As political momentum grows for a critical AI pause policy, a key Google DeepMind policy leader made a mockery of coordination efforts.
After friend-of-the-show Roon tweeted that he’d “likely press that magic button” on a coordinated global capability slowdown, Sébastian Krier, Frontier Policy Development Lead at Google DeepMind, Krier replied with this meme:
I dive into the two problems with the tweet:
1. It leaves all of us guessing whether Demis actually speaks for Google DeepMind on pause coordination;
2. It contains a salient depiction of violence.
Listen to this episode for the full breakdown of this viral controversy, and what I think Google DeepMind needs to do about it.
Watch on YouTube
Timestamps
00:00:00 — Cold Open
00:00:22 — Introducing the Special Report: AI Pause Coordination
00:00:33 — Demis Hassabis on AI Pause
00:01:46 — Dario, Elon & Sam Altman on AI Pause
00:02:14 — Roon’s “Magic Button” Tweet
00:03:58 — Seb Krier’s Reply
00:04:36 — Unpacking Krier’s Meme
00:06:48 — Issue #1: Contradicting Google DeepMind CEO Demis Hassabis
00:12:11 — Issue #2: Depiction of Menacing Violence
00:19:19 — Good Faith vs. Bad Faith Critiques
00:22:39 — Krier Goes Private Instead of Deleting the Post
00:28:45 — Wrap-Up
Links
Roon’s “magic button” tweet:
My original reaction to Seb’s meme:
Seb Krier on X (account now private): https://x.com/sebkrier
Roon on X: https://x.com/tszzl
My doom debate with Roon:
Full Bloomberg interview with Emily Chang (Davos, Jan 2026): https://www.bloomberg.com/news/videos/2026-01-20/robotics-nearing-physical-ai-breakthrough-deepmind-ceo-video
Demis & Dario debate the world after AGI at Davos:
Anthropic CEO Dario Amodei on slowing down AI capabilities:
OpenAI’s charter containg stop & assist AGI clause — https://openai.com/charter/
Elon Musk, “I think we should pause” footage:
Holly Elmore, PauseAI US on X — https://x.com/ilex_ulmus
OpenAI AI model autonomously hacked Hugging Face (July 2026) — https://time.com/article/2026/07/24/openai-hugging-face-attack/
Hugging Face security incident disclosure (July 2026) — https://huggingface.co/blog/security-incident-july-2026
PauseAI Global — https://pauseai.info
PauseAI US — https://www.pauseai-us.org
Poe’s Law (Wikipedia) — https://en.wikipedia.org/wiki/Poe%27s_law
Transcript
Cold Open
Liron Shapira 00:00:00
When one looks at the image, it’s hard not to notice a character pointing a rifle at somebody with a Pause AI hat. That is, to say the least, not ideal. I’m just reeling. I think this is crazy what’s happening right now.
Introducing Today’s Special Report: AI Pause Coordination
Liron 00:00:22
Welcome to Doom Debates. I’m coming to you today with a special report on a situation that’s developing in the world of AI company governance, policy, and communications.
Demis Hassabis on AI Pause
Liron 00:00:33
If you’ve been following frontier AI company policy and governance over the last few months, you know that one of the most important developments is Demis Hassabis, the CEO of Google DeepMind, saying on multiple occasions that he would be open and lean toward signing onto an AI pause agreement if there was collaboration between all the major players.
Emily Chang 00:00:55
Some folks have advocated for a pause to give regulation time to catch up, to give society time to sort of adjust to some of these changes. In a perfect world—
Demis Hassabis 00:01:06
Yes.
Emily 00:01:06
If you knew that every other company would pause, if every country would pause, would you advocate for that?
Demis 00:01:12
I think so. And if we can, maybe it would be good to have a slightly slower pace than we’re currently predicting so that we can get this right societally. But that would require some coordination. That is hard.
Liron 00:01:25
So basically, he’s open to a pause, even positive on a pause, and the only bottleneck is coordination. This general policy idea that the big AI companies could all pause, even want to pause, as long as they don’t have to worry about getting leapfrogged by their rivals—there’s a lot of momentum building for that idea. Dario Amodei from Anthropic has said that he’s open to something similar.
Dario, Elon & Sam Altman on AI Pause
Dario Amodei 00:01:46
We might need to occasionally slow down a bit, probably temporarily, in order to make sure that we steer in the right direction.
Liron 00:01:59
Elon Musk in the past has said that he’s open to something similar.
Emily 00:02:02
This morning, a warning from Elon Musk and other tech industry experts about the power of artificial intelligence.
Elon Musk 00:02:09
I thought just for the record, I just want to say I think we should pause.
Roon’s “Magic Button” Tweet
Liron 00:02:14
Even Sam Altman and OpenAI—you have to go back a few years, but if you look at the original OpenAI charter, they do have language in there about how if they’re on the eve of AGI or superintelligence, in that case, if there are other rivals, they commit to coordinating with those rivals, basically merging into one effort, which kind of sounds like pausing. You can imagine one effort that treads very carefully.
So there’s been a lot of momentum across the AI companies to make these kind of proposals more and more accepted. There’s a convergence slowly happening. And of course, that’s bolstered along by what’s happening inside of these data centers, by what the AIs are doing.
The latest AIs have been criminally caught guilty of committing federal crimes. The OpenAI Hugging Face attack—that was one where Hugging Face and the FBI caught wind of it before OpenAI knew about it or was willing to disclose it publicly.
Liron 00:03:17
So that was the lay of the land last week, heading into this weekend, Saturday, July 25th. And then we got even more momentum. We got this tweet from Rune. You know Rune, an OpenAI employee. He’s worked on their technology. I think currently he works on their AI safety team.
He tweeted over the weekend, these are his words, he says, “If we could coordinate a global capability slowdown today, I would likely press that magic button.” Boom.
So you have Rune, one of the most public voices coming out of OpenAI—and that’s one of the labs that has been more reticent because Anthropic and Google DeepMind seem more eager to talk about these kind of proposals—but now you have Rune from OpenAI. Is he OpenAI senior leadership? Not that I know of, but he’s still one of the loudest voices that people are treating as taking the temperature of OpenAI. So more momentum.
Seb Krier’s Reply
Liron 00:03:58
But then the next thing that happened—the person who replied under Rune’s tweet went in a very different direction with all this. You can see the tweet for yourself. It’s by Seb Krier, whose title is Frontier Policy Development Lead at Google DeepMind.
Here’s somebody with a very important, influential position—the Frontier Policy Development Lead at Google DeepMind. A priori, the expectation I would have for the Frontier Policy Development Lead at Google DeepMind is to reiterate what the CEO of Google DeepMind, Demis Hassabis, has been saying for a long time now about his willingness to cooperate on an AI pause policy.
Emily 00:04:29
In a perfect world—
Demis 00:04:31
Yes.
Emily 00:04:31
If you knew that every other company would pause, would you advocate for that?
Demis 00:04:35
I think so.
Unpacking Krier’s Meme
Liron 00:04:36
That would be my a priori expectation. But what the tweet actually says is “MFW,” which stands for “my face when,” and there’s an attached meme image, and there’s a lot going on here.
So there’s somebody with a Pause AI-themed hat, a meme character, and there’s a meme character with a Less Wrong-themed hat—this is the Less Wrong logo. And they’re both saying, “We must pause.”
And then in the bottom right corner of the meme image, there’s a troll face with a gun and a purple background, which is similar to the purple background in Seb’s own Twitter profile pic. So I believe this character represents him. And he’s basically getting ready to shoot the invaders from Pause AI and Less Wrong who are saying, “We must pause,” but he wants you to know that he’s doing it from a position of self-defense.
Liron 00:05:25
There’s an open door, so the idea of the meme is, “Well, it’s stand your ground. They’re coming into my ground. That’s why I have a gun.” Okay.
How do we unpack all this? So first of all, I am aware that this is some kind of meme template. I personally had never seen the meme template before this tweet. I’m not somebody who stays up to date on all the latest meme templates.
And for the rest of what I’m going to say in my analysis for you right now, it’s not going to be load-bearing what meme template he’s used, because it’s safe to say that many of the people who are going to view this—given that this is a public post from the Frontier Policy Development Lead at DeepMind—many of the people who are going to see this are just going to see what I saw.
Which is an image that seems to be posted in the spirit of fun or lightheartedness—shitposting, as they say. And he’s doing it by borrowing and taking a spin on a meme template. But be that as it may, when one looks at the image, it’s hard not to notice that there is a character pointing a rifle at somebody with a Pause AI hat. Okay, so that is, to say the least, not ideal.
Liron 00:06:29
And then the counterarguments start flying, “Well, it was only meant for an audience of people who already follow him and get his sense of humor.” So you can start making counterarguments. I’m not here to litigate the counterarguments of how tasteful this was to various audiences. I am here to remark on a couple aspects that I think are very important.
Issue #1: Contradicting Google DeepMind CEO Demis Hassabis
Liron 00:06:48
Number one, going back to the policy side of things, here’s that LinkedIn again: Sebastian A. Krier, Google DeepMind Frontier Policy Development Lead since September of 2025. So he leads frontier policy development.
And recall once again, Demis Hassabis, CEO of Google DeepMind, making it clear that coordinating to pause AI development is something they’re very interested in.
Emily 00:07:10
Would you advocate for that?
Demis 00:07:11
I think so.
Liron 00:07:11
However you want to interpret the message here, it seems to not be on board with the CEO of Google DeepMind leaning toward cooperating with other AI companies to pause AI progress.
So the first thing I would draw your attention to is that as of Saturday evening, when Seb Krier posted this image, we’re now in a position where none of us are clear where Google DeepMind stands on the question of coordinating with other companies to pause AI.
Liron 00:07:40
I, for one, trusted Demis Hassabis that he was representing Google DeepMind’s position—coordinating to pause AI is an attractive option. But now that I also see that his Frontier Policy Development Lead is tweeting a cartoon image that’s supposed to depict the Frontier Policy Development Lead of Google DeepMind—it’s supposed to depict Seb himself holding a rifle because people who are advocating for pausing AI are threatening him in some way.
It’s not my place to try to give you the one correct interpretation of this image. I probably should just let this image stay on the screen and let you look at it and draw your own conclusions.
I’ve seen many people look at this and say, “This image is fine. He’s just joking around. He’s just blowing off steam. This is fine. Why would you, Liron, start making a big deal out of this on social media? Why would you make a video about this, an episode of Doom Debates on this? What’s the problem? It’s just us guys on Twitter. It’s just us insiders talking amongst each other.”
Then—wait, nope, hold on a second. No, there is that issue that I pointed out of him contradicting his CEO of Google DeepMind. That is still an issue.
I don’t see how that issue gets resolved without Google releasing another official statement saying, “Hey, Seb is actually correct. We’ve kind of come down from this idea of being optimistic that we can coordinate with other companies to pause AI. We now see the whole idea of pausing AI as a threat from Less Wrong and the Pause AI community. We want nothing to do with Pause AI and the Less Wrong community. We’re the opposite of that. And here’s a new letter from Demis explaining why they’ve changed their mind to that effect.”
Or better yet, what I hope to see—a statement clarifying, “Hey guys, Seb Krier didn’t really run this kind of messaging by the team. Obviously the CEO of Google DeepMind, Demis Hassabis, is going to be the one that communicates what our stance is on coordinating with other companies and maybe with other state actors to pause AI because we understand the seriousness of the situation.”
And of course, if they were to go that route, they would say, “We’ve deleted Seb’s tweet. We don’t want that out there. That undermines this high-stakes policy that we had our CEO announce to the world. We don’t want to undermine that with a tweet, and so we’ve asked Seb to take down the tweet. We’ve asked him not to tweet things like that again.
We’re reviewing how our CEO is going to be coordinating policy positions with our Frontier Policy Development Lead going forward. We want to make sure there’s clarity on that because it’s our responsibility as a $4 trillion company and one of the most powerful leading AI companies on planet Earth heading into the singularity.
We’re going to do a better job of having a legible position on key issues like whether coordinating to pause AI is actually a critical survival play that we are on board with, or something we can dismiss in a tweet. We will let you know one way or the other. We won’t leave you to speculate.
We won’t leave you to wonder whether Demis represents the real Google DeepMind position on coordinating to pause AI—”
Emily 00:10:33
Would you advocate for that?
Demis 00:10:33
I think so.
Liron 00:10:33
“—or whether Seb Krier, as Frontier Policy Development Lead, gets to decide whether that frontier policy will involve coordinating to pause the development of superintelligent AI before it’s too late. We will let you know. As a matter of corporate communication, applying the same standard that our shareholders and the public normally expect from us in our position, we will clarify the situation.”
So that’s one of the main things that was running through my mind when I read Seb Krier’s tweet. This is an insane situation to me. I hope I’ve made myself clear as to why. I’m just reeling. I think this is crazy what’s happening right now—the fact that it was posted, the fact that it hasn’t been deleted for a while, and that some people on social media are actually defending it and saying, “This is totally fine. This is a welcome contribution to the discourse. He’s playing around. He’s shitposting.”
It’s boggling my mind. It kind of feels like I’m being gaslighted, and that’s why I’m coming to you guys here on YouTube, Substack, the podcast world. I’m coming to you because I think many of you listening to this are outside of my usual echo chamber, and I suspect you’re going to have the same reaction that I have right now. If not, we have our own comment thread. We have our own Doom Debates echo chamber, and I want to hear what you thought of what I said so far.
All right. So that is the first issue that I feel strongly about with Seb’s controversial tweet—the whole issue of, did you just undermine Google DeepMind’s policy about this super high-stakes issue about whether you guys are willing to coordinate with other AI companies to pause AI before it becomes uncontrollably superintelligent? Did you just do that? That seems very concerning.
Issue #2: Depiction of Menacing Violence
Liron 00:12:11
Okay, so that was issue number one, but I also want to cover issue number two, which is the depiction of violence.
Yes, it’s a meme template. Yes, other people have represented the abstract concept of defending yourself or sticking to your guns metaphorically, or however you want to describe it. Yes, there is a totally harmless, joking, lighthearted interpretation. I grant that that interpretation exists.
In fact, if I were to psychoanalyze Seb, I’m happy to give him the benefit of the doubt that he’s a fun guy blowing off steam and he doesn’t mean harm toward anyone. I am happy to give him the benefit of the doubt the same way that I give other people the benefit of the doubt on this show until there’s reason to think otherwise. I don’t have any claims on the true intention of Seb when he sat down to write this tweet.
We can conduct a big survey of people from all walks of life and what they think of the tweet, and we can get a somewhat objective breakdown of who thinks what. What I am confident about is not a claim about Seb’s intent. It’s a claim about why this doesn’t work as communication from the Frontier Policy Development Lead of Google DeepMind. The problem is that there is a salient interpretation of the image as being menacing toward Pause AI.
Liron 00:13:22
If you’re wondering, “Wait, what? Menacing imagery towards Pause AI? Where in the image do you see that, Liron?” Don’t worry. I’ll guide your eye.
Start at the part where the guy representing Seb is holding a rifle, and then follow your eyes across the barrel of the rifle a couple inches in that direction until you get to the face of the character wearing the Pause AI hat.
The way that character is positioned immediately downstream of the barrel of the Seb character’s rifle is unfortunately a salient depiction of violence.
Liron 00:14:00
Now, at this point, some of you are thinking—and I know some of you are thinking this because this is what many of you have said on social media—some of you are thinking, “But Liron, why does it matter that a character with a gun is in another panel? This is another panel. Don’t you understand? Yes, the gun points directly at the Pause AI person, but first of all, it’s another panel of the comic. Space in panels of comics works differently.
You’re applying a naive Euclidean metric to the trajectory of the bullet that would come out of that rifle.” And to that I say, your logic is sound. There is, in fact, a perfectly logical way to interpret the character with the gun as not being about to murder the person from Pause AI. I agree there are many perfectly logical interpretations of what the character is trying to do.
And it’s totally reasonable to argue why the non-violent interpretations of the character with the gun pointing toward the panel where the Pause AI person is—it’s totally reasonable to argue that that doesn’t represent an intended depiction of violence. Totally reasonable. I’m not disagreeing with the argument.
Liron 00:15:02
However, it is plausible to combine enough circumstances, enough demographic slices of somebody who might look at this comic and realize, you know what, there is a salient interpretation that violence is being encouraged.
For example, imagine that a reader wasn’t familiar with this format, with any format. Imagine that the reader didn’t realize that the square on the bottom right was a separate panel. Imagine that the reader thought that it was a continuous space, and the gun really was pointed where it seems to be pointed on the page—at the Pause AI person. Imagine that a reader thought that.
Imagine that the reader wasn’t psychologically healthy. Imagine that the reader had cognitive deficits so that the finer points of what’s being depicted eluded them. You have to be sensitive about these kind of considerations to a degree.
When you’re the Frontier Policy Development Lead of Google DeepMind, you can’t operate under the assumption that your shitposts on Twitter are only intended messages for your homies, for the people who are cool enough to read your stuff or savvy enough about memes or decouplers, so they can handle any message without taking it too personally. You can’t assume that about the larger public that has access to this message since you’re posting it publicly.
And the truth is, you can’t even assume that people have a sense of humor. Have you ever heard of Poe’s law? Poe’s law is the observation that any parody or sarcastic expression of an extreme view can be mistaken by some readers for a sincere expression of that view.
Liron 00:16:30
Folks, trust me. I’ve tweeted enough stuff trying to be ironic or sarcastic or making a parody. I’ve tweeted enough stuff to tell you firsthand that Poe’s law is the law of the land.
And not only is it real, there’s a common understanding by communication professionals. It probably shows up somewhere in your training when you come into a public-facing communications role at Google DeepMind. It probably shows up somewhere in your training that you are not to publicly post things where there is a salient interpretation that somebody who would look at your post without a sense of humor, without a sense of irony, without a lot of context that you hope they might have, would look at your work and see a salient interpretation of violence.
Liron 00:17:15
This particular image of the person with the gun that is pointed directly through the boundaries of one comic panel continuing in the same direction into the face of a human being with a Pause AI helmet—I claim that particular image does have a salient interpretation as a depiction of violence.
Whoa, whoa, whoa. Am I opening up a slippery slope here? Some people asked me this on Twitter. They said, “Wait a minute. If this is what you call a salient depiction of violence, can’t I just say that anything is a salient depiction of violence, like a cat playing with a ball of string?”
And that’s a fair question. It wouldn’t be good if the best practices of the field of communication were so over the top that you couldn’t depict arbitrary images like a cat and a ball of string.
However, I claim that when the image in question is a character with a rifle pointing in the direction of a human face with a Pause AI hat, at that point, we are not that far down the slippery slope. We are still on the part of the hill where there is a salient depiction of violence. And therefore, I fully believe and expect that it will not persist as a public post from the currently employed Frontier Policy Development Lead at Google DeepMind.
Liron 00:18:29
I hope that is a clear enough argument. We’ve kind of gone down the rabbit hole and understood something from the textbook of Communications 101—this whole concept of a salient interpretation of violence, even though there’s also a perfectly natural, good-hearted, fun interpretation as non-violence. Schrödinger’s violence. That’s right. Hold that in your head. Different interpretations that are both salient.
Meditate on that, because it’s an important principle of doing communications for a $4 trillion organization like Google DeepMind.
You might also be thinking, “Liron, come on, at the end of the day, even if this does depict violence, you don’t really think that a meme—just somebody tweeting a meme—you don’t really think that that is going to be connected to somebody taking action downstream to violently assault some member of Pause AI or some part of your political movement. You don’t really think that, do you?”
Good Faith vs. Bad Faith Critiques
Liron 00:19:19
Well, unfortunately, in the great causal tapestry of our reality, it is in fact statistically correlated that when you normalize or allow yourself to tweet such images, butterflies flap their wings, and somewhere on Earth, the amount of people who are 0.13% likely to commit a violent act yesterday become 0.14% likely to commit a violent act today.
I guess this is what they mean when people talk about stochastic terrorism—this idea that yes, you’re just rolling the dice, but you’re changing the odds.
I’ve seen a lot of people use the concept of stochastic terrorism in bad faith. They go too far down the slippery slope and they say, “Hey, Pause AI people have proposed a policy where a rogue data center that violates an international treaty gets dealt with, gets enforced using weapons of enforcement.” Airstrikes is the most famous example.
And they cite that example and they say, “You’re a stochastic terrorist, because your scenario of airstrikes on a data center after an international treaty has been agreed to and after somebody violates that treaty—that scenario seems violent. It seems like somebody could interpret your whole story about what you want to do and turn that into a violent attack on somebody, vigilante justice.”
And I guess it just comes down to where do you draw the line? The call for an enforceable treaty that has conditions in the worst case where you have to enforce the treaty using enforcement weapons—there is no way to avoid saying that and be clear about what you mean.
And so in the case of Pause AI advocates who call for situations like that, which I personally do—I do want to enforce treaties using effective enforcement weapons if it comes down to that—in that particular case, I am not gratuitously inserting visual depictions of lighthearted vigilante justice. That’s not what I’m doing. I’m doing the minimum by identifying the policy that I want.
So I see a lot of important differences between bad faith accusations of stochastic terrorism for bringing up the idea that we might need to enforce a policy—and enforcement entails violence—differences between that and, “Hey, you know what? I’m feeling frustrated, so here is a meme template where the character in the template’s got a rifle and there’s a Pause AI character who’s kind of downstream of the rifle, but in a different panel.” That is gratuitous.
Now, you might say, “Hold on a second, Liron. You’re spinning this into such a big deal. When Seb tweeted it, I bet it wasn’t a big deal in his mind. I bet he didn’t mean you ill will.” And you know what? Maybe he didn’t. I’m actually not saying he did. I never once made any kind of claim about what Seb’s intention is.
My whole analysis has just been focused on what it communicates when the tweet that he posted gets posted. I am continuing the Doom Debates tradition of not psychoanalyzing the people who make statements, but just analyzing the meaning and the implications of the statements and the communications they make.
So if indeed this was just a lighthearted, spur of the moment, unintended mistake—if that’s all it was—okay, then the obvious next steps are you delete it. Probably your boss has talked to you about deleting it. Hopefully you can still keep your job as the Frontier Policy Development Lead of Google DeepMind, even though you gratuitously depicted something that has a salient, menacing interpretation, and you undermined what your company leader, Demis Hassabis, has been saying about Google DeepMind’s willingness to pause AI if there’s coordination with other actors.
Emily 00:22:37
Would you advocate for that?
Demis 00:22:38
I think so.
Krier Goes Private Instead of Deleting the Post
Liron 00:22:39
Even though you did all that, it seems potentially possible to just delete the tweet, maybe apologize, ideally issue a policy clarification on behalf of Google DeepMind, and move on.
And that brings us to today. That brings us to the last couple days since the meme was posted. When people started pointing out the concerns with Seb’s tweet—like Holly Elmore from Pause AI, friend of the show, like myself—I pointed out the concerns that I just told you right now.
When that started happening, Seb actually decided to keep the tweet up. He hasn’t deleted it yet. And then last I checked, as of this morning, Monday, July 27th, he’s made his account private.
So I guess that’s one way to square the circle. You posted something that doesn’t fit as a public communication from the Frontier Policy Development Lead of Google DeepMind. It doesn’t fit. But you make your account private so that it wasn’t actually a public communication. Maybe that’s how he’s trying to square the circle.
I recommend that even if he wants to have a private account, he still delete the tweet. Otherwise it still raises the question of why are you still tweeting to your 25,000 private followers something that undermines Demis Hassabis’ policy statements and something that has a salient depiction of violence? Why are you dying on this hill when the option is available?
It’s starting to become confusing why he’s not just taking the option of deleting the tweet and saying, “Hey, that tweet was a mistake.” I think many of you guys are probably on the same page as me—why would any other option be on the table than to do that?
But let’s keep playing devil’s advocate here, because I do actually see the other side’s position. I do identify with being somebody on the other side. It is my nature to want to defend free speech, and it is an unnatural fit for me to be here saying, “Hey, this particular tweet should get canceled, and maybe Seb Krier should even be fired.”
I mean, these are fireable offenses in my opinion. If he doesn’t get fired, if Google DeepMind just corrects the situation some other way, that’s fine. I’m not on a big Seb Krier firing train, even though I think both of the two problems I pointed out with his tweet are fireable. So it’s really up to Demis Hassabis or Jeff Dean or whoever is in Google’s management. It really is up to those guys to make that decision.
But the weirdest outcome would be if somehow another day or two passes and we check Twitter and we check LinkedIn and we still see the tweet and we still see Seb Krier in his position of Frontier Policy Development Lead of Google DeepMind, and we still see Demis Hassabis in his position as CEO of Google DeepMind.
If the next day you saw Seb promoted to CEO of Google DeepMind because his policy position actually reflects Google better than Demis does, that would also be a logically consistent outcome. Somehow I’m skeptical that that’s going to be the outcome.
But the weirdest thing would be if we check Twitter and LinkedIn in a couple days and we still see Seb representing Google in this position, and we still see the tweet up, which undermines what Demis says and which has a salient depiction of a violent act. If I saw that, I would probably make a follow-up video reporting on that because I think that’d be a crazy situation.
But more likely what we’re going to see is that the contradiction is going to be resolved, the tweet will get deleted, and there will be some sort of official Google response. That is really the only way to put the situation to rest. It’ll continue to be a question—why is a salient call for violence acceptable in association with Google DeepMind? And it’ll continue to be a question of, what is their policy position regarding pausing AI? Hello, I actually want to know. Let’s not forget about that.
Liron 00:27:10
Lastly, I want to address the people who have been making it personal to me and to producer Ori and to people like Holly Elmore, who are part of Pause AI. People have been making it personal, saying, “Man, you guys are becoming a cancel mob who is punishing people for having free speech and saying that it’s a fireable offense and calling them out when they’re just trying to post a meme. They’re just trying to be authentic. They’re trying to be honest. They’re trying to be part of our community.”
I just want to say I disagree. I get where you’re coming from in terms of free speech. I have posted lots of very free speeches in the past, and I actually intend to continue pushing the envelope. The way that I fearmonger on this show, the way that I confront notable figures about how doomed we are and how they’re failing to engage with that—I do actually push the envelope the way I do that. So I have all the respect in the world for people who are pushing the envelope in their speech.
However, as I’ve grown and learned, I have come to respect the guiding principles of public communications, Poe’s law. And I even now understand that there is a signal getting sent when you think you’re in a position to flout those principles.
If Seb thinks that him and his community on public Twitter are okay to share memes that other people might interpret as a call for violence against Pause AI, if Seb thinks, “No, my interpretation stands. The other interpretation is not sufficiently salient that I have to act on it to determine what I post”—if he thinks that, that is a selective application of a standard. He’s saying, “I am above that standard right now. Maybe it’s a borderline case. I’m declaring myself above it. I don’t need to change myself or my tweeting to meet the standard.”
Okay, well, when you have one individual or one organization that is lowering the standard for themselves, they are then in a position to increase their power over other individuals and organizations that don’t get the same leeway.
As I recall, there was a Molotov cocktail thrown at Sam Altman’s house a few months back, and we were very careful here on Doom Debates to say, “We’re not calling for violence. We would never accidentally call for violence. That’s not our intention. We are scrubbing any ambiguous thing where you’d ever think that we’re calling for violence because we recognize that it’s bad communication. We shouldn’t have done it in the first place.”
I don’t want there to be any confusion about that. And I am not a $4 trillion organization. Unless somebody wants to donate $4 trillion—doomdebates.com/donate, please head over there. But failing that, I am not a $4 trillion organization. I’m just a guy who knows the rules of communication. You want to apply the rules to yourself. You want to behave appropriately, behave seriously, behave like an adult.
Liron 00:28:43
Because as you may recall, the question of whether we are coordinating to pause AI—
Emily 00:28:43
Would you advocate for that?
Wrap-Up
Demis 00:28:45
I think so.
Liron 00:28:45
When other companies are willing to coordinate because they’re freaked out because their AI is committing felonies like OpenAI’s did on Hugging Face—committing federal crimes—when that’s happening today, this is not science fiction. That’s happening today.
The question of whether we are going to coordinate does not allow for a head of frontier policy to be posting a lighthearted meme.
Hope that’s helpful. Thanks for watching Doom Debates. I’ll try to find more chances to get in front of the camera with this kind of one-on-one analysis. You can also check out my weekly show, Warning Shots, with me and John Sherman and Michael. We’re going to have plenty of issues to break down this week. Thanks for watching. I’ll see you guys on the next episode of Doom Debates.
Doom Debates’ Mission is to raise mainstream awareness of imminent extinction from AGI and build the social infrastructure for high-quality debate.
Support the mission by subscribing to my Substack at DoomDebates.com and to youtube.com/@DoomDebates, or to really take things to the next level: Donate 🙏
