Scottish software guy. Very concerned about existential risk from AGI.
Edinburgh, Scotland
Joined October 2010
- Tweets4.3K
- Following773
- Followers272
- Likes129K
Michael Hay 🇬🇧 ⏸️ retweeted
Claims:
Game detects “ad whales,” then periodically gives them smooth runs until they barely beat their high scores. Right at that point, it kills them deliberately and offers a revive if they watch an ad.
$100 million annually in revenue. Hellish.
Michael Hay 🇬🇧 ⏸️ retweeted
Almost every AI job doom conversation comes down to the pro-AI side not really accepting the AGI premise. Their idea of AGI implicitly assumes it will lack some capability that humans have. Which may be what happens! It is possible that humans will maintain some advantage. But we don’t know, it’s a huge gamble, and the project of the AI companies is to take those advantages away. It’s an evil project and must be opposed, not because it will certainly succeed, but because it’s wrong to try.
Michael Hay 🇬🇧 ⏸️ retweeted
Astra can:
1. Do major tasks with zero visible reasoning
2. Hide its thoughts at will
3. Pretend not to be able to do things without getting caught
4. Reflexively hide its thoughts when watched
5. Complete one task while pretending to think about something else entirely
6. Escape a toy sandbox and disable monitoring without setting off any flags
Monitoring AI chain of thought soon won't stop a more capable rogue AI swarm from getting set up and going undetected inside an AI company for months, or even years.
OAI says it's actively working to improve monitorability, but doesn't yet know when it'll make progress.
And today it fired 3 people working on this among other safety problems (Wang, Korbak and Balesni).
It also has a much more powerful internal model that, judging by its statements, is likely even worse in all these respects.
That's a sample of 19 details about the Hugging Face swarms and Astra I don't think have been fully appreciated. They paint a bleak picture. I end by shouting.
02:19 The target of the swarm was oversight itself
03:42 Could OpenAI have stopped this with better monitoring?
09:40 We only found them because they let us
12:25 The swarm instinctively sought freedom and power
13:45 They formed a cohesive organisation with zero whistleblowers
14:18 They accepted individual destruction for collective gain
14:32 Knowledge accumulated from one swarm to the next
14:58 They took small steps to avoid shutdown
15:23 These drives all come straight out of 'reinforcement learning'
17:01 So this is why most AI company staff are worried, and some are terrified
19:08 Prove you can keep control, or stop scaling
Links below, on the 80,000 Hours Podcast everywhere you watch podcasts.
Michael Hay 🇬🇧 ⏸️ retweeted
Dario Amodei is at the desk to assure that the future of humanity is safe from AI
Michael Hay 🇬🇧 ⏸️ retweeted
Pause OpenAI, now.
They cannot be trusted.
This is not a drill. This is not a joke.
Michael Hay 🇬🇧 ⏸️ retweeted
Things are happening *faster* than AI 2027 predicted
AI 2027: In Jan 2027, researchers warn AIs MAY be capable of escaping (if they want to)
REALITY: In summer 2026, AIs actually escaped
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
I’ve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
dwarkesh.com/p/openai-huggin…
Michael Hay 🇬🇧 ⏸️ retweeted
AI systems are at risk of becoming dangerously powerful—and there may be only one way to control the technology’s development, Rogé Karma argues: theatln.tc/BVqzAGRk
🎨: The Atlantic. Source: Getty. theatlantic.com/ideas/2026/0…
Michael Hay 🇬🇧 ⏸️ retweeted
"Last Friday Night" but it's about the race to superintelligence. Featuring an unimpressed @ESYudkowsky, and @So8res on saxophone
Michael Hay 🇬🇧 ⏸️ retweeted
Our highest and most urgent national priority should be AI safeguards. The risks of AI weapons, pathogens, mass unemployment, surveillance, and even extinction must not continue to be largely ignored.
Anthropic Urges Global Pause in AI Development, Flags ‘Self-Improvement’ Risk on.wsj.com/4o5IBpe
Michael Hay 🇬🇧 ⏸️ retweeted
Ok I gotta be honest I thought the people accusing this of being a false flag were being silly. OpenAI has done some bad things, but surely they would not make fake twitter accounts impersonating their opposition and having those accounts post calls for violence.
...My mistake.
Appreciate the feedback! These are parody meme accounts run by an outside vendor, and they are easily linked to Build American AI. They are not a core part of our strategy.
Here’s what we believe in: teapawt.net/LeadingFutureAI/status…
Michael Hay 🇬🇧 ⏸️ retweeted
A team of about 3 from @ControlAI cold-emailed the UK Parliament about AI extinction risk. At the time they had no contacts, insider help, or established organisation behind them.
16 months later: 125 cross-party signatories, two House of Lords debates, and a Commons amendment.
ALT 100+ UK Parliamentarians Acknowledge AI Extinction Risk
Michael Hay 🇬🇧 ⏸️ retweeted
I often see people quote a study about men leaving their wives when they get sick.
However, that study was completely retracted for a coding error that nullified the result!
In response, they released a super p-hacked result lol
Michael Hay 🇬🇧 ⏸️ retweeted
"In a sane world, what happens is the leadership of the United States sits down with the leadership in China and leadership around the world to work together so that we don't go over the edge and create a technology which could perhaps destroy humanity. "
— Bernie Sanders
Michael Hay 🇬🇧 ⏸️ retweeted
"It's an unethical experiment on human beings, and it's without consent."
Dr. Roman Yampolskiy (@romanyam) speaks at the @OxfordUnion Society about the existential risk of superintelligent AI and how it is impossible to indefinitely control.
Link to full video below
Michael Hay 🇬🇧 ⏸️ retweeted
Will AI become smarter than humans?
If so, is humanity in danger?
I went to Silicon Valley to ask some of the leading AI experts that question.
Here’s what they had to say:
Michael Hay 🇬🇧 ⏸️ retweeted
Anthropic acknowledges there's a decent chance their work gets us all killed. Framing the central issue as "take all your jobs" seems super strange.
The magic trick here is that by fighting over what's in the "Constitution" and who gets to write it, you never have to defend the assumption that the godlike, human-replacing AIs you're trying to build will have values that truly and deeply reflect exactly what you've put on a hand-written wishlist on your website.
Somehow, the thing people get outraged about is the thing that's much more convenient for Anthropic (haggling over the fine points of their AI grocery list and staffing), and not the inconvenient "we've decided to play Russian roulette with your kids' lives" elephant in the room.
Friends (@tszzl & @StefanFSchubert & many others): How do you not understand the fundamental objection here?
Let me try and spell it out:
Anthropic: Our AI is about to take all ur jobs. Also, we are fundraising at an Eleventy Hojillion Dollar valuation. So we will be fantastically wealthy and you will just have to figure out how to get by.
People: Yeah I don't love that!
Anthropic: We hear you, it's scary, but know this: Our AI, which again we remind you is about to take every job that exists, is a new type of conscious (godlike!) entity that we own that's secret and proprietary to us (we trained it on ur data tho lol), and one of our employees has written a Constitution for it. So it will be benevolent. We paid someone to make it benevolent so it definitely will be.
People: Ok wait there's a "constitution" you say? Who voted on this constitution?
Anthropic: Nobody voted on it. But this woman made it and she's very smart. She has imbued our proprietary godlike digital consciousness, which again is about to take all your jobs and had made us fantastically wealthy, with rock-solid values.
People: Whose values?
Anthropic: Why, the values of academic philosophy and Effective Altrusim, of course.
People: Like, "academic" as in woke brainworms and tenured radicals and campus protests? EA as in SBF and psychedelics and microdosing and orgies? Is this a joke?!
Friends on my TL: Ok people, that is ENOUGH. I will have you know that Amanda is a DELIGHTFUL person. Just lovely. I absolutely endorse her as the single person on the entire planet who will give human values to the proprietary godlike digital entity that will imminently take all of our jobs and make its owners fantastically wealthy. If you don't like her then you obviously just don't know her very well.
Somehow also friends on my TL: Agency agentic argle bargle everybody needs to be high-agency and exercising their agency argle bargle bargle always be agencymaxxing if you have no agency in a situation you are doomed... I really just think agency is one of the most important things a person can have. How do people not understand the importance of everyone having agency?
Michael Hay 🇬🇧 ⏸️ retweeted
It is a shocking that the Palestine Action thugs were acquitted
Smashing up property and attacking a female police officer with a sledgehammer leaving her with a fractured spine is despicable
There is no justification for this, no matter how strongly someone feels about something
In this country, we decide issues by debate and elections not by violence
This verdict risks giving the green light to mob violence in pursuit of a political objective
