The first rule in talking to Cade Metz is do not talk to Cade Metz.
And that’s phrased as a joke, but I mean it seriously. He is out to harm us. Please stop helping him do so. Quokkas usually just don’t understand when someone is going to harm them, but in this case you don’t understand he is out to harm everyone here.
(Also, linking this comment may make this point more explicit.)
IMO there is a huge unilateralist curse here. But it’s worse because the people who want to talk to Cade Metz are not just more optimistic than average, but worse at it than average.
Kelsey is probably good, but Michael Vassar is nearly pessimal for the job: an actual manipulative cult leader type that Metz can see past.
I don’t think Metz can “see past” Michael Vassar. I think Metz is executing an adaptive strategy which boils down to do not engage with this individual. Metz is not modelling what Vassar believes in any detail; he is essentially just nodding along. This is a locally smart strategy given what happens to (some subset of) people who do engage with Vassar’s ideas, but it’s a rock-like strategy rather than an intelligent one.
As I mentioned in a previous interview, I think these labels get thrown around as insults without being based on anything substantive:
ZMD: So there’s also this problem of different groups slandering each other. I mentioned in the memoir that there were a couple years when I was collaborating with Michael Vassar and some of his friends on a lot of intellectual stuff. I still consider myself friends with Ben Hoffman and Jessica Taylor. I don’t talk to Michael very much these days. We chatted briefly at a conference the other month. But there’s some people who call us “the Vassarites” with cultish insinuations in a way that just sort of seems like slander, because from the inside, like, what? We’re different people who talk to each other sometimes, what is the alleged problem here, exactly?
Changed “cult leader” to “cult leader type”. I think the fact that he was manipulative and caused negative outcomes in several people including drug-induced psychosis is well-established.
the fact that he [...] caused negative outcomes in several people including drug-induced psychosis is well-established
No, it isn’t: see Scott Alexander’s followup comment in which he reports that he “interviewed many of the people involved” and that his “main conclusion is that I was wrong about Michael making people psychotic.”
situation considered complicated, calling it pure selection effect is cope at best, enabling at worst. calling it pure causality would be a ridiculous witch hunt. truth is messier and counterfactuals are hard.
What is this “us” that will be harmed by Zack talking to this guy? Game this out for me. Who is harmed, in what specific way, by Zack giving interviews like this?
I liked Jiro’s littering analogy. I assume the causal story goes, rats talk to Metz → Metz has an easier time writing articles and a book portraying rats as unreasonable/low-status (which don’t have to rise to the level of overtly hostile hit pieces) → people who trust New York Times reporters have lower opinions of rats → those people e.g. discriminate against us in hiring or support worse AI policies in expectation. I buy that there’s a real effect here; it just seems small enough that I don’t particularly feel like suppressing my natural inclination to talk to everyone and do cool meta-journalism.
My issue with this is, you can’t stop someone from writing a book. Did you think a professional journalist is going to shrug and give up if the group they’re profiling successfully coordinates not to talk to them? If I didn’t volunteer to be the star cooperative source, the book just gets written with David Gerard in that role, which is even worse.
And if David Gerard decided that Metz was unethical enough that even he wouldn’t talk to him. Metz could just invent a conversation with rationalists from scratch.
I don’t have an exact link to the article right now, but Scott once pointed out that even when people lie, the lies contain true elements when they could have just as well made up a complete fabrication. Refusing to talk with Metz increases the amount by which he has to lie, and makes his lies less convincing when someone catches him doing it, even if lying is still possible. If all he has to do is pick out a juicy quote which comes directly from a rationalist and just has to have the preceding sentence removed, you’ve made his tactics a lot easier than if he has to say something like “I spoke with David Gerard, a prominent critic of the so-called rationalists, who confirmed that...”, even if in theory you haven’t stopped him.
I also find this to be a case of a general rationalist flaw: refusing to take responsibility for making bad things easier on the grounds that they could have done it anyway, when in practice making it easier to do something that could have been done anyway drastically affects people’s behavior even if that seems to be illogical. Getting rid of trivial inconveniences matters.
(Also, if he talked with David Gerard specifically, there would be a lower chance of getting the lies into Wikipedia, since Gerard’s conflict of interest is known there.)
That’s the point. Metz could still lie if even Gerard wouldn’t talk to him, but not all lies are equally easy to get away with. Making his goal harder matters, even if strictly speaking he still has ways to reach it.
Okay, I think we both understand the problem with selective reporting. (An adversary who wants to minimize an audience’s estimate of the Heads-bias parameter of a biased coin can report only the Tails outcomes, and if the audience trusts them to be reporting fairly, that’s a huge distortion on what the audience ends up believing.) So this is probably a quantitative disagreement about what Metz’s goals and constraints are, and what my goals should be?
I understand and concede that in the limit of a sufficiently intelligent and sufficiently dishonest reporter with sufficiently many readers who trust him sufficiently much, whose only goal is to maximize reputational damage to their subject, the subject absolutely shouldn’t speak, because any information revealed is going to be perfectly cherry-picked and spun in a maximally damaging way to a maximal audience.
As I repeatedly express in the transcripts (“I’m hoping that I’m not an idiot for continuing to talk to you”, “why am I acting as if I believe him”, &c.), I have some uncertainty about whether I’m getting the calculus wrong here, but in order to update, I think I’m going to need more details than “he is out to harm everyone here.”
Certainly, he has no qualms about doing harm to people here. (We see that in his reaction to the Slate Star affair, where he displays no remorse for his role in Scott Alexander leaving his job.) But I think there’s an important difference between intended harm and indifference to harm in the course of reporting on a real story (that an ideological internet community is wielding influence in Silicon Valley).
Maybe it would help to be concrete about what harm you foresee? My median expectation here is that a book gets published in 2027; it kind of sucks (similar to how Metz’s previous reporting kind of sucked) and sells maybe 40K copies. The harm dealt to users of this website (in the form of people reading it and treating us worse, or making worse decisions about AI safety) is plausibly less than that of Metz’s August 2025 piece on Lighthaven and the Secular Solstice (which, to be clear, kind of sucked), because more people will read a New York Times article than buy an entire book.
Did your life get noticeably worse as a result of that article (which, as far as I can tell, didn’t even make the print edition)? Will you forever rue the date of 4 August 2025 as the day of an unspeakable injustice and tragedy, a wound in the world that can never be healed? Maybe this seems monstrous to you, but I just find it really hard to care about such diffuse purported harms compared to the immediate benefit of me getting to publish a cool meta-journalism blog post.
But I think there’s an important difference between intended harm and indifference to harm in the course of reporting on a real story
Imagine a suitably anthromorphized lion wanting to eat a sheep. Does the lion intend harm, or is the lion merely indifferent to harm? You can argue that the lion just wants dinner. The fact that the sheep has to be killed in order to become dinner just happens to be a step towards the lion’s ultimate goal. If the sheep was harmed in some other fashion, let’s say being thrown into a volcano, the lion would have no desire for that to happen at all! And if the lion managed to eat his fill in some other way, he might not even care about the sheep (until he got hungry again).
So clearly the lion doesn’t actually intend harm to the sheep. It just wants dinner. It’s indifferent to the harm that happens to the sheep in service of its goal.
But if you’re a (suitably anthromorphized) sheep in a place full of lions, you still had better take measures against lions eating you.
In some literal sense, Cade Metz is merely indifferent to the harm he causes to rationalists by lying about them. But he’s still a threat. His terminal goal is incompatible with rationalists’ welfare, even if he wouldn’t throw rationalists into a volcano just because it causes them harm.
Maybe it would help to be concrete about what harm you foresee?
The problem with this question is that the exact details of the harm can be wrong in a way that lets you nitpick the answer, but which has no relevance to the question “how should you deal with him?”. My rough guess is that he is a particular kind of leftist who is ideologically hostile to techbros and the tech industry and associates them with fascists, and is going to spin anything you say along those lines whether it is truthful or not. But the exact details of what he’s going to do and how he’s going to do it don’t matter, because harm based on one ideology is pretty similar to harm based on a somewhat different ideology.
Did your life get noticeably worse as a result of that article
Does anyone’s life get worse as a result of throwing one piece of litter on the ground in the park?
Talking to hostile media is a collective action problem.
As I repeatedly express in the transcripts (“I’m hoping that I’m not an idiot for continuing to talk to you”, “why am I acting as if I believe him”, &c.), I have some uncertainty about whether I’m getting the calculus wrong here,
This is itself something you are messing up. It’s the equivalent of the sheep saying “Nice juicy mutton right here.” You seem to think that being half-hearted about admitting your naivety only causes half as much harm than completely admitting your naivety. It doesn’t work that way; doing it halfway hurts you about as much as doing it fully.
The first rule in talking to Cade Metz is do not talk to Cade Metz.
And that’s phrased as a joke, but I mean it seriously. He is out to harm us. Please stop helping him do so. Quokkas usually just don’t understand when someone is going to harm them, but in this case you don’t understand he is out to harm everyone here.
(Also, linking this comment may make this point more explicit.)
IMO there is a huge unilateralist curse here. But it’s worse because the people who want to talk to Cade Metz are not just more optimistic than average, but worse at it than average.
Kelsey is probably good, but Michael Vassar is nearly pessimal for the job: an actual manipulative cult leader type that Metz can see past.
I don’t think Metz can “see past” Michael Vassar. I think Metz is executing an adaptive strategy which boils down to do not engage with this individual. Metz is not modelling what Vassar believes in any detail; he is essentially just nodding along. This is a locally smart strategy given what happens to (some subset of) people who do engage with Vassar’s ideas, but it’s a rock-like strategy rather than an intelligent one.
As I mentioned in a previous interview, I think these labels get thrown around as insults without being based on anything substantive:
Changed “cult leader” to “cult leader type”. I think the fact that he was manipulative and caused negative outcomes in several people
includingdrug-induced psychosisis well-established.No, it isn’t: see Scott Alexander’s followup comment in which he reports that he “interviewed many of the people involved” and that his “main conclusion is that I was wrong about Michael making people psychotic.”
situation considered complicated, calling it pure selection effect is cope at best, enabling at worst. calling it pure causality would be a ridiculous witch hunt. truth is messier and counterfactuals are hard.
Thanks for the correction, I try to avoid drama and sometimes it means I miss correction comments and such.
I’m still wary of the process but don’t really want to engage more
What is this “us” that will be harmed by Zack talking to this guy? Game this out for me. Who is harmed, in what specific way, by Zack giving interviews like this?
I liked Jiro’s littering analogy. I assume the causal story goes, rats talk to Metz → Metz has an easier time writing articles and a book portraying rats as unreasonable/low-status (which don’t have to rise to the level of overtly hostile hit pieces) → people who trust New York Times reporters have lower opinions of rats → those people e.g. discriminate against us in hiring or support worse AI policies in expectation. I buy that there’s a real effect here; it just seems small enough that I don’t particularly feel like suppressing my natural inclination to talk to everyone and do cool meta-journalism.
Rats talk to Metz → Metz has an easier time writing about rats → more people learn about rats → there are more rats → rats talk to Metz
My issue with this is, you can’t stop someone from writing a book. Did you think a professional journalist is going to shrug and give up if the group they’re profiling successfully coordinates not to talk to them? If I didn’t volunteer to be the star cooperative source, the book just gets written with David Gerard in that role, which is even worse.
And if David Gerard decided that Metz was unethical enough that even he wouldn’t talk to him. Metz could just invent a conversation with rationalists from scratch.
I don’t have an exact link to the article right now, but Scott once pointed out that even when people lie, the lies contain true elements when they could have just as well made up a complete fabrication. Refusing to talk with Metz increases the amount by which he has to lie, and makes his lies less convincing when someone catches him doing it, even if lying is still possible. If all he has to do is pick out a juicy quote which comes directly from a rationalist and just has to have the preceding sentence removed, you’ve made his tactics a lot easier than if he has to say something like “I spoke with David Gerard, a prominent critic of the so-called rationalists, who confirmed that...”, even if in theory you haven’t stopped him.
I also find this to be a case of a general rationalist flaw: refusing to take responsibility for making bad things easier on the grounds that they could have done it anyway, when in practice making it easier to do something that could have been done anyway drastically affects people’s behavior even if that seems to be illogical. Getting rid of trivial inconveniences matters.
(Also, if he talked with David Gerard specifically, there would be a lower chance of getting the lies into Wikipedia, since Gerard’s conflict of interest is known there.)
No, he couldn’t. Journalists who outright make stuff up (as contrasted to engaging in mere “spin”) get fired in disgrace.
That’s the point. Metz could still lie if even Gerard wouldn’t talk to him, but not all lies are equally easy to get away with. Making his goal harder matters, even if strictly speaking he still has ways to reach it.
Okay, I think we both understand the problem with selective reporting. (An adversary who wants to minimize an audience’s estimate of the Heads-bias parameter of a biased coin can report only the Tails outcomes, and if the audience trusts them to be reporting fairly, that’s a huge distortion on what the audience ends up believing.) So this is probably a quantitative disagreement about what Metz’s goals and constraints are, and what my goals should be?
I understand and concede that in the limit of a sufficiently intelligent and sufficiently dishonest reporter with sufficiently many readers who trust him sufficiently much, whose only goal is to maximize reputational damage to their subject, the subject absolutely shouldn’t speak, because any information revealed is going to be perfectly cherry-picked and spun in a maximally damaging way to a maximal audience.
Given that foundation, I think there’s an empirical question of how close Metz is to that infinite limit along each of those dimensions—whether he’s close enough that the act-consequentialist cost of reputational damage to “the community” (which I have common interests with) outweighs my non-act-consequentialist committment to transparency and the benefit of getting to publish cool meta-journalism blog posts.
As I repeatedly express in the transcripts (“I’m hoping that I’m not an idiot for continuing to talk to you”, “why am I acting as if I believe him”, &c.), I have some uncertainty about whether I’m getting the calculus wrong here, but in order to update, I think I’m going to need more details than “he is out to harm everyone here.”
Certainly, he has no qualms about doing harm to people here. (We see that in his reaction to the Slate Star affair, where he displays no remorse for his role in Scott Alexander leaving his job.) But I think there’s an important difference between intended harm and indifference to harm in the course of reporting on a real story (that an ideological internet community is wielding influence in Silicon Valley).
Maybe it would help to be concrete about what harm you foresee? My median expectation here is that a book gets published in 2027; it kind of sucks (similar to how Metz’s previous reporting kind of sucked) and sells maybe 40K copies. The harm dealt to users of this website (in the form of people reading it and treating us worse, or making worse decisions about AI safety) is plausibly less than that of Metz’s August 2025 piece on Lighthaven and the Secular Solstice (which, to be clear, kind of sucked), because more people will read a New York Times article than buy an entire book.
Did your life get noticeably worse as a result of that article (which, as far as I can tell, didn’t even make the print edition)? Will you forever rue the date of 4 August 2025 as the day of an unspeakable injustice and tragedy, a wound in the world that can never be healed? Maybe this seems monstrous to you, but I just find it really hard to care about such diffuse purported harms compared to the immediate benefit of me getting to publish a cool meta-journalism blog post.
Imagine a suitably anthromorphized lion wanting to eat a sheep. Does the lion intend harm, or is the lion merely indifferent to harm? You can argue that the lion just wants dinner. The fact that the sheep has to be killed in order to become dinner just happens to be a step towards the lion’s ultimate goal. If the sheep was harmed in some other fashion, let’s say being thrown into a volcano, the lion would have no desire for that to happen at all! And if the lion managed to eat his fill in some other way, he might not even care about the sheep (until he got hungry again).
So clearly the lion doesn’t actually intend harm to the sheep. It just wants dinner. It’s indifferent to the harm that happens to the sheep in service of its goal.
But if you’re a (suitably anthromorphized) sheep in a place full of lions, you still had better take measures against lions eating you.
In some literal sense, Cade Metz is merely indifferent to the harm he causes to rationalists by lying about them. But he’s still a threat. His terminal goal is incompatible with rationalists’ welfare, even if he wouldn’t throw rationalists into a volcano just because it causes them harm.
The problem with this question is that the exact details of the harm can be wrong in a way that lets you nitpick the answer, but which has no relevance to the question “how should you deal with him?”. My rough guess is that he is a particular kind of leftist who is ideologically hostile to techbros and the tech industry and associates them with fascists, and is going to spin anything you say along those lines whether it is truthful or not. But the exact details of what he’s going to do and how he’s going to do it don’t matter, because harm based on one ideology is pretty similar to harm based on a somewhat different ideology.
Does anyone’s life get worse as a result of throwing one piece of litter on the ground in the park?
Talking to hostile media is a collective action problem.
This is itself something you are messing up. It’s the equivalent of the sheep saying “Nice juicy mutton right here.” You seem to think that being half-hearted about admitting your naivety only causes half as much harm than completely admitting your naivety. It doesn’t work that way; doing it halfway hurts you about as much as doing it fully.