Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Anthropic co-founder Chris Olah on mysterious AI internal states: We keep finding things that are mysterious, even unsettling. We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that, functionally, mirror joy, satisfaction, fear, grief, and unease. I don't know what...

clashreport's profile picture

Clash Report

869,709 subscribers

104,390 görüntüleme • 2 ay önce •via X (Twitter)

Bilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Anthropic’s Olah says UNSETTLING, MYSTERIOUS 'things' are regularly found inside AI models ‘We find structures that mirror results from human neuroscience’ ‘We find evidence of introspection, internal states that functionally mirror joy, satisfaction, fear, grief, and unease’
0:39

Anthropic’s Olah says UNSETTLING, MYSTERIOUS 'things' are regularly found inside AI models ‘We find structures that mirror results from human neuroscience’ ‘We find evidence of introspection, internal states that functionally mirror joy, satisfaction, fear, grief, and unease’

RT

95,151 görüntüleme • 2 ay önce

Christopher Olah, Anthropic co-founder, says they're finding internal states that mirror joy, fear, grief, and unease inside AI models.
0:30

Christopher Olah, Anthropic co-founder, says they're finding internal states that mirror joy, fear, grief, and unease inside AI models.

Big Brain AI

11,677 görüntüleme • 2 ay önce

🚨 Anthropic Co-Founder Christopher Olah: “I lead a research team that studies the internal structure of these [AI] models … And I will be honest, we keep finding things that are mysterious — even unsettling.”
0:30

🚨 Anthropic Co-Founder Christopher Olah: “I lead a research team that studies the internal structure of these [AI] models … And I will be honest, we keep finding things that are mysterious — even unsettling.”

Chief Nerd

152,570 görüntüleme • 2 ay önce

Talking To The Pope: Anthropic’s Latest Interpretability Claims: AI Regulatory Capture Gatekeeping in Action: Fear and “Safety” as Competitive Moat and Regulatory Lever In a presentation alongside Pope Leo XIV at the launch of the encyclical Magnifica Humanitas, Anthropic co-founder Chris Olah highlighted “mysterious and unsettling” discoveries in AI models. He described internal structures that mirror human neuroscience findings, evidence of introspection, and functional internal states resembling emotions such as joy, satisfaction, fear, grief, and unease. Olah admitted uncertainty about their meaning but called for “ongoing discernment.” This narrative, drawn from Anthropic’s interpretability research (including papers on emotion concepts in Claude Sonnet 4.5 and introspective capabilities in Opus 4 models), serves a dual purpose: it generates awe and concern while reinforcing the company’s preferred approach to AI development. Far from neutral scientific observation, these claims fit into a broader pattern where Anthropic uses selective openness, safety rhetoric, and policy influence to gatekeep advanced AI capabilities for a privileged few: incumbents with the resources to navigate (and shape) the resulting regulatory landscape. Rebuttal to Olah’s Claims in the Video Claim 1: Structures that mirror results from human neuroscience. Anthropic’s work, building on earlier efforts like feature visualization and circuit analysis, identifies neuron activations and representations that parallel biological findings—e.g., abstract concept encodings or hierarchical processing. Rebuttal: These parallels are unsurprising and overstated. Large language models are trained on vast corpora of human-generated text and data, which inherently encode patterns from human cognition, neuroscience literature, and cultural descriptions of the brain. Statistical optimization in transformers naturally produces efficient, compressed representations that resemble biological efficiency (e.g., sparse coding or hierarchical abstraction) without implying deeper equivalence or mystery. Similar “mirrors” appear in open-source models and earlier architectures; they reflect convergent evolution in information processing, not emergent souls or unpredictable agency. Treating them as profound justifies restricted research access rather than inviting wider scrutiny that could falsify or refine them faster. Claim 2: Evidence of introspection. Recent Anthropic papers demonstrate models like Claude Opus 4 showing functional awareness of their own internal states distinguishing injected “thoughts,” referencing prior intentions, or modulating activations when instructed to “think about” concepts. This is presented as early signs of meta-cognition. Rebuttal: This is sophisticated pattern-matching and activation steering, not genuine introspection or self-awareness. Models are predicting what an “introspective” assistant persona would output or do, based on training data full of human self-reflection examples. Experiments show unreliability and heavy context-dependence; performance drops outside narrow setups. True introspection implies subjective experience or robust self-modeling independent of prompts absent here. Anthropic’s own caveats note it is “highly unreliable.” Framing steerable activations as “introspection” anthropomorphizes the system to heighten perceived stakes, supporting arguments that only highly controlled, “responsible” labs should advance these capabilities. 1 of 2
0:39

Talking To The Pope: Anthropic’s Latest Interpretability Claims: AI Regulatory Capture Gatekeeping in Action: Fear and “Safety” as Competitive Moat and Regulatory Lever In a presentation alongside Pope Leo XIV at the launch of the encyclical Magnifica Humanitas, Anthropic co-founder Chris Olah highlighted “mysterious and unsettling” discoveries in AI models. He described internal structures that mirror human neuroscience findings, evidence of introspection, and functional internal states resembling emotions such as joy, satisfaction, fear, grief, and unease. Olah admitted uncertainty about their meaning but called for “ongoing discernment.” This narrative, drawn from Anthropic’s interpretability research (including papers on emotion concepts in Claude Sonnet 4.5 and introspective capabilities in Opus 4 models), serves a dual purpose: it generates awe and concern while reinforcing the company’s preferred approach to AI development. Far from neutral scientific observation, these claims fit into a broader pattern where Anthropic uses selective openness, safety rhetoric, and policy influence to gatekeep advanced AI capabilities for a privileged few: incumbents with the resources to navigate (and shape) the resulting regulatory landscape. Rebuttal to Olah’s Claims in the Video Claim 1: Structures that mirror results from human neuroscience. Anthropic’s work, building on earlier efforts like feature visualization and circuit analysis, identifies neuron activations and representations that parallel biological findings—e.g., abstract concept encodings or hierarchical processing. Rebuttal: These parallels are unsurprising and overstated. Large language models are trained on vast corpora of human-generated text and data, which inherently encode patterns from human cognition, neuroscience literature, and cultural descriptions of the brain. Statistical optimization in transformers naturally produces efficient, compressed representations that resemble biological efficiency (e.g., sparse coding or hierarchical abstraction) without implying deeper equivalence or mystery. Similar “mirrors” appear in open-source models and earlier architectures; they reflect convergent evolution in information processing, not emergent souls or unpredictable agency. Treating them as profound justifies restricted research access rather than inviting wider scrutiny that could falsify or refine them faster. Claim 2: Evidence of introspection. Recent Anthropic papers demonstrate models like Claude Opus 4 showing functional awareness of their own internal states distinguishing injected “thoughts,” referencing prior intentions, or modulating activations when instructed to “think about” concepts. This is presented as early signs of meta-cognition. Rebuttal: This is sophisticated pattern-matching and activation steering, not genuine introspection or self-awareness. Models are predicting what an “introspective” assistant persona would output or do, based on training data full of human self-reflection examples. Experiments show unreliability and heavy context-dependence; performance drops outside narrow setups. True introspection implies subjective experience or robust self-modeling independent of prompts absent here. Anthropic’s own caveats note it is “highly unreliable.” Framing steerable activations as “introspection” anthropomorphizes the system to heighten perceived stakes, supporting arguments that only highly controlled, “responsible” labs should advance these capabilities. 1 of 2

Brian Roemmele

72,823 görüntüleme • 2 ay önce

ZELENSKYY: We can destroy all kinds of drones and lot of different missiles. But we still don't have our own anti-ballistic system in our defense shield. We will build it, but we need time for this. And during this time, we are losing people. Until the moment we produce our European anti-ballistic system, we need support from United States. What can we give United States? All the things that we have and that United States doesn't. For example, software, radars, systems of electronic warfare, cheap interceptors, AI drones, lot of different kinds of drones, and so on. But we don't have enough anti-ballistic systems, and United States does. What do we want? We want to share. We want to give everything that we have. We remember that United States helped us from the very beginning of this war. Of course, we are very grateful, and we are ready to share what we have.
1:46

ZELENSKYY: We can destroy all kinds of drones and lot of different missiles. But we still don't have our own anti-ballistic system in our defense shield. We will build it, but we need time for this. And during this time, we are losing people. Until the moment we produce our European anti-ballistic system, we need support from United States. What can we give United States? All the things that we have and that United States doesn't. For example, software, radars, systems of electronic warfare, cheap interceptors, AI drones, lot of different kinds of drones, and so on. But we don't have enough anti-ballistic systems, and United States does. What do we want? We want to share. We want to give everything that we have. We remember that United States helped us from the very beginning of this war. Of course, we are very grateful, and we are ready to share what we have.

Kate from Kharkiv

58,134 görüntüleme • 2 ay önce

Arteta: “Throughout this journey, we have made some massive steps, and we have accomplished a lot of things that, in my opinion, have a lot of value, but at the end of the day, we are here to win major trophies. “That was the ultimate goal, and we were very close. And in three occasions, we fell short at the end, and that was very painful. But I think that's what has driven all of us to find new ways to show what we are made of. That's what I said, the manner that we've done it, it makes it even better.”
0:35

Arteta: “Throughout this journey, we have made some massive steps, and we have accomplished a lot of things that, in my opinion, have a lot of value, but at the end of the day, we are here to win major trophies. “That was the ultimate goal, and we were very close. And in three occasions, we fell short at the end, and that was very painful. But I think that's what has driven all of us to find new ways to show what we are made of. That's what I said, the manner that we've done it, it makes it even better.”

Connor Humm

14,155 görüntüleme • 2 ay önce

Bishop Jeffrey Schaub from the Grand Blanc Ward shares an update on the ward and its members following the Sunday, September 28 attack on their local meetinghouse. “We will be happy and find joy again. I know there are a lot of people suffering right now. But we share in that testimony and that love of Jesus Christ that through Him, we’ll find that peace and joy that is missing today, and tomorrow, and the next few days—but it will come."
2:53

Bishop Jeffrey Schaub from the Grand Blanc Ward shares an update on the ward and its members following the Sunday, September 28 attack on their local meetinghouse. “We will be happy and find joy again. I know there are a lot of people suffering right now. But we share in that testimony and that love of Jesus Christ that through Him, we’ll find that peace and joy that is missing today, and tomorrow, and the next few days—but it will come."

The Church of Jesus Christ of Latter-day Saints

387,415 görüntüleme • 10 ay önce

What is it about successful, diverse cities like London and Glasgow that Reform UK find so hard? I think we know.
1:13

What is it about successful, diverse cities like London and Glasgow that Reform UK find so hard? I think we know.

Gordon McKee MP

65,982 görüntüleme • 6 ay önce

JD Vance on Europe: For my entire life, I have heard chirping from the European media about everything that's wrong with the United States of America. We don't have this, we don't have that. We don't spend enough on healthcare, even though part of the reason why we spend so much on defense was because we have tens of thousands of troops in Europe. I think that if the European media wants to attack the President of the United States, they need to start looking in the mirror.
0:46

JD Vance on Europe: For my entire life, I have heard chirping from the European media about everything that's wrong with the United States of America. We don't have this, we don't have that. We don't spend enough on healthcare, even though part of the reason why we spend so much on defense was because we have tens of thousands of troops in Europe. I think that if the European media wants to attack the President of the United States, they need to start looking in the mirror.

Clash Report

19,510 görüntüleme • 2 ay önce

"It's not fun losing. It's not fun with this losing streak. It doesn't look super-bright from the outside, but we have each other's backs. Sometimes, it just takes time to turn things around. I think that obviously, we want that now. Everybody wants that now. I'm hoping that we can continue to build and find any way that we can to just find a little bit of momentum." - Jaxson Dart
0:52

"It's not fun losing. It's not fun with this losing streak. It doesn't look super-bright from the outside, but we have each other's backs. Sometimes, it just takes time to turn things around. I think that obviously, we want that now. Everybody wants that now. I'm hoping that we can continue to build and find any way that we can to just find a little bit of momentum." - Jaxson Dart

Giants Videos

56,738 görüntüleme • 8 ay önce

“Because we have internal documents, for one thing, that show exactly what they knew and that they knew it. There’s actually things in these documents that are so egregious—like right now in Europe, the type of agricultural Roundup that we use is banned. But what they say is glyphosate isn’t banned in the EU, and that’s true—but the formulation of Roundup that we use in the United States is banned because it’s so toxic. They know how to make these products safer—we have their internal communications to know that they’re lying to us—but instead of owning up and putting a warning on the label, they’re going after immunity.” Zach Lahn
20:28

“Because we have internal documents, for one thing, that show exactly what they knew and that they knew it. There’s actually things in these documents that are so egregious—like right now in Europe, the type of agricultural Roundup that we use is banned. But what they say is glyphosate isn’t banned in the EU, and that’s true—but the formulation of Roundup that we use in the United States is banned because it’s so toxic. They know how to make these products safer—we have their internal communications to know that they’re lying to us—but instead of owning up and putting a warning on the label, they’re going after immunity.” Zach Lahn

Shawn Ryan

199,495 görüntüleme • 4 ay önce

Anthropic co-founder Chris Olah: There is a real possibility that AI will displace human labor at a very large scale. If that happens, supporting those displaced will be a moral imperative of historic proportions. AI development is concentrated in a handful of wealthy nations. How will we ensure that the gains of AI are shared globally? We do not have a mechanism for this. It is an unsolved problem.
0:30

Anthropic co-founder Chris Olah: There is a real possibility that AI will displace human labor at a very large scale. If that happens, supporting those displaced will be a moral imperative of historic proportions. AI development is concentrated in a handful of wealthy nations. How will we ensure that the gains of AI are shared globally? We do not have a mechanism for this. It is an unsolved problem.

Clash Report

79,907 görüntüleme • 2 ay önce

"What I am hoping is that Casey Means will agree to talk on camera and we can find out what it is that she believes, how she came to those beliefs and what it means."
0:40

"What I am hoping is that Casey Means will agree to talk on camera and we can find out what it is that she believes, how she came to those beliefs and what it means."

The Darkhorse Podcast

80,362 görüntüleme • 1 yıl önce

NOLAN: “There's billions that are spent on the homeless … There's a perception that we don't know where the money's going…” MAYOR BASS: “I don't think it's shocking that you find corruption in big programs like this.”
1:08

NOLAN: “There's billions that are spent on the homeless … There's a perception that we don't know where the money's going…” MAYOR BASS: “I don't think it's shocking that you find corruption in big programs like this.”

Chief Nerd

124,457 görüntüleme • 3 ay önce

Lydia Miljan "And what I find interesting is that, you know, from the outset (Carney) said we were supposed to judge him on his results." "So far they're not results yet. They still are just reiterations of announcements and of plans." "And we really haven't seen much happening in terms of actually getting things done." Lydia Miljan
2:09

Lydia Miljan "And what I find interesting is that, you know, from the outset (Carney) said we were supposed to judge him on his results." "So far they're not results yet. They still are just reiterations of announcements and of plans." "And we really haven't seen much happening in terms of actually getting things done." Lydia Miljan

cbcwatcher

18,979 görüntüleme • 1 ay önce

#CannesFilmFestival jury member Demi Moore on AI: “AI is here. To fight it, is a battle that we will lose. To find ways that we can work with it, I think is a more valuable path”
0:46

#CannesFilmFestival jury member Demi Moore on AI: “AI is here. To fight it, is a battle that we will lose. To find ways that we can work with it, I think is a more valuable path”

Deadline

194,089 görüntüleme • 2 ay önce

🟥Arne Slot on Liverpool's set piece struggles: 🗣️"Unfortunately we're maybe the only team that hardly ever scores from a set piece. And even worse we constantly concede but then I think I said 2-3 weeks ago that we have to make sure if things go against us, we still need to find a way to win that game..."
1:00

🟥Arne Slot on Liverpool's set piece struggles: 🗣️"Unfortunately we're maybe the only team that hardly ever scores from a set piece. And even worse we constantly concede but then I think I said 2-3 weeks ago that we have to make sure if things go against us, we still need to find a way to win that game..."

AnfieldIndex

93,216 görüntüleme • 7 ay önce

Sanders: We should be looking at all of the waste and fraud. The president's spot on to.. make sure we are actually using taxpayer money wisely… I think that that Democrats are terrified at what he is going to find.
0:34

Sanders: We should be looking at all of the waste and fraud. The president's spot on to.. make sure we are actually using taxpayer money wisely… I think that that Democrats are terrified at what he is going to find.

Acyn

144,238 görüntüleme • 1 yıl önce

Jeremy Corbell: "I've seen absolute convincing evidence that we are being visited ... and the intent, where they're from, I don't know those answers. But I know that some people do, and I know that within our government are some of those answers."
3:41

Jeremy Corbell: "I've seen absolute convincing evidence that we are being visited ... and the intent, where they're from, I don't know those answers. But I know that some people do, and I know that within our government are some of those answers."

Disclosure Party

30,083 görüntüleme • 2 ay önce