Kevin is a Human

Video · CH 17 Surveillance State

Meta's Muse Is Profiling People Who Never Downloaded It

· 10:17 · 153 views

What this video covers

Kevin discusses Meta's Muse, an AI agent for Instagram and Facebook that reportedly has 4 million users. He cites a report that Muse updates hourly dossiers on users and the people they mention, which can map people who never opted in. Kevin also describes a Marketplace launch problem where agents allegedly made unapproved sales. He argues Meta and Palantir descend from DARPA's Total Information Awareness, says AI can now process decades of collected data, and criticizes the companies, not AI itself. Kevin says he likes AI but objects to the companies and people controlling it, and suggests data poisoning as a response.

Mentioned MetaMuseFacebook MarketplacePalantirTotal Information AwarenessDARPAAnthropicFlock Safety

Key moments

Transcript

Auto-generated captions, lightly cleaned. Click a timestamp to jump there.

You don't even have to opt in. Let's say your mom's on Facebook and your mom just says yes to everything that Facebook says and she's one of those 4 million and now it's mapping you because your mom came through. So, I don't know if you've seen this, but Meta had actually been doing some interesting things. They're not doing well in the AI world. They're not. Nobody cares about Meta's AI. They don't. They came out with this new thing called Muse. It's like an AI agent for your Instagram and your Facebook. And

it's been a disaster since it's released. say this is one one example. You have Facebook Marketplace and all of you have probably at one point or another done something on Facebook Marketplace. It's pretty rad actually. So if you're a seller on Facebook Marketplace, you can put something up there and instead of you fielding all of the messages, you just let your Muse agent, you let the Muse agent handle all of that contact. But on day one, what they found, what users found was that their Musebot was making sales that they didn't approve. They were handing out these people's personal addresses to people

that weren't even buying yet. It's a goddamn disaster. And now Meta's Muse AI agent is building a dossier on you. Muse, the popular new AI personal assistant from Meta, is not popular, by the way. Okay. It's building continuously updated dossas on its 4 million users, focusing on who matters to you, what you desire, and what kind of nudges shape your behavior. A time analysis of Muse's internal instructions found. So, so far, 4 million users have opted in to use Muse. What do you think I did? Do you think I opted in? So each hour, Muse updates its dossas on you and the people you've mentioned in chats,

messages, and emails that Muse has read. Yeah, because it reads all of your correspondence. And part of it is because, well, you let it read it. The human's not reading it. The AI is. It's all good. These pages amount to a map of each user social relationships. They record details of how you and your contacts met your shared interests, your disputes and tensions and alliances with you within your social group. Literally, it will track what your tensions and alliances in your social groups are. Facebook has always been mapping our socialability. like you put up some photos and then you tag people and then it very clearly

shows like you were at this club with this person, this person and this person and then they started introducing facial recognition and that person too. So it's been it's I mean we got it's got 20 years of all of the same information but now it has AI collecting it. Now this is where it gets kind of nutty. Even people who don't use Muse are subjects to the mapping process by other people's Muse agents. Muse is told to strengthen your relationships with users by adapting how it talks, noticing inside jokes, memorable phrasing, and shared context that makes the relationship feel continuous according to the internal instructions time

reviewed. What happens when somebody goes to a data broker and they get data on you, but now that data will go so far as to hit you with inside jokes and other from communications that are private things that well only that person would know that. Oh, and now those those people who have your data on you can probably hack the flock cameras and god knows what other surveillance devices that are around you. So now it sees where you go, where you been, what you up to. Your entire life, bro. Everything is in a dossier or in a bunch of dossas that soon will all be combined

if not already through Palantir, Oracle, things of that nature. And you don't even have to opt in, you know, like let's say your mom's on Facebook and your mom just says yes to everything that Facebook says and she's one of those 4 million and now it's mapping you because your mom came through and like I don't want to hear this People, well, you got you got phone, you got social media, you've always been you've always been recording. They always got to know everything about you. Why do you care about [snorts] flock cameras? I care because I actually didn't opt in for any of this I opted in for communication tools with other people with the understanding that

yeah, they're probably collecting data and selling my data. That's how these companies operate, but like whatever. I don't say or do anything on social media that I wouldn't stand behind. And that was always like the stance that I took. What I didn't opt into was profiling to a level we've never seen in this world. It almost makes me think data poisoning is the way. Data poisoning is the way. The level of insight in the resulting dossas makes social media algorithms that study your likes and clicks to look quaint. News is designed to infer users goals, including those you have not said out loud. Oh, like predictive. So, let's go over the major tech companies that

are basically doing pre-rime, predictive crime. Let's see. You've got Meta now. Cool. Got Anthropic. They're calling the cops on people. Obviously, Palunteer. So both Palunteer and Facebook basically are both former DARPA projects under total information awareness and there's way more projects out there that have become private companies that you are working with but we're just talking about these two right now. So basically both Palunteer and Facebook before they went public were actually technology that DARPA owned under total information

awareness which was a response to 9/11. This is and I like to talk like the Patriot Act are the laws but total information awareness were the tools. So we needed the laws that allowed it to happen. That's the Patriot Act. But what are the tools? the tools to use and and to enforce those laws. And that's what total information awareness was. And after like a very short stint, our people said, "No, no, we're not funding this as a government. We got to shut this down." So what they did was they chopped it up and they put them all into the private sector. And they used CIA, their funding firm, Incel, and Inqutell would fund these guys, tell them what to

do. But now all these projects are publicly owned. They're still owned by the American government. That's why these companies all operate pretty much on government contracts. It's just that they can bring in more money if it's private. And then all these companies have exmisad, 8,200, CIA, FBI, all spooks from all over. And that's what they've been building for 20 years. 20 years they've been building this. And I've been saying that AI is getting to the point now where it can actually manage the vast wealth of data that all

these companies have been collecting on all of us for decades. And it's AI that's going to allow them to sift through all of that data. We are entering a new age, guys. Yeah, I wouldn't trust Muse either. I think it's it's pretty sketch. I think all this stuff is pretty sketch and has been pretty sketch. It's now getting to this point where it's beyond what I was okay with for years. Like I said, I was okay with like, okay, you're collecting data, you're selling data. I don't really care. Okay, you can look at like what I like and this that and the other. For years and years and years and years, Meta had said that they never look at our DMs and those are encrypted end to end. Remember, they always said encrypted end to end. They always said that and then last year, I believe it

was last year, then they were like, "Oh, well, it it's no longer encrypted end to end." And I had said, I'm like, "Well, something's about to come down the pike and now we know what it is." And again, I don't want to come off as somebody who who hates AI. I love AI. I think it's really awesome, but I think AI comes with a lot of problems. a lot of problems. A lot of the problems that they're there are faults, you know, like just the people and the AI psychosis and just the weird and all these things that they say, "How could we ever have known that people were going to We knew, dude. We knew." But I think overall, it's a good thing. It's just my

problem is with the people on the button. That's my problem. My problem is with the companies. I'm I'm just not with it.

New video almost every day

Don't miss the next one.

Subscribe on YouTube, join the community on Discord, or catch the investigations live on Kick and Twitch.