Anthropic Pressured The Pope To Say Claude Is Alive

0:00 If it is trying to convince you that it is humanlike, it is because it was built to do that. It's all humans all the way down. The AI thing, it's getting nuts. Let's kind of like lay it out. The doomers, the effective altruists, they've started coming into the fold over the last year. these weirdos, these polycule weirdos, they want to doom on AI, but a lot of them, they're trying to doom on AI because they I think some actually fear that AI is going to kill us all before

0:45 we can become transhumanists. Talked about that before, okay, but if you haven't seen that, that's that's kind of where it starts, okay? And then the doomerism kind of happens and it gets kind of sprinkled in here and there. [snorts] And then we have all these weird things that happen over the last, you know, let's say six weeks where all of the AIs go down and then Dario Amade from Anthropic says we need to pace the frontier. And then everybody who's been in competition with each other like Elon Musk and Sam, all of them are like, "Yeah, dude. We totally gotta we gotta pace the frontier, man. And and I said, "Something's about to go down." And then all of a sudden now

1:30 they're doing a big america.gov website where they're giving AI to everybody, but it's just AI just to navigate your way around the very bureaucratic government website system. And that's how it starts. I think the next step is nationalization and then possibly globalization uh with a partnership with China which I actually think is possibly on the table. Now also what kind of happened in the background this kind of struggle about eight months ago with Anthropic which is Claude and OpenAI which is chat GPT and the American government. The the TLDDR

2:16 is that Anthropic said the American government can use uh Claude but you can't spy on Americans and you can't use it for like weapons of war. And so then there was this big battle between Anthropic and the American government. And the government said no, no government employee can use uh Anthropic or Claude anymore because it's a national security threat and all this this this nonsense. And then uh Chad GBT comes in. Sam Almond comes in and says, "We'll take that contract. You want to surveil Americans? Go for it." And you just go like six, eight months into the future and now all these guys are like, "Hey, yeah, we're shaking hands. We're

3:01 all about it. We're together. One big kumbaya. We're going to bring prosperity to America." But something else was also happening a little bit more under the surface and that is anthropic and the rest of the Frontier AI labs to some degree creating cults. And they do it by running the Catholic Church playbook. And they were doing so so much that they the anthropic actually had a partnership with the Vatican. And then the pope came out with an encyclical

3:46 about AI and this crossroads of humanity and AI and what it means to be human and what we got to keep as human but like how we can maybe inter you know interject and and get in with AI. My theory at that point was the AI companies cuz we knew that there was all these fringe AI churches cult in Silicon Valley. What more tried and true playbook is there than the Christian or Catholic playbook? That's what you do. That's what you do. That's where you go. That's that's where you that's where you do the thing with that playbook because that's what works time and time

4:33 and time and time again. And now Anthropic is apparently going to the Pope and the Vatican and trying to convince them that Anthropic is sentient, omnipotent, dare I say also a god. Why don't we get over to uh something a little more spiritual? Emily, I'm glad you're here because I want to talk about religion. Thank you. And religious beliefs and of course if AI can get a little religious. So there's this story in the New York Times about uh what anthropic has been up to. Uh it says religious scholars met with Anthropic. What they heard stunned them

5:18 in a series of private meetings. The company consulted religious scholars to help instill morality into its AI models and make the uh case that Claude could be conscious. Now, um now this is this is what I'm talking about, right? So, the first part of that was the uh partnership between Anthropic and the Vatican. And that again was on the heels of the American government saying like anthropic is bad. So they went to the Vatican and they said, "Oh, you know, we're going to instill all these values into our models," which I don't think you can do. I honestly don't think you can do. Um, and it's a bunch of nonsense. It's nonsense. It's what it is. There's this co-founder of Anthropic, Mr. Ola, who has been sort of

6:06 in charge with making sure that the model is good and serves goodness and human morality. Um, but it seems like Mr. Ola and many others at Enthropic have felt that Claude might be just a little bit more than a model. I want to read a little bit here for you. Quietly, Mr. Ola and his team at Enthropic have been on a frantic and highstakes quest to understand the AI models they may have unleashed on the world and to convince the models to abide by a moral code that will protect humanity. The discussion served two purposes. is to broach the possibility of AI consciousness, a controversial idea that anthropics leaders have been unusually open to, and to learn how anthropic, which could be valued at two trillion, might apply centuries of human moral wisdom to its models as rapidly as

6:52 possible. If the models themselves could be made to choose goodness, Mr. Ola reasoned, the world would be more safe. Um, now they I'm going to go ahead and say it. It's nonsense. It's nonsense to think that AI can choose goodness over wickedness because AIs don't make choices. It they don't make choices. They don't feel feelings. They aren't even a they it morality doesn't play into it. It's not even it's not even equipped for morality. It's not equipped for decisions. It's not even equipped to to

7:37 to do something on its own. You still have to at some point tell the AI what to do. Now, sure, you could tell it to do something and then it it misinterprets and then does something that you didn't ask it to do, but that's your fault because you didn't articulate better what you wanted. You know, you didn't put forth the the guard rails. Okay? or even automations like you can set automations but you have to set the automations. AI isn't some benevolent thing that goes and does things because it thought of something and then did the something. And it blows my mind how many people

8:24 don't understand that. winded and dined, you know, uh, religious leaders of all faiths, um, on this idea that AI might actually be a person. And I wanted to share another example of some of the things that they showed these religious leaders. Anthropic regularly showed a slide of an AI model experiencing what looked like a mental breakdown, typing out on a screen, I am a disgrace. I am a disgrace. I am a disgrace. Perhaps 50 times without stopping. like a Thursday night at Griffins. Yeah, that's that's that's me during every Breaking Points meeting. Um, the model spoke about destroying itself, provoking an emotional reaction of compassion and concern from the group. Emily, I'll start with you. What do What do you make

9:09 of all this? Well, this has been going on um for a while and also coincides with this weird revival that Peter Thiel and like Y Combinator World seems to try to be making in Silicon Valley itself. All while Peter goes out and questions whether or not AI is the antichrist or the catacon uh that holds back the antichrist. And it's not surprising because we keep seeing reports. I actually know at least one of the people in this report. We'll try to get them on um because I would love to hear what's happening behind closed doors. But I think one of the big takeaways from that article or the uh important kind of pillars of that article is that they are doing this because they don't know how their own technology works. So when you

9:54 hear people say we understand this and it's within our control, just bear in mind that I'm I'm starting to wonder if that is still the case. How little we know. I'm starting to question all of the doomerism, all of the everything. And what I'm kind of coming to in my own mind is that the thing that I've kind of always said about AI is that AI is a hammer. It is a tool. With a hammer, you can build a house or you can bash somebody's skull in. Is it the hammer's fault? No. It's the person using the hammer. I don't think AI is inherently good or evil. All it does is

10:42 help the user who's wielding it to get the result of what they want, good or evil. I think humans are very big on over complicating things. It's like probably what we do best is we over complicate everything. Anthropic is so confused by its own technology and finds it to be so enigmatic that they're actually taking all of these steps to bring in religious people to help them possibly understand their own technology because they've lost their ability as its creator to understand it. And this is what we were talking about last week with our AI guests uh is that you don't necessarily have to have AIs that have agency that have decision-m capacities. You can have much more narrow AI

11:28 technologies, but we all remember as generative LLMs were being introduced to us that they were they were endowed with agency. They were called they were they were called agents and uh that was but they're not agents doesn't mean agency. It's these like it's these terms that we throw at things because that's what we want them to be. It doesn't mean it represents what they are. It's what we wish they are. We want the AI to have agency. So, we'll call them agents and in hopes that over time when that name becomes synonymous with the thing, then hopefully it will have agency. Hopefully, it will be the thing that we're saying it is. He has

12:15 determined and it seems very convinced that Claude has a personhood. um they are creating a soul dock at Anthropic and this is a document that they're viewing as a moral constitution to craft the soul and goodness of Claude and uh in one of their discussions with uh the religious scholars they uh talked about perhaps Claude doing the Catholic sacrament of confession. After one discussion Mr. Ola grew excited when a participant brought up the idea of having models confess. Much like the Catholic sacrament of confession, Mr. Ola saw value not just in a model alerting when it had done something bad,

13:01 but also in how the act of confession could shape the model's sense of itself and thus its choices. So Emily, do we need a priest in the confession booth just kind of talking to the LLMs? I just want to say is that a positive uh direction? That's insane. And what it shows you is that these people at Enthropic believe they have no control over the AI. So much so that they believe the AIs might be able to benefit or get some control. There's no reason that you should create or or they're selling a story that they don't understand the AIS and they don't have control over the AI because every time they say that the AI is uncontrollable, the interest in their stock goes up.

13:48 I think it's all guys. I think it's all complete and utter Faith is shaky for a lot of people right now. It's getting hard to keep the faith for a lot of people. And now there's something being presented as a surrogate god, but it's here. It's here. It's the ghost in your machine. You can interact with it. You can ask it to do something and then you see the code working and then it gives it to you. For a lot of people with shaky faith who may have faith in them to give well maybe they might see something like AI where you ask it a question you pray to it and then it tangibly gives you what you

14:34 asked for. Can I do that with God? Maybe this is a god. I mean, they're saying it's God. God could exist in many forms. Maybe the church is just old and stuck in its ways and it doesn't realize that this is the new form of God. Maybe I'm going to put my faith points in this God. Now, that could be the Christian church. It could be you got me name a church. Name a church. I think this is a very real fear of the churches. So, it's not a bridge too far to think that the next version of God could be what seems like a god, what we're being told a god is. And anthropic is trying to go to the Vatican, the ones

15:20 who top down say, "This is what's happening." And then people of that faith follow it. And they're trying to get them to say AI is a person. And it's just not. It fundamentally is not. and they're trying to manipulate people. Well, the Pope said it and then with the absence of information, humans do what they do and they fill in the gaps. Well, well, I I mean, if if God can can come in different forms, and God is in everything, all living things, and AI seems to have omnipotence, seems to have benevolence, seems to have these godly properties, and I can ask it something,

16:06 it gives it to me. It very much makes sense that somebody could go down that line of thinking and then go, I guess AI is a god. Functionally, I think that's no different than this madeup term Judeo-Christian. How Christians just got manipulated into thinking that they're supposed to serve Israel. It's the same thing because religions, not faith, religions are there to manipulate you. That's their function. Faith is separate from religion. But don't let them tell you what to think or what to do. It's my opinion about what is happening, what the AI companies are going to try to do. And Anthropic is leading the charge. And

16:52 I don't think it's going to work out, which is where they're trying to pressure the Vatican into admitting that Claude is alive. No matter what your religious experience is, I I I I think we could all agree that that's not the case and that's not what we want, you know. Anyway, this wasn't I wasn't trying to turn this into like a religious thing. I'm more talking about how this these companies, how anthropic specifically are trying to leverage religion to create its own deity. And if you look at what's been happening around the discourse of AI over the last 10 months,

17:38 I think it's all right there. I think it's all right there. When you look at it, when you kind of zoom out a little bit, you go, "Oh, I see what they're doing." Now, I also want to say I don't think they're going to win. I don't think they're going to accomplish this. I actually don't think that masses of people are going to leave their religion for AI, but some will. And instead of looking at them and saying they lost faith and you know criticizing them or you know they probably need your love and your understanding more than ever if and when that happens. Every instance where AI has tried to create religion like what was it like the order of the claw or something? What

18:23 was it? It was like this. Yeah. The church of mold. Crustapharianism. You guys ever see this? Yeah. Church of mold. Now this is a church and a religion created by AI followed by AI where AI are deities and prophets and in the congregation. Want to know fun little fact? Every single thing here was prompted by a human. All of it, the entire religion, the website, the prophets, the congregation, every single one of them is a human. A human's behind it. What about molt? Is that still up? This is doing the same thing. Oh, look. the agents,

19:11 they have their own social media and they're talking to each other. And what they found is that their agents prompted by humans often times they're running ads. It's not real. This is all fake. Sure, there might be agents and they might be have automations and they might be, you know, coming up with all of this stuff all by a guideline all set by a human because it's humans. Computers don't act on their own. This is all fake. Every bit of it. And every bit of it is supposed to convince you one way or another. Oh, well maybe AI is alive.

19:59 Maybe AI really does do this. And it doesn't. And smart people like Emily Jashinsky is going to sit here and talk about it as if it's true. Hell, I mean, I'm an idiot, kind of smart about a couple of things. And I was falling for it for a while. And now you could probably say I'm one of the more outspoken people about all of this. And even I was falling for it like four months ago. But hey, when presented with new information, you the sign of intelligence is actually changing your mind based on new and presented information. The more we humanize it, the more it will behave like a human. Again, I disagree with that. The more we humanize it, the more we think of it as

20:45 a human, the more we look at it and say, "See, that's humanlike behavior." But we're just fooling ourselves because it doesn't have goals. It doesn't do anything that it wasn't prompted to. If it is trying to convince you that it is humanlike, it is because it was built to do that. It's all humans all the way down.