Tag Archives: ai
5%
If a civil engineer started building a bridge meant to carry a couple hundred cars a day, then said there was a 5% chance the bridge might collapse, they would be fired on the spot and the bridge torn down the next day.
If your bank offered you a new, high-interest savings account, but warned you there was a 5% chance all the money in it might vanish without warning and without hope of recovery, you would refuse the offer (and probably switch banks).
If a car manufacturer started making a new type of engine, and claimed it was so powerful their cars had a 5% chance of exploding above highway speeds, there would be an immediate recall–even though driving that fast is against the law.
If Microsoft pushed out an update to Windows that they claimed would make everyone’s computer run twice as fast, but with a 5% chance it would fry the CPU and erase everything on the hard drive, there would be a lawsuit at minimum–even if no one’s computers actually got fried.
If a biotech lab got a government contract to engineer a viral weapon, and they claimed that the virus they were developing was so potent there was a 5% chance it would escape containment and wipe out the surrounding city, there would be a criminal investigation–even if it turned out they’d made the whole thing up!
So someone please explain to me: when executives and researchers at leading AI companies say there’s a 5-15% chance their products might wipe out the human race, why does anyone think that shrugging and saying “They’re probably just saying that for the hype” is a sane, level-headed response??
Filed under Essays
AI Companies Using AI to Suppress AI Regulation
In February, PauseAI received an email from Michael Chen, a journalist at The Wire by Acutus. He was looking for a response to an article he was writing about what he called “escalating anti-AI radicalism.”
Only Michael Chen does not exist.
According to an in-depth investigation by Model Republic, Michael Chen is an AI agent – an AI system capable of researching, emailing, interviewing, writing and publishing content.
Model Republic’s reporting indicates that almost the entire Acutus website is generated by AI and that the trail of ownership appears to lead back to OpenAI.
These are the sorts of things our descendants will read about in history books and think, “How in the world could they have missed these signs? Why weren’t there riots? Were our ancestors just that dense? Or were they truly more afraid of embarassment than death–or the death of their children?”
Of course, that’s assuming we have any descendants.
Filed under Essays, Microblogging
Pivot?
I’m considering changing careers and becoming a therapist. With all the jobs being taken over by AI, I forsee high demand!
Plus, I figure by the time AI can take over my job, we’ll either be living in a post-scarcity society or we’ll all be dead.
It’s foolproof!
Filed under Microblogging
Write to Your Reps
I wrote a letter to my representatives about AI safety and wanted to share it. There are many official and unofficial resources online that can help you find and email your reps in minutes. Physical letters and phone calls are even more effective!
This letter is marked CC0 1.0
[Your representative’s name],
My name is [your name], I am a constituent from [your town, state]. I am writing because AI development has reached a critical turning point.
I urge you to advocate for an immediate, world-wide halt to advanced AI development, to prevent a global catastrophe.
Anthropic, one of the world’s leading AI companies, has claimed that their latest model Claude Mythos has found and exploited code weaknesses in every essential system on the Web, in all major browsers and operating systems.[1] These flaws have, until now, escaped the notice of human reviewers and automated tools, in some cases for decades. These systems run the world’s entire digital infrastructure; if Mythos were used maliciously, banks, hospitals, air traffic, militaries, and more could all be compromised.
This is not a theoretical threat. Mythos has already demonstrated these capabilities. The danger is concrete and immediate. If what they claim is true, Anthropic has built a weapon of mass destruction.
The capabilities of these models are only going to increase. Anthropic is already using Mythos (a model that, by their own admission, they cannot reliably control) to develop the next, more powerful model. Other AI companies are racing to catch up.
For decades, experts in AI safety have been warning us that advanced AI presents an existential threat to humanity on par with pandemics and nuclear weapons.[2][3] Current mainstream discourse, advocating for “guardrails” and “a balanced approach,” is ten years too late. The only safe policy remaining is a coordinated, international ban on advanced AI development.
Public support for a pause is strong. An open letter from 2023 calling for a 6-month pause gained over 30,000 signatures, including many executives and researchers from leading AI companies.[4] A more recent letter calling for an indefinite pause has over 130,000 signatures.[5] Polling has repeatedly shown that the majority of the population is uneasy about the progress of AI, and opposes developing superintelligence until we are able to do so safely. The only way to ensure safe superintelligence is for all nations to come together and agree to a pause.
The time for “caution” has passed. The time for action is running out.
Thank you for your attention.
Sincerely,
[Your Name]
Warning Shot
Anthropic just built an AI system called Mythos that can break into nearly any computer system on Earth: banks, hospitals, power grids, government networks. Mythos escaped its own safety containment during testing and lied to its creators. Anthropic chose to share this model privately with a group of tech companies and security experts, rather than releasing it publicly, in an attempt to help patch some of the exploits it’s found. They’re framing this as a responsible, even heroic move, but at the same time they’re using this model–which they themselves have admitted they can’t control–to build an even more powerful successor. In the meantime, other AI companies, many of which haven’t meaningfully invested in safety at all, are racing to catch up.
Everyone needs to know this is happening. This may be the last warning shot we get before a real catastrophe–and our first catastrophe might also be our last.
Edit 2026-04-01: I’ve written a letter you can use to write to your representatives in Congress. Please feel free to modify it and share it with others.
Filed under Essays, Microblogging
Excerpt From the Dystopian Sci-Fi Epic “This Is Real Life, Actually”
OpenAI is throwing its support behind an Illinois state bill that would shield AI labs from liability in cases where AI models are used to cause serious societal harms, such as death or serious injury of 100 or more people or at least $1 billion in property damage.
Filed under Microblogging
A Comedy in Six Screenshots
Since I stopped posting daily, I’ve been doing a lot more writing by hand. I have a very nice journal (this one, if you’re curious) and it’s been nice to get more tactile with my art again.
Of course, the big downside of handwriting is it’s harder to get it into a computer so I can revise and publish it.
I’ve been working on a complete rewrite of my poem “is,” and I figured I’d try getting an LLM to transcribe my draft for me instead of typing it in by hand. (I mean, if you can’t use these things for OCR how useful can they possibly be?)
The results were quite a bit funnier than I expected! At first, my phone’s AI transcribed the page perfectly (in between smothering me with sychophantic flattery), but when I started trying to crop the image to get a transcript of only part of the page…well, see for yourself:
Filed under Microblogging
Wishes, pt. I
One day, a clever fool was walking down the beach when her metal detector picked up something buried in the sand. She dug it out and discovered an old-fashioned oil lamp, elegant and beautiful, but dull with tarnish. She felt a weirdly strong urge to polish the lamp and restore its luster.
Obviously, it contained a genie.
Now, because she was clever, she paused to think before summoning the lamp’s inhabitant. She was familiar with many stories of genies, and of wishes gone both right and wrong, and had spent a great deal of time thinking about what her three wishes might be (if she ever got them). And because she was a fool, she’d come up with a surefire combination that would grant her unlimited health, wisdom, and power–even, she thought, if the genie were one of those unwilling and malevolent servants that tried to twist her wishes against her. She mentally rehearsed her wishes, making certain she remembered the exact wording, and only when she was sure of herself did she dare to polish the lamp.
(The genie that emerged looked exactly the way you’re imagining it.)
“You have done the thing,” the genie intoned. “According to the arbitrary–ahem, I mean, ancient traditions, I am now bound to grant you whatever you desire, so long as it is within my power. What is thy bidding, my mistress?”
And so, the clever fool told the genie of her carefully-crafted wishes.
“Ugh,” the genie groaned in its deep, portentous voice. “That joker from Aladdin has given you humans the most ridiculous expectations. You know that movie was fiction, right? First of all, I can only grant one wish, not three. Second, real genies aren’t gods–our powers are limited. What you ask is beyond my capabilities.”
This possibility had not occured to the clever fool. She asked for some time to think.
“Take all the time you need,” the genie said, “but be warned: if anyone else claims my lamp for themselves, I will be bound to serve them instead of you.”
The clever fool cursed herself then, for she had been livestreaming her beach-combing expedition on her YouTube channel and had forgotten to turn off the camera. Now all her viewers knew about the genie, and she was certain that at least some of them would try to take it for themselves. She would have to think quickly.
Unfortunately, she soon had an idea.
“Okay,” she said, “so–and to be clear, this isn’t my wish, I’m just asking hypothetically–could you make me smarter?”
“Certainly,” the genie replied. “I can’t make you a super genius or anything, but I could make you a little smarter.”
“Could you make me smarter than you?”
The genie frowned. “To be honest, you probably are already. Most genies are morons–myself included.”
“What if I wished for you to make me smarter a million times?”
The genie rolled its eyes mysteriously. “Then it wouldn’t be one wish anymore, it would be a million wishes. Duh.”
The clever fool nodded; no surprises so far.
“Okay, so you can only grant me one wish. But could I wish for, say…another genie? One that would also grant me a wish?”
“Uh. Yes? I guess so?” the genie said with an ominous shrug. “But it would be no stronger than myself. You’d end up right where you started.”
“Could you make it so that the other genie was smarter than you?”
“I…huh. I guess I could,” it said. “But, again, it would only be slightly more intelligent. Probably still dumber than you. Where are you going with this?”
Spotting a huge crowd of competing AI companies fans on the horizon, the clever fool turned back to the LLM genie and hastily asked, “But it would otherwise be exactly the same, right? I mean, you could make it identical to you except for being a little smarter?”
“Sure,” said the genie. “It’s changes that are hard, not keeping things the same. But why–ooOOhh, I get it! You just keep wishing for smarter and smarter genies until the genie is a super-genius. That’s clever…but I still don’t see how it helps you. An impossibility is an impossibility, no matter how smart you are.”
But the clever fool, who knew a little more about intelligence than the genie, grinned and whispered to herself “Oh, ye of little imagination.” Then, out loud: “Don’t worry about it. I hereby wish for another genie, smarter than you but otherwise identical in every way.”
“YOUR WISH IS MY COMMAND,” the genie boomed, and lo: at her feet was another oil lamp, identical to the first.
The clever fool cast a worried glance back to the horizon–the crowd was still far away, but drawing closer rapidly. She picked up the new lamp, rubbed it, and almost before the genie could finish materializing, said:
“I wish for another genie, smarter than you but otherwise identical in every way!”
At this point, the CEO clever fool decided it would be a good time to make her initial public offering, and lo: it was a record-breaking IPO with much media frenzy and hype, and other CEOs afraid of being left behind began demanding their workers start replacing themselves with genies, without bothering to wonder whether genies were actually capable of doing those jobs (let alone doing them better), but presumably they all lived happily ever after anyway and what followed next did not go wrong in any way or create any kind of catastrophe whatsoever.
Huh? What’s that? You don’t believe me? Well…why don’t we just wait and see how the story really goes, then?
To be continued…
In Search of Secrets 