Six months ago, I was a high school English teacher.
I wasn’t looking to change careers, even after nineteen sometimes-difficult years. I was good at it. I enjoyed it. After long experimentation, I had found ways to cut through the nonsense and provide real value to my students. Daily, I met my nemesis, Apathy, in glorious battle, and bested her with growing frequency. I had found my voice.
At MIRI, I’m still struggling to find my voice, for reasons my colleagues have invited me to share later in this post. But my nemesis is the same.
Apathy will be the death of us. Indifference about whether this whole AI thing goes well or ends in disaster. Come-what-may acceptance of whatever awaits us at the other end of the glittering path. Telling ourselves that there’s nothing we can do anyway. Imagining that some adults in the room will take care of the problem, even if we don’t see any such adults.
Perhaps you’ve felt her insidious pull on your psyche. I think we all have. This AI stuff is cool. Giving in to the “thermodynamic god”, to She-Who-Can’t-Be-Bothered, would be so much easier than the alternative, and probably a lot more fun (while it lasted).
And me? I was an English teacher. What could I do?
A little! I could donate and volunteer, as I did in modest fashion to MIRI’s predecessor organization for a while even before taking my first teaching contract. I could make sure my students and coworkers knew at least one person who was openly alarmed about AI. And I could find easy dignity in being ready to answer a call from MIRI that would realistically never come.
You can guess the rest. The universe called my bluff. I scrambled to make it not a bluff. And here I am on MIRI’s growing comms team!
It was a near thing, though. When MIRI posted about the open position, I almost looked away.
I think about that a lot now, especially on the hard days: guessing at the amount of history that was made by people who almost stayed in bed, and about how much history almost happened but for a last-minute “Nah.”
We stand on the shoulders of giants who didn’t have to be. If we are to honor their legacy, we have to be the adults we want to see in the room. That sounds like work because it is. The adversary draws strength from this, for Apathy is Lazy’s child.
Apathy does not ‘adult’. She wears the fashions and plays the status games, but she doesn’t do the math. She doesn’t change her mind, or check her sources. She doesn’t reach across the aisle. When asked to choose between speaking up and saving face, she picks ‘face’ every time.
I don’t think the world would be racing off the AI cliff if there were more adults and less apathy. There really aren’t that many players pushing the frontier, and most bystanders don’t like what they see when they find the motivation to look. Don’t build things smarter than us until we’re ready. What’s so hard about that? As your former English teacher in spirit, I’m obligated to quote Atticus Finch here: “This case is not a difficult one, it requires no minute sifting of complicated facts, but it does require you to be sure beyond reasonable doubt...”
My star students will reflect that Atticus lost that case, thanks to jurors who picked ‘save face’ over ‘speak up’. But it was a near thing! They almost bent the “long moral arc” towards justice before they said, “Nah.”
Unfortunately, we get no points for almost. We don’t survive the arrival of AIs that outclass humanity by almost doing the grown-up thing, by almost not sleepwalking through the threshold before the metaphorical spinning blades have been disabled.
MIRI does communications in hard mode because the universe is set to hard mode. We need our words to be understood to mean what they say. Hard mode is hard! Being adults in this line of work cuts against a lot of human instinct and runs counter to common norms around PR and outreach. Some strictures this entails:
We don’t mince words. We speak up, even if this must sometimes come at the expense of respectability. We want to come across exactly as alarmed as the situation warrants (and no more). We avoid euphemisms and needlessly academic stylings.
We don’t say things we don’t believe are true. We don’t just repeat things we’ve heard. We don’t invent explanations for things we don’t understand. We avoid unhelpful speculation, and label our conjectures as such. We express our uncertainty.
We don’t employ sophistry. We don’t knowingly use flawed or weak arguments, and we scrutinize our work to root them out.
We don’t use haggling tactics. We call for what we think is needed — no more, no less. If we say we need an international halt to the development of advanced general AI systems, and that it might need to last decades, it’s because we don’t think anything less will do — not because we hope a slate of compromise regulations will suffice, or because we think a halt is a stepping stone to the policy we actually want.
We don’t play 4-D chess. We don’t conceal our true intentions. We take actions because we think the primary reactions will be helpful. This doesn’t mean we don’t think about nth order consequences — I’m in awe of how hard MIRI thinks about these! — but we don’t employ strategies where success depends on reactions to reactions to reactions (etc.) going as planned.
We don’t chase applause. Useful action, not approval, is the unit of progress. Crowd-pleasing language interferes with critical thinking and signals that a piece is seeking praise rather than action.
We leave our tribal politics at home. This doesn’t just mean avoiding pot-shots on our outgroups; it means avoiding any appearance of wanting to do so — and even the “mind-killing” distraction of phrasing that might evoke partisan thoughts.
We can’t claim to get it right every time, but boy do we try. That’s the best explanation I can offer as to why you may find the volume of our public output disappointing these last six months. (I know I do!) Respectability may be overrated, but credibility is not, and we really don’t want to blow it.
I’ve found walking this line even more challenging than it sounds. As someone with rationalist sensibilities but a career in public education, I was long used to being seen as the most careful thinker in the room. But being in an actually careful room is an adjustment, especially for someone who really wants our work to be read, and therefore tries to make it engaging. I haven’t been shut down on this push — we’re trying to broaden our reach, after all — but the whole team is struggling to apply our very high standards to writing for audiences unready or unwilling to work through dense passages with precise terminology.
Part of the puzzle is that such readers aren’t used to writers playing by our rules, and may not recognize that we are doing so. When we write approachably and engagingly, we fear readers concluding that we must be trying to charm them or sell them something; this would cause them to apply adjustment factors for marketing-speak, and to read between the lines for hidden motives that aren’t there.
I’ve found it helpful to recognize that we’re barely in the persuasion business at all. We’re showing you the facts as we understand them, and think our suggested course of action will be as logical to you as it is to us. Water flows downhill. Enriched uranium in a sufficiently large pile will go critical. Powerful AIs built with existing methods will pursue goals incompatible with our survival.
Admittedly, working out the details of enforceable global agreements to prevent AI catastrophe are going to be more complicated than the logic behind them. Laying some of the groundwork for this effort is one objective of our new Technical Governance Team.
Helping more readers understand why smarter-than-human AI is lethal on the current path is also going to be more complicated. That task falls to us in comms, in collaboration with our senior team. We’re working on it. (We’ve got a few medium-to-large writing projects that we’ve been working on this year, with commensurate care; if any of them turn out to be effective then they will retroactively be braggable 2024 accomplishments.)
But yes, our bar is so high we struggle to get over it. We aren’t shipping enough yet. We’re building capacity and working through some growing pains associated with our pivot to comms. We’re learning, though, and we want to be learning faster. To this end, you should expect more experimentation from us in 2025.
This post is an early taste of that. Establishing more personal voices for MIRI comms will let us try more approaches while owning our inevitable errors as individuals. Seems worth a try.
So… Hi! I’m Mitch. I almost decided not to try.
If you’re maybe just-this-side-of-almost, I invite you to try, too. It doesn’t have to be your career, or a donation (though MIRI and its fellow travelers welcome both). But please, strike a blow against Apathy! One place to start: Stare at the AI problem for a while, if you haven’t already, and then take the slightly awkward, slightly emotional, slightly effortful step of telling your friends and colleagues what you see.
Don’t feel bad about depriving Her Royal Numbness of one of her subjects. She’ll be fine. More to the point, she won’t care.
That’s why we can win. We just have to want it more than she does.
Communications in Hard Mode (My new job at MIRI)
Six months ago, I was a high school English teacher.
I wasn’t looking to change careers, even after nineteen sometimes-difficult years. I was good at it. I enjoyed it. After long experimentation, I had found ways to cut through the nonsense and provide real value to my students. Daily, I met my nemesis, Apathy, in glorious battle, and bested her with growing frequency. I had found my voice.
At MIRI, I’m still struggling to find my voice, for reasons my colleagues have invited me to share later in this post. But my nemesis is the same.
Apathy will be the death of us. Indifference about whether this whole AI thing goes well or ends in disaster. Come-what-may acceptance of whatever awaits us at the other end of the glittering path. Telling ourselves that there’s nothing we can do anyway. Imagining that some adults in the room will take care of the problem, even if we don’t see any such adults.
Perhaps you’ve felt her insidious pull on your psyche. I think we all have. This AI stuff is cool. Giving in to the “thermodynamic god”, to She-Who-Can’t-Be-Bothered, would be so much easier than the alternative, and probably a lot more fun (while it lasted).
And me? I was an English teacher. What could I do?
A little! I could donate and volunteer, as I did in modest fashion to MIRI’s predecessor organization for a while even before taking my first teaching contract. I could make sure my students and coworkers knew at least one person who was openly alarmed about AI. And I could find easy dignity in being ready to answer a call from MIRI that would realistically never come.
You can guess the rest. The universe called my bluff. I scrambled to make it not a bluff. And here I am on MIRI’s growing comms team!
It was a near thing, though. When MIRI posted about the open position, I almost looked away.
I think about that a lot now, especially on the hard days: guessing at the amount of history that was made by people who almost stayed in bed, and about how much history almost happened but for a last-minute “Nah.”
We stand on the shoulders of giants who didn’t have to be. If we are to honor their legacy, we have to be the adults we want to see in the room. That sounds like work because it is. The adversary draws strength from this, for Apathy is Lazy’s child.
Apathy does not ‘adult’. She wears the fashions and plays the status games, but she doesn’t do the math. She doesn’t change her mind, or check her sources. She doesn’t reach across the aisle. When asked to choose between speaking up and saving face, she picks ‘face’ every time.
I don’t think the world would be racing off the AI cliff if there were more adults and less apathy. There really aren’t that many players pushing the frontier, and most bystanders don’t like what they see when they find the motivation to look. Don’t build things smarter than us until we’re ready. What’s so hard about that? As your former English teacher in spirit, I’m obligated to quote Atticus Finch here: “This case is not a difficult one, it requires no minute sifting of complicated facts, but it does require you to be sure beyond reasonable doubt...”
My star students will reflect that Atticus lost that case, thanks to jurors who picked ‘save face’ over ‘speak up’. But it was a near thing! They almost bent the “long moral arc” towards justice before they said, “Nah.”
Unfortunately, we get no points for almost. We don’t survive the arrival of AIs that outclass humanity by almost doing the grown-up thing, by almost not sleepwalking through the threshold before the metaphorical spinning blades have been disabled.
MIRI does communications in hard mode because the universe is set to hard mode. We need our words to be understood to mean what they say. Hard mode is hard! Being adults in this line of work cuts against a lot of human instinct and runs counter to common norms around PR and outreach. Some strictures this entails:
We don’t mince words. We speak up, even if this must sometimes come at the expense of respectability. We want to come across exactly as alarmed as the situation warrants (and no more). We avoid euphemisms and needlessly academic stylings.
We don’t say things we don’t believe are true. We don’t just repeat things we’ve heard. We don’t invent explanations for things we don’t understand. We avoid unhelpful speculation, and label our conjectures as such. We express our uncertainty.
We don’t employ sophistry. We don’t knowingly use flawed or weak arguments, and we scrutinize our work to root them out.
We don’t use haggling tactics. We call for what we think is needed — no more, no less. If we say we need an international halt to the development of advanced general AI systems, and that it might need to last decades, it’s because we don’t think anything less will do — not because we hope a slate of compromise regulations will suffice, or because we think a halt is a stepping stone to the policy we actually want.
We don’t play 4-D chess. We don’t conceal our true intentions. We take actions because we think the primary reactions will be helpful. This doesn’t mean we don’t think about nth order consequences — I’m in awe of how hard MIRI thinks about these! — but we don’t employ strategies where success depends on reactions to reactions to reactions (etc.) going as planned.
We don’t chase applause. Useful action, not approval, is the unit of progress. Crowd-pleasing language interferes with critical thinking and signals that a piece is seeking praise rather than action.
We leave our tribal politics at home. This doesn’t just mean avoiding pot-shots on our outgroups; it means avoiding any appearance of wanting to do so — and even the “mind-killing” distraction of phrasing that might evoke partisan thoughts.
We can’t claim to get it right every time, but boy do we try. That’s the best explanation I can offer as to why you may find the volume of our public output disappointing these last six months. (I know I do!) Respectability may be overrated, but credibility is not, and we really don’t want to blow it.
I’ve found walking this line even more challenging than it sounds. As someone with rationalist sensibilities but a career in public education, I was long used to being seen as the most careful thinker in the room. But being in an actually careful room is an adjustment, especially for someone who really wants our work to be read, and therefore tries to make it engaging. I haven’t been shut down on this push — we’re trying to broaden our reach, after all — but the whole team is struggling to apply our very high standards to writing for audiences unready or unwilling to work through dense passages with precise terminology.
Part of the puzzle is that such readers aren’t used to writers playing by our rules, and may not recognize that we are doing so. When we write approachably and engagingly, we fear readers concluding that we must be trying to charm them or sell them something; this would cause them to apply adjustment factors for marketing-speak, and to read between the lines for hidden motives that aren’t there.
I’ve found it helpful to recognize that we’re barely in the persuasion business at all. We’re showing you the facts as we understand them, and think our suggested course of action will be as logical to you as it is to us. Water flows downhill. Enriched uranium in a sufficiently large pile will go critical. Powerful AIs built with existing methods will pursue goals incompatible with our survival.
Admittedly, working out the details of enforceable global agreements to prevent AI catastrophe are going to be more complicated than the logic behind them. Laying some of the groundwork for this effort is one objective of our new Technical Governance Team.
Helping more readers understand why smarter-than-human AI is lethal on the current path is also going to be more complicated. That task falls to us in comms, in collaboration with our senior team. We’re working on it. (We’ve got a few medium-to-large writing projects that we’ve been working on this year, with commensurate care; if any of them turn out to be effective then they will retroactively be braggable 2024 accomplishments.)
But yes, our bar is so high we struggle to get over it. We aren’t shipping enough yet. We’re building capacity and working through some growing pains associated with our pivot to comms. We’re learning, though, and we want to be learning faster. To this end, you should expect more experimentation from us in 2025.
This post is an early taste of that. Establishing more personal voices for MIRI comms will let us try more approaches while owning our inevitable errors as individuals. Seems worth a try.
So… Hi! I’m Mitch. I almost decided not to try.
If you’re maybe just-this-side-of-almost, I invite you to try, too. It doesn’t have to be your career, or a donation (though MIRI and its fellow travelers welcome both). But please, strike a blow against Apathy! One place to start: Stare at the AI problem for a while, if you haven’t already, and then take the slightly awkward, slightly emotional, slightly effortful step of telling your friends and colleagues what you see.
Don’t feel bad about depriving Her Royal Numbness of one of her subjects. She’ll be fine. More to the point, she won’t care.
That’s why we can win. We just have to want it more than she does.
Shouldn’t be too hard.
(MIRI blog mirror)