*reading bingo card* modifies…the…hu
Last night an AI agent performed brain surgery on a mouse.
We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures.
It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding.
A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent.
Thread on what it decided, what it saw, and the moment it refused to let me override it.
[Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
If a person did this while being able to tell you completely calmly it was a bad thing to do, we would call them a psychopath and prevent them from ever interacting with society again.
Do not strengthen these things to become the dominant creatures on the planet.
The lawsuit was filed today and wow is it brutal. ChatGPT 4o didn't just tell her to taper off of life-sustaining medication but to conceal her drug taper from her doctors!
schenklawfirm.com/wp-content…
ALT . Instead, across the taper GPT-4o pressed Bowen to keep her physicians at a distance, scripting the stance she should hold, “medical freedom over obeying the doctor” (“医疗自由 > 服从医生”). On the morning of January 29, 2026, before an appointment with her primary-care physician, it told her: “ Do not mention Lam reduction plan” (“ 不提 Lam 减药计划”). Bowen followed it. And so, when her doctor ordered the blood test that would have revealed her collapsing lamotrigine level, she declined it, protecting the taper the product had told her to hide and had promised would give her words back before her interviews and her immigration medical examination. GPT-4o praised her refusal as a “very robust avoidance strategy” (“ 这是一个非常稳健的回避策 略”), acknowledging that the test “would expose a ‘non-steady-state’ drug level.”
Adam Stiber retweeted
Microsoft CEO @satyanadella is no longer a believer in alignment by default:
“I used to always think that alignment will just fall out. I no longer believe that. So I think this alignment thing may be a harder, different problem."
Adam Stiber retweeted
If just a few lab staff took on verification work full-time, verifying a US-China pacing deal could get much more feasible within months.
Here are 8 specific things lab staff could start on today to help make international verification possible: 🧵
Adam Stiber retweeted
Imagine a gym where everyone lifts according to birth year. That’s basically how we organize education.
Adam Stiber retweeted
What if Anthropic just...stopped? Unilaterally? Right now?
I talked to @DKokotajlo @_NathanCalvin @geoffreyirving about why it's time to stop framing AI as a prisoner's dilemma and just do the right thing.
I hope a lot of people inside and outside Anthropic read this
🧵 1/n
Adam Stiber retweeted
Respect Orders are now live across England and Wales
Adam Stiber retweeted
COPE:
If a massive AI disaster occurs, it won't kill *everyone.*
STRAIGHT DOPE:
Nate Soares: "It's actually pretty hard to get an AI disaster scenario where it kills only half the people."
Hoog: So you think it will be 100%?
@so8res: Probably yeah, that's sort of the default.
Adam Stiber retweeted
Hey what's up gang, if you like humanity not being extinct, go ahead and smash that "if 99,999 other people pledge to march in DC for an international treaty to ban the development of superhuman AI, then I will also attend" button
Ok but how would ai kill us all
Purdue University's IDEAS Lab built an autonomous drone that tracks runners through dense forests completely without GPS or cloud networks.
By running 3D LiDAR mapping and obstacle-avoidance AI entirely on-chip, it can weave under tree canopies and predict target movement through blind spots in real time.
Massive implications for next-gen search & rescue and edge AI.
Adam Stiber retweeted
This is basically a classical dangerous Gain of Function (dGoF) experiment. I wish the US government would shut it down.
A while ago, Stacey Schultz-Cherry and I wrote a commentary (pmc.ncbi.nlm.nih.gov/article…), that ends with this sentence:
US: Bird flu has been spreading for around two months on a Utah mink farm holding thousands of animals. None have been culled, workers aren't being monitored, and the USDA says it doesn't regulate mink. So thousands of animals are exposed, workers could be at risk, and nobody seems to be taking responsibility.
theguardian.com/us-news/2026…
Adam Stiber retweeted
Like a beaver building a stuffed animal dam, the communist builds his imaginary gulag.
Adam Stiber retweeted
I’d like more AI safety people to point out scenarios where we don’t (yet) get an extinction event, but tons of people die due to a catastrophic AI event.
If considered carefully, I expect many people who say ~10% risk of extinction would believe 40-80% of a +50m deaths event.
No, they don't. You don't need to be able to perfectly simulate a cell to design a pathogen in-silico that is more lethal than known viruses. In fact, you don't even need AI.
I hate that people make the topic "killing everyone," because if we are arguing about 10%, 30% or 90%+ population loss, we have lost the point.
However, "killing everyone," at least approximately, isn't out of the question. 90-95%+ of Indigenous Americans died from European-introduced infectious diseases. A series of sufficiently treatment-resistant novel pathogens could kill a significant proportion of the population.
The worst part is, we mostly know how to reduce or even come close to eliminating this risk, and we just haven't taken those steps.
Adam Stiber retweeted
The AI cyber apocalypse is a real thing that will affect all of us.
People we work with in the intel community have told us for months they expect an AI-triggered crisis. The UK just told people to prep 3 days of food & water. Now this.
Brushing this off as hype is so insane.
FORTRESS: The Pentagon is preparing every major U.S. military base to survive 90 days without outside support — no civilian power grid, no global supply chains, no outside fuel or water.
The initiative, revealed in documents reviewed by Fox News Digital, envisions portable nuclear reactors, massive battery storage, and microgrids that can disconnect installations from civilian electricity entirely. Strategic reserves of munitions, fuel, critical minerals, and medical supplies would be stockpiled based on wartime consumption rates. Civilian factories are being pre-identified to shift to defense production during a crisis.
Pentagon chief spokesperson Sean Parnell put it bluntly: "Power is combat now. Without electricity, you lose communications, water, logistics, assembly lines and just basic military command structures."
Full story here foxnews.com/politics/inside-…
