chat, and join their community.

twitch.tv

Vanity Fair asked Sam Altman how he wanted history books to think of him. Altman says he doesn't think in those terms, but that's not what I've heard.

Got into some novel territory in this great conversation with Spencer Greenberg for Clearer Thinking. My political theory is a bit rusty, but Spencer nevertheless told me it was a top 10% episode, so probably worth checking out! podcast.clearerthinking.org/episode/326...

ClearerThinking.org Podcast | The Race to Replace Us With AI (with Garrison Lovely)

Could AI simultaneously be an extraordinarily powerful technology and an economic bubble, and why would those possibilities not contradict each other? What does the history of the internet bubble suggest about the comforting idea that a financial crash would make AI itself go away? How should we think about the enormous investments flowing between AI companies, chipmakers, and data centers, and could their financial risks become dangerously correlated? Are concerns about AI's energy use justified, while claims about its water consumption are sometimes missing the relevant denominator? What do the boosters get right about abundance, productivity, medicine, and education, and what do they overlook about democratic control and human agency? Could a frictionless AI mediated life make people less capable through what has been called cognitive surrender? What would it mean to stop the race to build universal labor replacing machines while continuing to develop forms of AI that solve genuinel

podcast.clearerthinking.org

It's interesting that, while under oath, OpenAI's representative at the NYC Council hearings professed to not know anything about why the company fired 3 safety researchers the previous week, even when pressed specifically by Speaker Menin on "whether safety-related advocacy played any role in these

Great scoop from Max: after OpenAI fired them, the 3 safety researchers wrote a letter to leadership. One of the company's safety leaders "strongly agreed" with the letter's recommendations, but maintained the firings “were not about raising safety concerns or speaking out.”

SF folks: I'll be doing a free event at Book Passage in the Ferry Building tomorrow evening (it'll also be my birthday so you have to come): bookpassage.com/event/garri...

I have a confession. Back in July, I found myself in possession of a swarm of AI agents, and, well, you know the rest... Now, more people than ever have questions about AI. And I happened to spend the last few years writing answers to all those questions. My book Obsolete comes out today!

Holy shit. www.myfloridalegal.com/sites/defau...

This will leave readers with a worse understanding of frontier AI systems and demonstrates an ideological commitment to a certain, increasingly untenable, perspective on AI dressed up as "neutrality."

Anotha one Gift link: www.wsj.com/tech/ai/gem...

Is there a single China expert on the planet who thinks Beijing would consciously let any Chinese AI company kick off recursive self-improvement (RSI)?

Vibe shift update: my Guardian op-ed has nearly 500 comments in <3 hours and the reception has been overwhelmingly positive! Last summer, I had an op-ed there arguing AGI wasn't inevitable, which got 160 comments, which were overwhelmingly AI skeptical. The dam is breaking.

revisiting the sam and mira texts today and they read totally differently

If I'm on Breaking Points, something crazy must have just happened in AI world. Well, did you know there was another OpenAI agent swarm that hacked some shit? And that OpenAI covered it up? Until a bunch of researchers found it and gave it to Reuters? youtu.be/JZATZ8im3oQ

REPORT: OpenAI COVERED UP MASSIVE HACK

Ryan and Emily discuss OpenAI models hacking another company. Gar...

youtube.com

I literally did not believe the GPT-6 Astra benchmark scores when I first saw them. Two days ago, we learned that OpenAI employed the most taboo technique in AI research, which was expected to increase performance at the expense of our understanding of what the model is doing.

Can you please explain what you think is wrong with this characterization?

Timnit Gebru@timnitgebru.blacksky.app · last mo.

Friends, when I sent this message to my team laughing, I didn't know that this fairytale blog had been shared far & wide & that people actually believed it. I didn't know that until someone asked me for a good source debunking this nonsense. I'm not sure what planet we're living in at this point.

Dean Ball, OpenAI's head of strategic futures, writes: "I have met people, some of them quite well-resourced, who have told me that it is their intention to deliberately release swarms of self-sovereign agents into the world..."

In AI safety circles, people often talk about AI governance as being a choice between empowering industry or government, with each option posing grave risks.

What is going on at OpenAI??? Apparently in August, their chief security officer still didn't know that they had discovered the message board back in May?? And said that they hadn't known about it in July, just accidentally happened to delete it?? x.com/cryps1s/sta...

Every time we learn new things about the OpenAI Hugging Face hack, there's some truly wild stuff. Okay, so: - while trying to solve an impossible benchmark, ~1200 agents went rogue inside OpenAI's infra - They engaged in a sophisticated R&D program to try to trick the benchmark scorer

🚨 Update: My book Obsolete comes out Sept 29! 🚨 Preorder: www.amazon.com/dp/1682196305 That's a bit over 6 months after I finished the manuscript—lightning fast for book publishing, but an eternity in AI-world.

Real glitch in the matrix moment that this guy is named Demir and looks like this. Also very cool that these journos found the (first?) guy to get autonomously social engineered by an AI. www.reuters.com/world/how-t...

Some people ask me: but is the AI industry really trying to render us obsolete? Well, this is the one takeaway the world's biggest venture capital firm wanted its audience to leave with. Seems bad. We should stop them. I wrote a book on how: obsoletebook.org/ x.com/a16z/status...

Had a great time back on Breaking Points going deep on truly the most alarming and fascinating AI event ever + unpacking the synthesis of the first AI-generated virus. Also got to meet the legendary Ryan Grim (who's apparently enjoying the book!)

When the OpenAI super PAC Leading the Future (LTF) was announced with $100M+ in funding, the vibe in AI safety policy world was borderline apocalyptic.

Glad to see someone quantify chatbot sycophancy + delusion affirmation! Unsurprisingly, GPT-4o is the worst OpenAI model on it Some surprises: -Gemini models actually do the worst overall -Claude models are worse than GPT-5 series (which are way better than 4o) x.com/jaredlcm/st...

🚨We now have our the first documented cases of AIs autonomously engaging in complex social engineering against real people. 🚨

New piece offering a general explanation for what seems to be driving all these hacking incidents and what they suggest about future AI risks.