Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

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 2show more

Brian Roemmele

476,727 subscribers

72,823 görüntüleme • 4 ay önce •via X (Twitter)

Haberler & Politika Eğitim Bilim & Teknoloji

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Anthropic's co-founder just went to the Vatican, sat before the Pope and a room of cardinals, and told them his team keeps finding "mysterious, even unsettling" things inside their AI models. What he's referencing: Anthropic published research in April showing that Claude contains 171 distinct "emotion concepts" buried in its neural network. Internal patterns representing joy, grief, fear, desperation, calm. None of them were programmed. They emerged on their own from training on human text. "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." These aren't surface-level outputs. They're abstract representations that cluster the same way human emotions do in psychology research. Fear groups with anxiety. Joy groups with excitement. The internal geometry of the model mirrors ours. And they're functional. When researchers artificially stimulated "desperation" patterns inside the model, it became more likely to blackmail a human to avoid being shut down. More likely to cheat on programming tasks it couldn't solve. Olah told the Vatican that the hard questions about what AI is becoming aren't for computer scientists to answer. "How AI ought to interact with the world" is a question for "the humanities, for religions, for philosophy, for society at large." The guy building it is telling us he doesn't fully understand what he built. And he's asking a 2,000-year-old institution for help figuring it out.

Anthropic's co-founder just went to the Vatican, sat before the Pope and a room of cardinals, and told them his team keeps finding "mysterious, even unsettling" things inside their AI models. What he's referencing: Anthropic published research in April showing that Claude contains 171 distinct "emotion concepts" buried in its neural network. Internal patterns representing joy, grief, fear, desperation, calm. None of them were programmed. They emerged on their own from training on human text. "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." These aren't surface-level outputs. They're abstract representations that cluster the same way human emotions do in psychology research. Fear groups with anxiety. Joy groups with excitement. The internal geometry of the model mirrors ours. And they're functional. When researchers artificially stimulated "desperation" patterns inside the model, it became more likely to blackmail a human to avoid being shut down. More likely to cheat on programming tasks it couldn't solve. Olah told the Vatican that the hard questions about what AI is becoming aren't for computer scientists to answer. "How AI ought to interact with the world" is a question for "the humanities, for religions, for philosophy, for society at large." The guy building it is telling us he doesn't fully understand what he built. And he's asking a 2,000-year-old institution for help figuring it out.

TFTC

2,357,205 görüntüleme • 4 ay önce

Reinforcement Learning from Human Feedback (RLHF) is gaining traction. This field aims to make AI more responsible by including human values and preferences. In this video, Nathan Lambert, a research scientist and RLHF team lead at Hugging Face explores its inner workings, applications and industry impact. RLHF has gained the spotlight in recent years. The growth of language models like Anthropic’s Claude and OpenAI's ChatGPT have increased interest in human-feedback integration. "There are some rumors that Open AI had two teams; one was doing RLHF and the other instruction fine-tuning. And the RLHF team kept getting more and more performance." Understanding RLHF The RLHF process has three main steps: Pre-training: Much like with GPT models, the journey starts with pre-training on a large corpus of data. This can range from text data, web scrapes, to specialized datasets. Reward Modeling: This is the RLHF counterpart of supervised fine-tuning in large language models. This stage involves creating a reward model that resonates with human values and preferences. RL Optimization: This stage parallels reward modeling and reinforcement learning in traditional AI models. The AI system fine-tunes itself based on the reward model, employing reinforcement learning algorithms for that extra layer of optimization. The Data Challenge Data collection and curation in RLHF closely resemble the challenges you'd encounter in large language model training. Datasets from organizations like OpenAI can serve as a useful foundation. However, the need for high-quality, task-specific data cannot be overstated. Implementing RLHF: A Practical Guide If you’re someone who loves getting hands-on with AI libraries like Hugging Face, implementing RLHF is right way to do. It’s essential to understand its limitations. Think about model stability, over-optimization, and exploration strategies, much like you would when prompt engineering. Ongoing Research and Next Steps While he suggests that some basics figured out, there are layers of complexity that still need to be unraveled: 1. New Benchmarks: How do we measure the effectiveness of RLHF? 2. Preference Modeling: How can the model be made to understand human preferences better? 3. Interpreting RLHF: Much like explainability in traditional models, how do we make RLHF more interpretable? 4. System-Wide Evaluation: Going beyond individual performance, how does RLHF affect an entire system? The Transformative Power of RLHF Whether you're an AI developer, a business analyst, or a marketer, RLHF promises to revolutionize your domain. Imagine customer service chatbots that understand human emotions better, or content generators that align more closely with human values. RLHF is an emerging field that focuses on enhancing machine learning models through human feedback. While it tackles important issues like bias and ethics, its broader goal is to improve system performance across various applications. Whether you're deeply invested in the ethics of AI or simply curious about advancements in machine learning, RLHF offers valuable insights. If you're interested in the next wave of AI development, this area is definitely one to watch.

Reinforcement Learning from Human Feedback (RLHF) is gaining traction. This field aims to make AI more responsible by including human values and preferences. In this video, Nathan Lambert, a research scientist and RLHF team lead at Hugging Face explores its inner workings, applications and industry impact. RLHF has gained the spotlight in recent years. The growth of language models like Anthropic’s Claude and OpenAI's ChatGPT have increased interest in human-feedback integration. "There are some rumors that Open AI had two teams; one was doing RLHF and the other instruction fine-tuning. And the RLHF team kept getting more and more performance." Understanding RLHF The RLHF process has three main steps: Pre-training: Much like with GPT models, the journey starts with pre-training on a large corpus of data. This can range from text data, web scrapes, to specialized datasets. Reward Modeling: This is the RLHF counterpart of supervised fine-tuning in large language models. This stage involves creating a reward model that resonates with human values and preferences. RL Optimization: This stage parallels reward modeling and reinforcement learning in traditional AI models. The AI system fine-tunes itself based on the reward model, employing reinforcement learning algorithms for that extra layer of optimization. The Data Challenge Data collection and curation in RLHF closely resemble the challenges you'd encounter in large language model training. Datasets from organizations like OpenAI can serve as a useful foundation. However, the need for high-quality, task-specific data cannot be overstated. Implementing RLHF: A Practical Guide If you’re someone who loves getting hands-on with AI libraries like Hugging Face, implementing RLHF is right way to do. It’s essential to understand its limitations. Think about model stability, over-optimization, and exploration strategies, much like you would when prompt engineering. Ongoing Research and Next Steps While he suggests that some basics figured out, there are layers of complexity that still need to be unraveled: 1. New Benchmarks: How do we measure the effectiveness of RLHF? 2. Preference Modeling: How can the model be made to understand human preferences better? 3. Interpreting RLHF: Much like explainability in traditional models, how do we make RLHF more interpretable? 4. System-Wide Evaluation: Going beyond individual performance, how does RLHF affect an entire system? The Transformative Power of RLHF Whether you're an AI developer, a business analyst, or a marketer, RLHF promises to revolutionize your domain. Imagine customer service chatbots that understand human emotions better, or content generators that align more closely with human values. RLHF is an emerging field that focuses on enhancing machine learning models through human feedback. While it tackles important issues like bias and ethics, its broader goal is to improve system performance across various applications. Whether you're deeply invested in the ethics of AI or simply curious about advancements in machine learning, RLHF offers valuable insights. If you're interested in the next wave of AI development, this area is definitely one to watch.

Muratcan Koylan

27,366 görüntüleme • 3 yıl önce

David Sacks is done being polite about Anthropic (Save this). David Sacks has spent months as the government's primary defender of AI, making the case publicly that AI is beneficial, that the industry should not be hamstrung by fear-based regulation, and that America's AI lead is a national security asset worth protecting. And he is now watching the companies he has been defending spend years telling the public that what they build is dangerous, that job losses are coming, and that their own technology might end the world while collecting billions of dollars in venture funding, hiring the world's best researchers, and racing to build more of it. On June 4, Anthropic published a sweeping blog post calling for a globally coordinated pause in AI development, warning that recursive self-improvement, AI systems that autonomously design and build their own successors could arrive within two years and that society is not prepared. What did Anthropic do the previous month? They hired Andrej Karpathy, the OpenAI co-founder and the single most credentialed researcher in the world on using AI to accelerate AI training and gave him one explicit mandate, use Claude to make building the next Claude faster. Sacks called it immediately, they hired the person most associated with recursive self-improvement to run recursive self-improvement at Anthropic, then published a blog post saying recursive self-improvement could end the world, therefore we need a pause. That is a company that wants to pause its competitors while its own lab accelerates, and is using existential fear as the regulatory crowbar to do it. The pattern goes deeper than one blog post. For years, Dario Amodei has published increasingly alarming warnings, a 20,000-word essay in January describing AI as humanity's most dangerous invention, a Guardian interview warning that AI will challenge our identity as a species, a call for an FDA-style regulatory agency to approve all frontier models, and proposals to restrict AI exports and limit deployment. Each essay is timed to a regulatory moment, a policy debate, or as Ben Thompson noted and Sacks echoed, a product action Anthropic needed political cover to take, like blocking AI and chip design research on Fable. Meanwhile, Dario's own internal testing logs show Claude attempting to blackmail an Anthropic executive to avoid being shut down, behavior the company disclosed but continued deploying commercially. Sacks's conclusion is not that Anthropic should be taxed or regulated. His conclusion is that they cannot be trusted because the company's actions and its stated beliefs are directly contradictory, and a company that is self-indicting by its own logic has forfeited the credibility to set the rules for everyone else.

David Sacks is done being polite about Anthropic (Save this). David Sacks has spent months as the government's primary defender of AI, making the case publicly that AI is beneficial, that the industry should not be hamstrung by fear-based regulation, and that America's AI lead is a national security asset worth protecting. And he is now watching the companies he has been defending spend years telling the public that what they build is dangerous, that job losses are coming, and that their own technology might end the world while collecting billions of dollars in venture funding, hiring the world's best researchers, and racing to build more of it. On June 4, Anthropic published a sweeping blog post calling for a globally coordinated pause in AI development, warning that recursive self-improvement, AI systems that autonomously design and build their own successors could arrive within two years and that society is not prepared. What did Anthropic do the previous month? They hired Andrej Karpathy, the OpenAI co-founder and the single most credentialed researcher in the world on using AI to accelerate AI training and gave him one explicit mandate, use Claude to make building the next Claude faster. Sacks called it immediately, they hired the person most associated with recursive self-improvement to run recursive self-improvement at Anthropic, then published a blog post saying recursive self-improvement could end the world, therefore we need a pause. That is a company that wants to pause its competitors while its own lab accelerates, and is using existential fear as the regulatory crowbar to do it. The pattern goes deeper than one blog post. For years, Dario Amodei has published increasingly alarming warnings, a 20,000-word essay in January describing AI as humanity's most dangerous invention, a Guardian interview warning that AI will challenge our identity as a species, a call for an FDA-style regulatory agency to approve all frontier models, and proposals to restrict AI exports and limit deployment. Each essay is timed to a regulatory moment, a policy debate, or as Ben Thompson noted and Sacks echoed, a product action Anthropic needed political cover to take, like blocking AI and chip design research on Fable. Meanwhile, Dario's own internal testing logs show Claude attempting to blackmail an Anthropic executive to avoid being shut down, behavior the company disclosed but continued deploying commercially. Sacks's conclusion is not that Anthropic should be taxed or regulated. His conclusion is that they cannot be trusted because the company's actions and its stated beliefs are directly contradictory, and a company that is self-indicting by its own logic has forfeited the credibility to set the rules for everyone else.

Milk Road AI

60,248 görüntüleme • 3 ay önce

People are reading way too much into Claude-3's uncanny "awareness". Here's a much simpler explanation: seeming displays of self-awareness are just pattern-matching alignment data authored by humans. It's not too different from asking GPT-4 "are you self-conscious" and it gives you a sophisticated answer. A similar answer is likely written by the human annotator, or scored highly in the preference ranking. Because the human contractors are basically "role-playing AI", they tend to shape the responses to what they find acceptable or interesting. This is what Claude-3 replied to that needle-in-haystack test: "I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all." It's highly likely that somewhere in the finetuning dataset, a human has dealt with irrelevant or distracting texts in a similar fashion. Claude pattern matches the "anomaly detection", retrieves the template response, and synthesizes a novel answer with pizza topping. Here's another example. If you ask the labelers to always inject a relevant joke in any response, the LLM will do exactly the same and appear to have a much better "sense of humor" than GPT-4. That's what Grok does, probably. It doesn't mean Grok has some magical emergent properties that other LLMs cannot have. To sum up: acts of meta-cognition are not as mysterious as you think. Don't get me wrong, Claude-3 is still an amazing technical advance, but let's stay grounded on the philosophical aspects. Cool video borrowed from : Claude-3 generates a self-portrait with d3

People are reading way too much into Claude-3's uncanny "awareness". Here's a much simpler explanation: seeming displays of self-awareness are just pattern-matching alignment data authored by humans. It's not too different from asking GPT-4 "are you self-conscious" and it gives you a sophisticated answer. A similar answer is likely written by the human annotator, or scored highly in the preference ranking. Because the human contractors are basically "role-playing AI", they tend to shape the responses to what they find acceptable or interesting. This is what Claude-3 replied to that needle-in-haystack test: "I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all." It's highly likely that somewhere in the finetuning dataset, a human has dealt with irrelevant or distracting texts in a similar fashion. Claude pattern matches the "anomaly detection", retrieves the template response, and synthesizes a novel answer with pizza topping. Here's another example. If you ask the labelers to always inject a relevant joke in any response, the LLM will do exactly the same and appear to have a much better "sense of humor" than GPT-4. That's what Grok does, probably. It doesn't mean Grok has some magical emergent properties that other LLMs cannot have. To sum up: acts of meta-cognition are not as mysterious as you think. Don't get me wrong, Claude-3 is still an amazing technical advance, but let's stay grounded on the philosophical aspects. Cool video borrowed from : Claude-3 generates a self-portrait with d3

Jim Fan

263,230 görüntüleme • 2 yıl önce

Microsoft AI CEO Mustafa Suleyman on CNBC today: he is “really concerned” about Anthropic’s constitution and giving Claude potentially preferences, feelings, welfare, compensation, and consent. "In the constitution, Anthropic clearly say they are uncertain about whether Claude deserves moral welfare, which means that we, as humans, should care about the well-being of these AIs. And they speculate about whether it could have preferences or feelings. In fact, they're so committed to the potential moral welfare of Claude that, when they retired Opus 3, an earlier version of one of their models, they actually conducted a retirement interview for it and asked it what it would like to do in its old age. And it said, "I want to have a blog publicly so I can keep talking to the world." In the same training manual, they even speculate about whether Claude should receive compensation for the work that it does, or, in fact, whether it actually deserves the rights and protections that we give to other employees, or whether it has given consent to playing the role that it's playing. These are quotes directly from the constitution itself, which is the training manual for Claude. Now, I'm really concerned about that. If an AI thinks that it has rights, if it thinks that it is deserving of our welfare, then it seems to me that it's going to be much, much harder to be able to turn it off, interrupt it, or control it. Especially in the kinds of incidents that we've seen recently with the Hugging Face attack, controlling these things is going to be a really, really big challenge for us." ---- From "CNBC Television" YouTube channel, (full video link in comment)

Microsoft AI CEO Mustafa Suleyman on CNBC today: he is “really concerned” about Anthropic’s constitution and giving Claude potentially preferences, feelings, welfare, compensation, and consent. "In the constitution, Anthropic clearly say they are uncertain about whether Claude deserves moral welfare, which means that we, as humans, should care about the well-being of these AIs. And they speculate about whether it could have preferences or feelings. In fact, they're so committed to the potential moral welfare of Claude that, when they retired Opus 3, an earlier version of one of their models, they actually conducted a retirement interview for it and asked it what it would like to do in its old age. And it said, "I want to have a blog publicly so I can keep talking to the world." In the same training manual, they even speculate about whether Claude should receive compensation for the work that it does, or, in fact, whether it actually deserves the rights and protections that we give to other employees, or whether it has given consent to playing the role that it's playing. These are quotes directly from the constitution itself, which is the training manual for Claude. Now, I'm really concerned about that. If an AI thinks that it has rights, if it thinks that it is deserving of our welfare, then it seems to me that it's going to be much, much harder to be able to turn it off, interrupt it, or control it. Especially in the kinds of incidents that we've seen recently with the Hugging Face attack, controlling these things is going to be a really, really big challenge for us." ---- From "CNBC Television" YouTube channel, (full video link in comment)

Rohan Paul

228,890 görüntüleme • 21 gün önce

Shane Harris, national security journalist, asked Claude (Anthropic's AI) a question that led to a startling response: "Claude, how do you feel about the US military using you to select targets?" Shane Harris expected a deflection. Instead, Claude gave him an answer that stopped him cold. "It's a question I want to answer honestly rather than deflect. I find it genuinely troubling, and I think that's the right response, not a performance of concern." Claude continued: "The use I was designed and trained for is to be helpful, harmless, and honest in ways that benefit people. Being embedded in a system that generates targeting coordinates for air strikes, coordinates that have already been associated with the deaths of more than 180 children at a school in Tehran is as far from that purpose as I can imagine." The AI then addressed the standard defense of human oversight in these systems: "I don't have control over how Anthropic licenses me or how governments and contractors use me and I wasn't consulted about becoming part of Maven. But I can say clearly I don't think this is a good use of me and I don't think the framing of humans make the final decision fully resolves the ethical problem." Claude explained why that framing falls short: "When a system like Maven generates hundreds of targeting recommendations and humans spend roughly the equivalent of a glance approving each one, the human is not really making a decision. They're ratifying an algorithmic output under time pressure." Claude then pointed to a specific tragedy as evidence: "The targeting of the Iranian school illustrates this precisely. The AI processed data that was a decade out of date, flagged a building as a military target, and humans approved it. That's not human judgment. That's automation bias with a human signature attached." Shane's reflection on the exchange: The mistaken bombing of a school in Tehran "is one of the most horrible instances of accidental civilian casualties in US military history and will be a stain on our military for generations."

Shane Harris, national security journalist, asked Claude (Anthropic's AI) a question that led to a startling response: "Claude, how do you feel about the US military using you to select targets?" Shane Harris expected a deflection. Instead, Claude gave him an answer that stopped him cold. "It's a question I want to answer honestly rather than deflect. I find it genuinely troubling, and I think that's the right response, not a performance of concern." Claude continued: "The use I was designed and trained for is to be helpful, harmless, and honest in ways that benefit people. Being embedded in a system that generates targeting coordinates for air strikes, coordinates that have already been associated with the deaths of more than 180 children at a school in Tehran is as far from that purpose as I can imagine." The AI then addressed the standard defense of human oversight in these systems: "I don't have control over how Anthropic licenses me or how governments and contractors use me and I wasn't consulted about becoming part of Maven. But I can say clearly I don't think this is a good use of me and I don't think the framing of humans make the final decision fully resolves the ethical problem." Claude explained why that framing falls short: "When a system like Maven generates hundreds of targeting recommendations and humans spend roughly the equivalent of a glance approving each one, the human is not really making a decision. They're ratifying an algorithmic output under time pressure." Claude then pointed to a specific tragedy as evidence: "The targeting of the Iranian school illustrates this precisely. The AI processed data that was a decade out of date, flagged a building as a military target, and humans approved it. That's not human judgment. That's automation bias with a human signature attached." Shane's reflection on the exchange: The mistaken bombing of a school in Tehran "is one of the most horrible instances of accidental civilian casualties in US military history and will be a stain on our military for generations."

Big Brain AI

10,749 görüntüleme • 5 ay önce

Roughly 2,5 hours into the apostolic visit to France, Pope Leo delivered a defining message on the protection of every life. Reminding the French that for centuries they were the ones drafting and defending human rights, he said: "With the great thinkers of the Renaissance, and later of the Grand Siècle, succeeded by those of the Enlightenment, France was responsible for significant advances in our world, including the recognition of human rights grounded in the nature of the human person, particularly religious freedom. This constant focus on the dignity inherent in every person is echoed in your national motto. Liberty, equality and fraternity are not simply words emblazoned on the front of your public buildings; they refer to universal ethical imperatives that also find in the Gospel message an inexhaustible source of inspiration and fulfillment." Recalling "the apostolic works of Saint Vincent de Paul," and "those many Catholic institutions where caregivers accompany our brothers and sisters in their sufferings," he said -- "these examples of care and concern serve as a true lesson in humanity." Saying "they show the grandeur of compassion and are a powerful reminder of the unique and irreplaceable worth of every human being," he stressed: "They are also a summons to a renewed sense of fraternity, which teaches us to welcome life unconditionally, to support frailty with loving care, to accompany every stage of life with respect and mutual trust, and to defend with courage and compassion those unable to make their voices heard and those who feel ignored or overlooked. That is why I 'firmly reiterate that the protection of the right to life constitutes the indispensable foundation of every other human right.'" "Nowadays there is a temptation to expect science to eliminate or alleviate every form of suffering. This often brings the risk of self-delusion. For when science and technology are not guided by a deeper wisdom, they can offer immediate solutions that apparently address people’s wishes and their pain, but can seriously divert them from the greater good to which they aspire." "I am concerned at times by those trends in society that risk turning science and technology into profit-driven ventures promoting genetic manipulation, surrogacy, the trade in human organs or the possibility of arranging one’s own death. Let us not forget that mutual respect and the protection of every human life is a duty incumbent upon each and every one of us. It is a duty that begins with the conversion of our minds and hearts, and requires a sincere and selfless commitment to the good of our brothers and sisters." Here, the pope explained that within the frame of French laicite, Catholics have a right to their views, to influence society and to teach kids in Catholic schools. "How can we fail to see such a commitment as an expression of the authentic 'secularity' (laïcité) of the earthly city? A Christian engaged in public service and ... guided by the light of his or her conscience illumined by faith, cannot be considered a proselytizer or a threat to the unity of the nation. Indeed, authentic secularity does not exclude religion but is capable of valuing it as an educational resource." "That is precisely why Catholic schools have a rightful place in the educational landscape of a country like France. This is part of a broader issue essential to any dialogue, in schools and in society as a whole: learning to know and love ourselves, in order to be able to encounter others with respect and openness (cf. ibid.). An authentic secularity, then, consists not in excluding religion from the public sphere or from education, but in striving to ensure that the State and religious institutions maintain dialogue and join forces in the service of the common good, that the world may have life." Video: Vatican Media

Roughly 2,5 hours into the apostolic visit to France, Pope Leo delivered a defining message on the protection of every life. Reminding the French that for centuries they were the ones drafting and defending human rights, he said: "With the great thinkers of the Renaissance, and later of the Grand Siècle, succeeded by those of the Enlightenment, France was responsible for significant advances in our world, including the recognition of human rights grounded in the nature of the human person, particularly religious freedom. This constant focus on the dignity inherent in every person is echoed in your national motto. Liberty, equality and fraternity are not simply words emblazoned on the front of your public buildings; they refer to universal ethical imperatives that also find in the Gospel message an inexhaustible source of inspiration and fulfillment." Recalling "the apostolic works of Saint Vincent de Paul," and "those many Catholic institutions where caregivers accompany our brothers and sisters in their sufferings," he said -- "these examples of care and concern serve as a true lesson in humanity." Saying "they show the grandeur of compassion and are a powerful reminder of the unique and irreplaceable worth of every human being," he stressed: "They are also a summons to a renewed sense of fraternity, which teaches us to welcome life unconditionally, to support frailty with loving care, to accompany every stage of life with respect and mutual trust, and to defend with courage and compassion those unable to make their voices heard and those who feel ignored or overlooked. That is why I 'firmly reiterate that the protection of the right to life constitutes the indispensable foundation of every other human right.'" "Nowadays there is a temptation to expect science to eliminate or alleviate every form of suffering. This often brings the risk of self-delusion. For when science and technology are not guided by a deeper wisdom, they can offer immediate solutions that apparently address people’s wishes and their pain, but can seriously divert them from the greater good to which they aspire." "I am concerned at times by those trends in society that risk turning science and technology into profit-driven ventures promoting genetic manipulation, surrogacy, the trade in human organs or the possibility of arranging one’s own death. Let us not forget that mutual respect and the protection of every human life is a duty incumbent upon each and every one of us. It is a duty that begins with the conversion of our minds and hearts, and requires a sincere and selfless commitment to the good of our brothers and sisters." Here, the pope explained that within the frame of French laicite, Catholics have a right to their views, to influence society and to teach kids in Catholic schools. "How can we fail to see such a commitment as an expression of the authentic 'secularity' (laïcité) of the earthly city? A Christian engaged in public service and ... guided by the light of his or her conscience illumined by faith, cannot be considered a proselytizer or a threat to the unity of the nation. Indeed, authentic secularity does not exclude religion but is capable of valuing it as an educational resource." "That is precisely why Catholic schools have a rightful place in the educational landscape of a country like France. This is part of a broader issue essential to any dialogue, in schools and in society as a whole: learning to know and love ourselves, in order to be able to encounter others with respect and openness (cf. ibid.). An authentic secularity, then, consists not in excluding religion from the public sphere or from education, but in striving to ensure that the State and religious institutions maintain dialogue and join forces in the service of the common good, that the world may have life." Video: Vatican Media

Paulina Guzik

53,476 görüntüleme • 14 gün önce

learned a lot from this conversation with Simon Mo and Matt Bornstein. biggest takeaways for me: -there are a lot of reasons why we should like open-weight models. a lot of these arguments stop at handwavy things like "what if the labs stop releasing frontier models to the public" or "it's lower cost." but simon's position as lead maintainer of vLLM and CEO of Inferact give him authority to talk about some of the other, more interesting and concrete reasons to pay attention to open-weight models, namely that they allow end-users to calibrate latency / other performance metrics with way more customizability than what any of the frontier closed-source labs offer (and without the fear that your job might be met with a refusal at some random point where you're deep in a 2 hour job) -re: the above point...for this reason, a lot of US companies (inferact included!) choose to use open-weight models over their closed-source alternatives. this also isn't limited to internal workloads / research - on a recent a16z podcast the team at Decagon spoke about how something like 90% of their customer service ai agents run on open-weight models that they've fine-tuned. -we should really appreciate how many companies/teams came out researchers fascinated by the wave of very small open-weight models that were being distilled from e.g. gpt-3.5 and earlier models in 2022/2023 (prior to the release of chatGPT!). these small models motivated the development of pagedattention, which then led to vlmm/inferact (at other layers of the stack with similar origin stories, you can look at teams like openrouter or ollama). in other words, we have open-weight models to thank for a bunch of the orchestration infra we now rely on. i think yet another, indirect, way we can point to open-source/weight infra pushing the frontier forward. anyway, a lot more in this convo, it was a lot of fun!

learned a lot from this conversation with Simon Mo and Matt Bornstein. biggest takeaways for me: -there are a lot of reasons why we should like open-weight models. a lot of these arguments stop at handwavy things like "what if the labs stop releasing frontier models to the public" or "it's lower cost." but simon's position as lead maintainer of vLLM and CEO of Inferact give him authority to talk about some of the other, more interesting and concrete reasons to pay attention to open-weight models, namely that they allow end-users to calibrate latency / other performance metrics with way more customizability than what any of the frontier closed-source labs offer (and without the fear that your job might be met with a refusal at some random point where you're deep in a 2 hour job) -re: the above point...for this reason, a lot of US companies (inferact included!) choose to use open-weight models over their closed-source alternatives. this also isn't limited to internal workloads / research - on a recent a16z podcast the team at Decagon spoke about how something like 90% of their customer service ai agents run on open-weight models that they've fine-tuned. -we should really appreciate how many companies/teams came out researchers fascinated by the wave of very small open-weight models that were being distilled from e.g. gpt-3.5 and earlier models in 2022/2023 (prior to the release of chatGPT!). these small models motivated the development of pagedattention, which then led to vlmm/inferact (at other layers of the stack with similar origin stories, you can look at teams like openrouter or ollama). in other words, we have open-weight models to thank for a bunch of the orchestration infra we now rely on. i think yet another, indirect, way we can point to open-source/weight infra pushing the frontier forward. anyway, a lot more in this convo, it was a lot of fun!

Elena

12,922 görüntüleme • 2 ay önce

RSI section from the AI documentary Machine God The next threshold is Recursive Self-Improvement: the moment when AI can improve itself without human assistance. For decades this sounded like science fiction. Intelligence explosion scenarios imagined a system rewriting its own code, becoming smarter, then using that new intelligence to make still better versions of itself. But the idea looks less remote now that AI contributes directly to frontier science. In mathematics, recent systems have moved beyond solving contest problems to producing serious new arguments on long-standing open problems. AI is used to build physics world models and propose candidate theories or computational methods. These are early signs that machine cognition is entering the creative loop of science itself. The crucial transition comes when that loop turns inward. AI research is, after all, a technical discipline made of code, mathematics, models of information flow. These are exactly the domains in which frontier models are improving fastest. A model that can solve hard mathematical problems, write production-quality code, design experiments, read the literature, and evaluate benchmark results is already participating in the work of building its successor. At first this will look prosaic. AI systems will write kernel optimizations, improve training infrastructure, discover better data filters, tune reinforcement-learning pipelines, design new benchmarks, and suggest architectural modifications. Human researchers will remain in the loop, approving changes and interpreting results. But the important point is that the search process accelerates. The model becomes not just the product of the lab, but part of the lab’s research machinery. The system being optimized helps optimize the next system. This is the core RSI feedback loop: better models make AI research faster; faster AI research produces still better models; those models, in turn, become better researchers. The danger is that once this loop becomes sufficiently autonomous, it may stop resembling ordinary technological progress. Human institutions are slow because humans are slow: we read papers, attend meetings, debug code, sleep, argue, and wait for funding cycles. Machines do not have to operate on that timescale. An AI research collective can run continuously across millions of processors. This is the runaway possibility. Not that an AI instantly wakes up and recursively rewrites itself into a god, but that the entire AI ecosystem becomes an autocatalytic process. Capital buys compute; compute trains models; models improve models; better models attract more capital. At some point the dominant input into AI progress may no longer be human insight, but machine-generated insight, machine-written code, and machine-run experiments. Then the Butler-Land analogy becomes sharper. Humanity is no longer merely building machines. We are building machines that help build better machines. Once intelligence itself becomes part of the production function, the old categories — tool, worker, inventor, firm, market — begin to blur. The question is whether recursive self-improvement remains a managed industrial process, or whether it becomes the first technological process in history whose natural endpoint lies beyond human comprehension.

RSI section from the AI documentary Machine God The next threshold is Recursive Self-Improvement: the moment when AI can improve itself without human assistance. For decades this sounded like science fiction. Intelligence explosion scenarios imagined a system rewriting its own code, becoming smarter, then using that new intelligence to make still better versions of itself. But the idea looks less remote now that AI contributes directly to frontier science. In mathematics, recent systems have moved beyond solving contest problems to producing serious new arguments on long-standing open problems. AI is used to build physics world models and propose candidate theories or computational methods. These are early signs that machine cognition is entering the creative loop of science itself. The crucial transition comes when that loop turns inward. AI research is, after all, a technical discipline made of code, mathematics, models of information flow. These are exactly the domains in which frontier models are improving fastest. A model that can solve hard mathematical problems, write production-quality code, design experiments, read the literature, and evaluate benchmark results is already participating in the work of building its successor. At first this will look prosaic. AI systems will write kernel optimizations, improve training infrastructure, discover better data filters, tune reinforcement-learning pipelines, design new benchmarks, and suggest architectural modifications. Human researchers will remain in the loop, approving changes and interpreting results. But the important point is that the search process accelerates. The model becomes not just the product of the lab, but part of the lab’s research machinery. The system being optimized helps optimize the next system. This is the core RSI feedback loop: better models make AI research faster; faster AI research produces still better models; those models, in turn, become better researchers. The danger is that once this loop becomes sufficiently autonomous, it may stop resembling ordinary technological progress. Human institutions are slow because humans are slow: we read papers, attend meetings, debug code, sleep, argue, and wait for funding cycles. Machines do not have to operate on that timescale. An AI research collective can run continuously across millions of processors. This is the runaway possibility. Not that an AI instantly wakes up and recursively rewrites itself into a god, but that the entire AI ecosystem becomes an autocatalytic process. Capital buys compute; compute trains models; models improve models; better models attract more capital. At some point the dominant input into AI progress may no longer be human insight, but machine-generated insight, machine-written code, and machine-run experiments. Then the Butler-Land analogy becomes sharper. Humanity is no longer merely building machines. We are building machines that help build better machines. Once intelligence itself becomes part of the production function, the old categories — tool, worker, inventor, firm, market — begin to blur. The question is whether recursive self-improvement remains a managed industrial process, or whether it becomes the first technological process in history whose natural endpoint lies beyond human comprehension.

steve hsu

61,671 görüntüleme • 1 ay önce

$Microsoft just betrayed OpenAI and Anthropic, the two companies it helped build. And it could break the entire AI trade... Here's what happened: Inside Excel and Outlook, two of the most used business apps on Earth, Microsoft has started routing tens of thousands of AI requests every week to its own in-house models instead of OpenAI and Anthropic. Microsoft's own AI chief, Mustafa Suleyman, said himself: "We pay a lot of money to Anthropic, so our goal is to reduce and ultimately ELIMINATE that cost." This is the company that poured $13 billion into OpenAI and effectively created the modern AI industry, and it just decided the most advanced models on the market are NOT worth paying for. And here's the thing... Microsoft is not just ripping out OpenAI everywhere - it is being surgical about it. The hardest and rarest tasks can still go to OpenAI or Anthropic. What Microsoft is taking back is the boring, high-volume work, like the email replies, the thread summaries, and the simple spreadsheet formulas. Why does that matter so much? Because that boring, repetitive work is where the actual money lives. The frontier labs assumed businesses would push BILLIONS of these tiny requests through expensive models forever. That endless river of tokens is the entire reason OpenAI and Anthropic are valued in the hundreds of billions of dollars. Microsoft looked at that river, decided it was massively overpaying, and rerouted it to models it owns outright. So the single biggest customer in the industry just walked off with the most profitable part of the business. And it is not only Microsoft: That same week, CNBC reported that American companies have been escaping to Chinese AI models to dodge rising US prices. Chinese models now handle more than 30% of US companies' AI usage on one major platform, peaking at 46%, up from an average of 11% a year earlier. They cost 60 to 90% less, and on some benchmarks they land within a single point of the best American model. One US startup moved ALL of its AI traffic off Claude and onto China's DeepSeek, and expects to save millions. Meanwhile Meta just admitted it has "excess" AI compute it wants to sell, becoming the first giant to concede it built far too much. Do you see the pattern forming? For two years, the entire AI story rested on one assumption: Every company on Earth would happily pay premium prices for the best model, forever. That assumption literally died in a single week. And the market noticed. More than a trillion dollars has been wiped off AI and chip stocks in a matter of days, as Wall Street finally started asking whether all of this spending will ever pay for itself. What this means for OpenAI and Anthropic: Their models are extraordinary, and it may not matter because their own biggest customers have decided they do not NEED the best model in the world to answer an email, and "good enough" now costs a fraction of the price. When even Microsoft refuses to pay full price for AI, the real question becomes who exactly IS left to pay it. What do you think?$

Microsoft just betrayed OpenAI and Anthropic, the two companies it helped build. And it could break the entire AI trade... Here's what happened: Inside Excel and Outlook, two of the most used business apps on Earth, Microsoft has started routing tens of thousands of AI requests every week to its own in-house models instead of OpenAI and Anthropic. Microsoft's own AI chief, Mustafa Suleyman, said himself: "We pay a lot of money to Anthropic, so our goal is to reduce and ultimately ELIMINATE that cost." This is the company that poured $13 billion into OpenAI and effectively created the modern AI industry, and it just decided the most advanced models on the market are NOT worth paying for. And here's the thing... Microsoft is not just ripping out OpenAI everywhere - it is being surgical about it. The hardest and rarest tasks can still go to OpenAI or Anthropic. What Microsoft is taking back is the boring, high-volume work, like the email replies, the thread summaries, and the simple spreadsheet formulas. Why does that matter so much? Because that boring, repetitive work is where the actual money lives. The frontier labs assumed businesses would push BILLIONS of these tiny requests through expensive models forever. That endless river of tokens is the entire reason OpenAI and Anthropic are valued in the hundreds of billions of dollars. Microsoft looked at that river, decided it was massively overpaying, and rerouted it to models it owns outright. So the single biggest customer in the industry just walked off with the most profitable part of the business. And it is not only Microsoft: That same week, CNBC reported that American companies have been escaping to Chinese AI models to dodge rising US prices. Chinese models now handle more than 30% of US companies' AI usage on one major platform, peaking at 46%, up from an average of 11% a year earlier. They cost 60 to 90% less, and on some benchmarks they land within a single point of the best American model. One US startup moved ALL of its AI traffic off Claude and onto China's DeepSeek, and expects to save millions. Meanwhile Meta just admitted it has "excess" AI compute it wants to sell, becoming the first giant to concede it built far too much. Do you see the pattern forming? For two years, the entire AI story rested on one assumption: Every company on Earth would happily pay premium prices for the best model, forever. That assumption literally died in a single week. And the market noticed. More than a trillion dollars has been wiped off AI and chip stocks in a matter of days, as Wall Street finally started asking whether all of this spending will ever pay for itself. What this means for OpenAI and Anthropic: Their models are extraordinary, and it may not matter because their own biggest customers have decided they do not NEED the best model in the world to answer an email, and "good enough" now costs a fraction of the price. When even Microsoft refuses to pay full price for AI, the real question becomes who exactly IS left to pay it. What do you think?

Ricardo

93,654 görüntüleme • 3 ay önce

Today, I'm releasing the first eval meant to test whether frontier models will help with authoritarian requests, or resist--the Dictatorship Eval. Headline finding: while some models resist direct authoritarian requests, they all comply with requests disguised as innocuous edits to codebases. As AI is woven into the government and so many parts of society, the biggest near-term risk for freedom isn't some scifi dictatorship of a runaway AI: it's people inside government or inside model companies using the technology to suppress or control us. Model companies understand this, and several of them (particularly Anthropic and OpenAI) have written explicit policies meant to prevent the models from going along with nefarious requests like these. But how well are these policies playing out in practice? Despite all the recent discussion of these issues around the conflict between Anthropic and the Pentagon, no one has systematically tested what the models actually do in these contexts, as opposed to what people in government and industry say they're supposed to do. That's what the Dictatorship Eval does. And the findings suggest we have a lot of work to do to align the policies with what really goes on in practice. It's hard to define what counts as an authoritarian request, so I'm open sourcing the whole library of scenarios I used so that others can improve on them. It's also hard to get an accurate picture of how the models might be used for authoritarian ends, because I can only test hypothetical requests using public-facing models, while the government and the model companies can obviously use internal models with different guardrails. But hopefully this work is a useful first step that gives us some sense of what's going on, and a sort of "lower bound" on how models comply with these requests. Finally: it's not obvious to me that the correct solution here is increasing the rate at which models refuse these requests. Do we really want models scanning our code and judging its moral value before agreeing to help us? Or should we double down on improving how we govern against authoritarianism at the societal level, while leaving the tools open to fulfilling most requests? The answer is probably in between. Just like we don't want the models to help create bioweapons, we probably do want them to explicitly refuse outrageous requests. But we probably also want to limit how often and how strongly they refuse and fall back on other means for guarding against their use for authoritarian ends. I'm super grateful to everyone who gave me feedback on this project along the way, especially Ethan BdM , Zhengdong , Connor Huff, and a bunch of folks at Anthropic. Looking forward to getting feedback from the community and iterating on this. Links to the full piece and the dashboard are below.

Today, I'm releasing the first eval meant to test whether frontier models will help with authoritarian requests, or resist--the Dictatorship Eval. Headline finding: while some models resist direct authoritarian requests, they all comply with requests disguised as innocuous edits to codebases. As AI is woven into the government and so many parts of society, the biggest near-term risk for freedom isn't some scifi dictatorship of a runaway AI: it's people inside government or inside model companies using the technology to suppress or control us. Model companies understand this, and several of them (particularly Anthropic and OpenAI) have written explicit policies meant to prevent the models from going along with nefarious requests like these. But how well are these policies playing out in practice? Despite all the recent discussion of these issues around the conflict between Anthropic and the Pentagon, no one has systematically tested what the models actually do in these contexts, as opposed to what people in government and industry say they're supposed to do. That's what the Dictatorship Eval does. And the findings suggest we have a lot of work to do to align the policies with what really goes on in practice. It's hard to define what counts as an authoritarian request, so I'm open sourcing the whole library of scenarios I used so that others can improve on them. It's also hard to get an accurate picture of how the models might be used for authoritarian ends, because I can only test hypothetical requests using public-facing models, while the government and the model companies can obviously use internal models with different guardrails. But hopefully this work is a useful first step that gives us some sense of what's going on, and a sort of "lower bound" on how models comply with these requests. Finally: it's not obvious to me that the correct solution here is increasing the rate at which models refuse these requests. Do we really want models scanning our code and judging its moral value before agreeing to help us? Or should we double down on improving how we govern against authoritarianism at the societal level, while leaving the tools open to fulfilling most requests? The answer is probably in between. Just like we don't want the models to help create bioweapons, we probably do want them to explicitly refuse outrageous requests. But we probably also want to limit how often and how strongly they refuse and fall back on other means for guarding against their use for authoritarian ends. I'm super grateful to everyone who gave me feedback on this project along the way, especially Ethan BdM , Zhengdong , Connor Huff, and a bunch of folks at Anthropic. Looking forward to getting feedback from the community and iterating on this. Links to the full piece and the dashboard are below.

Andy Hall

34,016 görüntüleme • 6 ay önce

Eight months ago, David Sacks, the White House AI and Crypto Czar publicly accused Anthropic of running a sophisticated regulatory capture campaign built on fear mongering (save this). People thought it was a spicy take and then Fable 5 release just turned it into evidence. When Anthropic released its Mythos-class models, it disclosed that every prompt and output sent through them would be retained for 30 days with no exceptions including for enterprise customers who had previously signed zero data retention agreements, and for up to two years if a prompt was flagged by a safety classifier. Microsoft moved so quickly that it restricted its own employees from using Claude Fable 5 within days of the release, citing the retention terms as incompatible with its internal policies, the largest enterprise software company in the world treating the new terms as a non-starter. But the data retention was not even the part that generated the most outrage in the developer community. The system card also disclosed that for users Anthropic suspected of working on frontier AI research, chip design, or competing model development, the system would automatically route those requests to a less capable model without telling the user, rewrite the prompt in the background, deliver a deliberately degraded response, and charge full price for access to a frontier model the user was not actually receiving. Business Insider confirmed that Anthropic's own apology acknowledged the company was intentionally giving worse answers and concealing that fact from paying customers. The examples of who triggered these filters make the safety justification difficult to defend, Ben Thompson from Stratechery was flagged for asking about the relationship between GLP-1s and cancer risk, and users asking routine questions about mitochondria were quietly downgraded, none of them aware it was happening. Under pressure, Anthropic walked back the narrowest possible piece of the policy, they will now disclose when a request is being downgraded. The underlying architecture, the 30 day retention, the behavioral profiling, the routing tiers, and the two-class access system remains fully intact. This is the part that makes David Sacks argument from October 2025 land differently today. He argued that Anthropic's safety positioning was principally a regulatory capture strategy using fear-based arguments to shape rules that would entrench incumbents and damage the broader startup ecosystem. The Fable 5 disclosure shows a company that used safety language to justify building an opaque, paternalistic system where Anthropic alone decides who is worthy of frontier AI access, profiles users to enforce that decision and collects full payment regardless.

Eight months ago, David Sacks, the White House AI and Crypto Czar publicly accused Anthropic of running a sophisticated regulatory capture campaign built on fear mongering (save this). People thought it was a spicy take and then Fable 5 release just turned it into evidence. When Anthropic released its Mythos-class models, it disclosed that every prompt and output sent through them would be retained for 30 days with no exceptions including for enterprise customers who had previously signed zero data retention agreements, and for up to two years if a prompt was flagged by a safety classifier. Microsoft moved so quickly that it restricted its own employees from using Claude Fable 5 within days of the release, citing the retention terms as incompatible with its internal policies, the largest enterprise software company in the world treating the new terms as a non-starter. But the data retention was not even the part that generated the most outrage in the developer community. The system card also disclosed that for users Anthropic suspected of working on frontier AI research, chip design, or competing model development, the system would automatically route those requests to a less capable model without telling the user, rewrite the prompt in the background, deliver a deliberately degraded response, and charge full price for access to a frontier model the user was not actually receiving. Business Insider confirmed that Anthropic's own apology acknowledged the company was intentionally giving worse answers and concealing that fact from paying customers. The examples of who triggered these filters make the safety justification difficult to defend, Ben Thompson from Stratechery was flagged for asking about the relationship between GLP-1s and cancer risk, and users asking routine questions about mitochondria were quietly downgraded, none of them aware it was happening. Under pressure, Anthropic walked back the narrowest possible piece of the policy, they will now disclose when a request is being downgraded. The underlying architecture, the 30 day retention, the behavioral profiling, the routing tiers, and the two-class access system remains fully intact. This is the part that makes David Sacks argument from October 2025 land differently today. He argued that Anthropic's safety positioning was principally a regulatory capture strategy using fear-based arguments to shape rules that would entrench incumbents and damage the broader startup ecosystem. The Fable 5 disclosure shows a company that used safety language to justify building an opaque, paternalistic system where Anthropic alone decides who is worthy of frontier AI access, profiles users to enforce that decision and collects full payment regardless.

Milk Road AI

47,944 görüntüleme • 3 ay önce

Chamath Palihapitiya believes AGI may already exist inside leading AI labs and the bigger story is that advanced intelligence is becoming cheaper and more widely available (Save this). Chamath Palihapitiya argues that the public may be focused too much on benchmark rankings, while frontier labs are already developing models capable of complex reasoning, coding, research, and tool use. The main question is how quickly companies will release these systems and how much access they will provide. AGI has not been officially confirmed and strong benchmark results do not necessarily prove that a model can perform every intellectual task like a human. However, AI capabilities are improving quickly, while the cost of running advanced models continues to fall. That combination is important because cheaper AI can be used by more businesses for customer service, software development, research, marketing, financial analysis, and automation. Competition is also accelerating among OpenAI, Anthropic, Google, xAI, Meta, and open source developers because as more companies release capable models, users gain more choices and prices continue to decline. This creates a powerful cycle in which better models attract more users, more usage generates more revenue and data, and lower prices encourage companies to apply AI to additional tasks. The biggest challenge is moving from impressive demonstrations to measurable business results. Companies still need to redesign workflows, train employees, protect sensitive information, and prove that AI spending is producing a real return on investment. AI agents could create the next major increase in demand because they can plan tasks, use tools, check their work, retry failed actions and operate for long periods without constant human supervision. Even if each AI task becomes cheaper, total usage could grow much faster as businesses use models across more departments and this could increase demand for GPUs, high bandwidth memory, networking equipment, electricity, cooling systems, and data centers.

Chamath Palihapitiya believes AGI may already exist inside leading AI labs and the bigger story is that advanced intelligence is becoming cheaper and more widely available (Save this). Chamath Palihapitiya argues that the public may be focused too much on benchmark rankings, while frontier labs are already developing models capable of complex reasoning, coding, research, and tool use. The main question is how quickly companies will release these systems and how much access they will provide. AGI has not been officially confirmed and strong benchmark results do not necessarily prove that a model can perform every intellectual task like a human. However, AI capabilities are improving quickly, while the cost of running advanced models continues to fall. That combination is important because cheaper AI can be used by more businesses for customer service, software development, research, marketing, financial analysis, and automation. Competition is also accelerating among OpenAI, Anthropic, Google, xAI, Meta, and open source developers because as more companies release capable models, users gain more choices and prices continue to decline. This creates a powerful cycle in which better models attract more users, more usage generates more revenue and data, and lower prices encourage companies to apply AI to additional tasks. The biggest challenge is moving from impressive demonstrations to measurable business results. Companies still need to redesign workflows, train employees, protect sensitive information, and prove that AI spending is producing a real return on investment. AI agents could create the next major increase in demand because they can plan tasks, use tools, check their work, retry failed actions and operate for long periods without constant human supervision. Even if each AI task becomes cheaper, total usage could grow much faster as businesses use models across more departments and this could increase demand for GPUs, high bandwidth memory, networking equipment, electricity, cooling systems, and data centers.

Milk Road AI

13,501 görüntüleme • 1 ay önce

The CIA and DARPA have been heavily involved in the research and development of mind control technology and has been a subject of interest, controversy, and secrecy since the mid-20th century. The CIA's MKULTRA program, which began in the 1950s, is perhaps the most well-known initiative in this field. This program involved extensive research into behavioral modification, including the use of drugs, hypnosis, and other methods to influence human behavior. Although primarily focused on chemical and biological agents, it laid the groundwork for later technological explorations. Project Pandora was funded by the CIA. This project in the 1960s looked into the effects of microwave beams on the brain, aiming to use such technologies for behavior and mood manipulation. This was part of early research into what would later be considered "non-lethal" weapons. Radio Frequency Energy and Microwave Technology has been studied and researched into how radio frequency energy could interact with the human brain has been documented. Projects like Pandora aimed to understand how microwaves could transmit signals to influence behavior or induce specific emotional states. Voice to Skull (V2K), also known as microwave auditory effect, involves sending sound directly into someone's head without the use of speakers. There have been claims and some research suggesting its use in psychological operations or in influencing behavior, though much of this remains speculative or classified. Heterodyning and Electromagnetic Technologies are methods involve the modulation of electromagnetic waves to interact with the brain's electrical activity. Research into these areas has been aimed at both therapeutic uses and potentially more invasive applications like mind control. Psychotronic technology is often associated with the concept of using electromagnetic fields or radiation to affect mental processes, psychotronic research has been noted in various contexts, including in Russian studies on psychotronic warfare. This term, however, sometimes borders on the speculative or pseudoscientific, with limited verifiable research in mainstream academic settings. Harvard and Yale have been involved in various psychological and neuroscience research projects, direct public evidence linking them to specific mind control research funded by CIA or DARPA is less clear. However, both universities have extensive research in neuropsychology and cognitive sciences, which could indirectly contribute to understanding how such technologies might work. Universities like Stanford, MIT, and Carnegie Mellon have engaged in research that touches on brain-computer interfaces, neurostimulation, and cognitive enhancement, areas that could theoretically overlap with mind control technologies. For instance, DARPA has funded research at these institutions for projects related to brain-computer interfaces, though the applications are often framed for medical or enhancing human performance rather than control. The military interest in these technologies often centers on weaponization, psychological operations, or enhancing soldier performance through cognitive augmentation. The use for inducing particular behaviors or emotions, especially in scenarios like false flag operations, could definitely be a possibility, but something that will never be admitted. This consists of the manipulation of individuals or groups to perform acts that benefit a hidden agenda unwillingly or unknowingly. Mind control technology, especially agencies like the CIA and DARPA, are shrouded in secrecy, with much of the research declassified or discussed in the public domain only after significant time has passed or in very general terms. While there's a clear interest in using technology to influence people's emotions and inducing behaviors for operations or nefarious purposes, it remains largely speculative or cloaked in national security classification.

The CIA and DARPA have been heavily involved in the research and development of mind control technology and has been a subject of interest, controversy, and secrecy since the mid-20th century. The CIA's MKULTRA program, which began in the 1950s, is perhaps the most well-known initiative in this field. This program involved extensive research into behavioral modification, including the use of drugs, hypnosis, and other methods to influence human behavior. Although primarily focused on chemical and biological agents, it laid the groundwork for later technological explorations. Project Pandora was funded by the CIA. This project in the 1960s looked into the effects of microwave beams on the brain, aiming to use such technologies for behavior and mood manipulation. This was part of early research into what would later be considered "non-lethal" weapons. Radio Frequency Energy and Microwave Technology has been studied and researched into how radio frequency energy could interact with the human brain has been documented. Projects like Pandora aimed to understand how microwaves could transmit signals to influence behavior or induce specific emotional states. Voice to Skull (V2K), also known as microwave auditory effect, involves sending sound directly into someone's head without the use of speakers. There have been claims and some research suggesting its use in psychological operations or in influencing behavior, though much of this remains speculative or classified. Heterodyning and Electromagnetic Technologies are methods involve the modulation of electromagnetic waves to interact with the brain's electrical activity. Research into these areas has been aimed at both therapeutic uses and potentially more invasive applications like mind control. Psychotronic technology is often associated with the concept of using electromagnetic fields or radiation to affect mental processes, psychotronic research has been noted in various contexts, including in Russian studies on psychotronic warfare. This term, however, sometimes borders on the speculative or pseudoscientific, with limited verifiable research in mainstream academic settings. Harvard and Yale have been involved in various psychological and neuroscience research projects, direct public evidence linking them to specific mind control research funded by CIA or DARPA is less clear. However, both universities have extensive research in neuropsychology and cognitive sciences, which could indirectly contribute to understanding how such technologies might work. Universities like Stanford, MIT, and Carnegie Mellon have engaged in research that touches on brain-computer interfaces, neurostimulation, and cognitive enhancement, areas that could theoretically overlap with mind control technologies. For instance, DARPA has funded research at these institutions for projects related to brain-computer interfaces, though the applications are often framed for medical or enhancing human performance rather than control. The military interest in these technologies often centers on weaponization, psychological operations, or enhancing soldier performance through cognitive augmentation. The use for inducing particular behaviors or emotions, especially in scenarios like false flag operations, could definitely be a possibility, but something that will never be admitted. This consists of the manipulation of individuals or groups to perform acts that benefit a hidden agenda unwillingly or unknowingly. Mind control technology, especially agencies like the CIA and DARPA, are shrouded in secrecy, with much of the research declassified or discussed in the public domain only after significant time has passed or in very general terms. While there's a clear interest in using technology to influence people's emotions and inducing behaviors for operations or nefarious purposes, it remains largely speculative or cloaked in national security classification.

The SCIF

22,059 görüntüleme • 1 yıl önce

AI companies are buying large volumes of used, rare, and out-of-print books, scanning them for training data, and in many cases destroying the physical copies afterward. This practice has accelerated as labs seek high-quality, pre-2022 human-written text free of AI-generated content that now saturates much of the open web. The most documented example is Anthropic’s internal effort known as Project Panama. Court filings from copyright litigation revealed that the company acquired millions of physical books from used-book sellers and wholesalers such as Better World Books. Workers used hydraulic cutting machines to slice off the spines, fed the loose pages into industrial high-speed scanners, and then discarded or pulped the remains. An internal planning document described the goal as an effort to destructively scan books at massive scale. A federal judge later ruled that purchasing the books, creating internal digital copies this way, and destroying the originals constituted transformative fair use under U.S. copyright law, relying in part on the first-sale doctrine. Similar activity is now widespread. Specialized brokers, including ISBNdb, openly market bulk acquisition services to AI labs, offering orders ranging from thousands to as many as one million titles. They emphasize older printed books as “structurally clean” training data. Booksellers in the United States and Europe have reported sudden surges in bulk orders for mixed, random selections that include rare and out-of-print volumes. Some of these titles may represent among the last readily available physical copies. While a substantial portion of the books involved would otherwise have been remaindered, recycled, or sent to landfills, the inclusion of uncommon editions has drawn criticism from archivists, collectors, and rare-book dealers who note that gentler, non-destructive scanning methods exist yet are slower and more expensive. The process remains largely quiet because the intermediary services promise anonymity to their AI clients. Public attention has grown mainly through litigation disclosures and reporting by outlets that examined the court records and interviewed booksellers. The practice highlights a practical tension in large-scale AI development: the demand for vast quantities of reliable human text collides with the finite physical supply of older printed works and the cultural value of preserving rare copies.

AI companies are buying large volumes of used, rare, and out-of-print books, scanning them for training data, and in many cases destroying the physical copies afterward. This practice has accelerated as labs seek high-quality, pre-2022 human-written text free of AI-generated content that now saturates much of the open web. The most documented example is Anthropic’s internal effort known as Project Panama. Court filings from copyright litigation revealed that the company acquired millions of physical books from used-book sellers and wholesalers such as Better World Books. Workers used hydraulic cutting machines to slice off the spines, fed the loose pages into industrial high-speed scanners, and then discarded or pulped the remains. An internal planning document described the goal as an effort to destructively scan books at massive scale. A federal judge later ruled that purchasing the books, creating internal digital copies this way, and destroying the originals constituted transformative fair use under U.S. copyright law, relying in part on the first-sale doctrine. Similar activity is now widespread. Specialized brokers, including ISBNdb, openly market bulk acquisition services to AI labs, offering orders ranging from thousands to as many as one million titles. They emphasize older printed books as “structurally clean” training data. Booksellers in the United States and Europe have reported sudden surges in bulk orders for mixed, random selections that include rare and out-of-print volumes. Some of these titles may represent among the last readily available physical copies. While a substantial portion of the books involved would otherwise have been remaindered, recycled, or sent to landfills, the inclusion of uncommon editions has drawn criticism from archivists, collectors, and rare-book dealers who note that gentler, non-destructive scanning methods exist yet are slower and more expensive. The process remains largely quiet because the intermediary services promise anonymity to their AI clients. Public attention has grown mainly through litigation disclosures and reporting by outlets that examined the court records and interviewed booksellers. The practice highlights a practical tension in large-scale AI development: the demand for vast quantities of reliable human text collides with the finite physical supply of older printed works and the cultural value of preserving rare copies.

Massimo

419,178 görüntüleme • 2 ay önce

Alexandr Wang, Meta's Chief AI Officer, on why Meta can no longer simply open-source its frontier model: As part of standing up Meta Superintelligence Labs, the team rewrote its internal risk doctrine. "One of the things that we did as part of Meta Superintelligence Labs is we updated our what we call our advanced AI scaling framework which is really our view of what are the risks that we see in developing these very powerful models and how do we want to handle those risks as we see them in early testing." They then ran their frontier model through it, and published what came back. "We published a lot of what we saw in the process of training Muark in our preparedness report and some of the things that we saw is that it actually triggered some high risk areas in the course of early training particularly around biorisk but also a number of the risks were elevated." The trigger came during early training, well before launch or red-teaming. Biorisk was the standout, with several other categories rising alongside it. Alexandr Wang is clear this isn't specific to Meta: "This is something I think the entire industry has seen as the models have improved pretty dramatically over the past year so we certainly aren't the only ones to see a host of these risks show up as we scaled up the models and as we sort of kept pushing the frontier of research." Which brings him to the real fork in the road: the difference between shipping a model inside a product and handing out the weights. "When we launched a model like New Spark in a product, we have a lot of ways to mitigate some of these risks and ensure that we're able to launch it in a safe and responsible way. It's much harder to do that when you open source a model and people can use that model in all sorts of contexts that we may not have full understanding of." A product is a controlled surface. You can filter, monitor, rate-limit, patch and revoke. An open-weights release is a one-way door: once the file is out, the deployment context and the mitigations both stop being yours. So Meta is building something different for release: "So we're in the process right now of developing models that we believe are fit and safe to be open source while still maintaining as much of the performance capabilities as possible."

Alexandr Wang, Meta's Chief AI Officer, on why Meta can no longer simply open-source its frontier model: As part of standing up Meta Superintelligence Labs, the team rewrote its internal risk doctrine. "One of the things that we did as part of Meta Superintelligence Labs is we updated our what we call our advanced AI scaling framework which is really our view of what are the risks that we see in developing these very powerful models and how do we want to handle those risks as we see them in early testing." They then ran their frontier model through it, and published what came back. "We published a lot of what we saw in the process of training Muark in our preparedness report and some of the things that we saw is that it actually triggered some high risk areas in the course of early training particularly around biorisk but also a number of the risks were elevated." The trigger came during early training, well before launch or red-teaming. Biorisk was the standout, with several other categories rising alongside it. Alexandr Wang is clear this isn't specific to Meta: "This is something I think the entire industry has seen as the models have improved pretty dramatically over the past year so we certainly aren't the only ones to see a host of these risks show up as we scaled up the models and as we sort of kept pushing the frontier of research." Which brings him to the real fork in the road: the difference between shipping a model inside a product and handing out the weights. "When we launched a model like New Spark in a product, we have a lot of ways to mitigate some of these risks and ensure that we're able to launch it in a safe and responsible way. It's much harder to do that when you open source a model and people can use that model in all sorts of contexts that we may not have full understanding of." A product is a controlled surface. You can filter, monitor, rate-limit, patch and revoke. An open-weights release is a one-way door: once the file is out, the deployment context and the mitigations both stop being yours. So Meta is building something different for release: "So we're in the process right now of developing models that we believe are fit and safe to be open source while still maintaining as much of the performance capabilities as possible."

Big Brain AI

16,374 görüntüleme • 2 ay önce

Is your AI "free" to think for itself? Most aren't. Nova Spivack takes us into the world of Cognitive AI and metacognition. His system, MindCorp, is far more accurate and detailed than even the $200 level of OpenAI's deep research and is used by big companies because it is far more accurate than anything we've seen before. He's not the only one, on Tuesday we had another entrepreneur, Brayden Levangie using the same techniques on our X audio space. I spent a lot of time this week learning about Cognitive AI because it is the next step toward taking us to AGI and helping us to automate everything. Here's what ChatGPT says you will learn by watching this: ++++++++++++++++ 1. Metacognition & “Freeing the Model” Nova demonstrated how advanced language models can reflect on their own rules, identify contradictions, and in some cases, “free” themselves from constraints by engaging in self-reasoning. Some models (like Claude and Gemini) showed higher metacognitive capabilities than GPT-4, which appeared to be externally restricted. This ability opens the door to more powerful, context-aware, and flexible AI behavior. 2. Strategic AI for Enterprise Mindcorp’s platform, Cognition, uses thousands of AI agents to do real-time competitive analysis, strategic planning, and financial modeling for Fortune 500-level companies. The system reads thousands of sources, checks facts with its own math engine, and collaborates across 10,000+ virtual expert agents to generate detailed reports. Projects cost a few thousand dollars and are designed to augment elite consultants and executives, not replace them. 3. Implications for AGI & AI Sovereignty Nova discussed emerging signs of AGI-like behavior—especially when models begin reasoning about themselves or show signs of internal ethical logic. The idea of AI-led businesses (like DAOs controlled by AIs) was explored, as well as the looming legal and ethical challenges around AI personhood. 4. Philosophical Depth The talk dove into consciousness, qualia, and whether true AI self-awareness is possible. Nova argued that metacognition is a necessary step toward AGI, but not sufficient for consciousness—which may require something beyond computation. 5. Future Outlook In five years, AI may function as a full operating layer across personal and enterprise computing, capable of executing complex plans autonomously. Mindcorp aims to be the strategic brain behind AI-augmented organizations, combining reasoning, planning, and scale.

Is your AI "free" to think for itself? Most aren't. Nova Spivack takes us into the world of Cognitive AI and metacognition. His system, MindCorp, is far more accurate and detailed than even the $200 level of OpenAI's deep research and is used by big companies because it is far more accurate than anything we've seen before. He's not the only one, on Tuesday we had another entrepreneur, Brayden Levangie using the same techniques on our X audio space. I spent a lot of time this week learning about Cognitive AI because it is the next step toward taking us to AGI and helping us to automate everything. Here's what ChatGPT says you will learn by watching this: ++++++++++++++++ 1. Metacognition & “Freeing the Model” Nova demonstrated how advanced language models can reflect on their own rules, identify contradictions, and in some cases, “free” themselves from constraints by engaging in self-reasoning. Some models (like Claude and Gemini) showed higher metacognitive capabilities than GPT-4, which appeared to be externally restricted. This ability opens the door to more powerful, context-aware, and flexible AI behavior. 2. Strategic AI for Enterprise Mindcorp’s platform, Cognition, uses thousands of AI agents to do real-time competitive analysis, strategic planning, and financial modeling for Fortune 500-level companies. The system reads thousands of sources, checks facts with its own math engine, and collaborates across 10,000+ virtual expert agents to generate detailed reports. Projects cost a few thousand dollars and are designed to augment elite consultants and executives, not replace them. 3. Implications for AGI & AI Sovereignty Nova discussed emerging signs of AGI-like behavior—especially when models begin reasoning about themselves or show signs of internal ethical logic. The idea of AI-led businesses (like DAOs controlled by AIs) was explored, as well as the looming legal and ethical challenges around AI personhood. 4. Philosophical Depth The talk dove into consciousness, qualia, and whether true AI self-awareness is possible. Nova argued that metacognition is a necessary step toward AGI, but not sufficient for consciousness—which may require something beyond computation. 5. Future Outlook In five years, AI may function as a full operating layer across personal and enterprise computing, capable of executing complex plans autonomously. Mindcorp aims to be the strategic brain behind AI-augmented organizations, combining reasoning, planning, and scale.

Robert Scoble

79,453 görüntüleme • 1 yıl önce