reSee.it - Related Video Feed

Customization allows using the same engine for each robot to rapidly create new robotic characters. This is presented as a very cool feature. One of the biggest problems faced is then mentioned, but not elaborated upon.

There's enormously more information in the physical world—vision, touch, and audition—than in all human texts, so human-level AI will not emerge without systems learning from observing the world. Psychologists note this vast amount of information in the real world compared with text. Babies acquire background knowledge early on through observation, a form of self-supervised learning we must reproduce to reach animal or human intelligence. They develop notions such as object permanence—the idea that hidden objects still exist—along with stability and natural object categories without names. They also grasp intuitive physics—gravity, inertia, momentum. By about nine months they have this; by six months a scenario where an object floats may not surprise them, while by ten months they are surprised, having learned that unsupported objects fall. This learning occurs through observation and limited interaction.

Speaker 0 argues that while AI systems can solve conjectures that already exist, they currently cannot generate genuinely new hypotheses or novel ideas about how the world might work. He suggests that achieving such a capability would require features that go beyond solving established problems, pointing to the need for long-term planning, improved reasoning, and a functioning world model. A world model would allow the system to have a more accurate internal understanding of the physics of the world, enabling it to run simulations and test its own hypotheses in its own mind—processes that human scientists typically employ when developing new theories or discoveries. He notes that this is the type of capability that appears to be missing in contemporary AI systems. Speaker 1 asks for clarification on the concept of world models, particularly how they differ from large language models (LLMs). Speaker 0 explains that while current models—such as LLMs—are predominantly text-based, there are foundation models like Gemini that can handle multiple modalities, including images, video, and audio. Nevertheless, even with multimodal capabilities, these systems still do not truly understand the physics or causality of the world, nor how one event affects another. The question of whether an AI can plan far into the future is linked to the broader idea of world models. Speaker 0 emphasizes that to truly understand how the world works—to potentially invent something new or to explain something that was previously unknown, effectively performing scientific theorizing—an AI needs an accurate model of how the world operates. This involves starting from intuitive physics and extending to more complex domains such as biology and economics. In essence, a robust world model would enable the AI to reason about causality, simulate outcomes, and test hypotheses over long timescales, mirroring the capabilities that characterize human scientific inquiry. The dialogue contrasts the current state of AI, which is strong in pattern recognition and problem-solving within existing knowledge, with the envisioned potential of AI to generate new theories through a comprehensive internal model of the world.

I'm using Jetson-powered robots learning to walk in Isaac Sim. This is the orange one and that's the famous green one.

We're XAI, and our mission is to understand the universe by rigorously pursuing truth, even if it's politically incorrect. We're excited to introduce Grok-3, a significant leap from Grok-2, thanks to our incredible team. Grok, from Heinlein's novel, means to fully and profoundly understand. Our progress in the last 17 months has been unprecedented, driven by a dedicated team and substantial compute power. To accelerate further, we built our own data center in just 122 days, housing 100k GPUs, and then doubled the capacity in 92 days. Grok-3 boasts 10x more compute and excels in math, science, and coding. A blind test showed Grok-3 leading across all categories. We're continuously improving it, so you'll see updates daily. We've added advanced reasoning capabilities to Grok, tested with physics problems and creative games, showcasing the beginnings of creativity.

The presentation outlines the rapid, multi-faceted progress of xAI over two-and-a-half years, emphasizing velocity, scope, and ambition across four main application areas and their supporting infrastructure. Key accomplishments and claims - xAI is two-and-a-half years old and has achieved leadership in voice, image, and video generation, with Grok forecasting (Grok 4.20) beating all others on forecasting. The team notes it is generating more images and video than all competitors combined. - Grokopedia is introduced as a forthcoming Encyclopedia Galactica, intended to distill all knowledge with video and image data not present on Wikipedia. - The company achieved a 100,000 GPU-hour training cluster and is about to reach 1,000,000 GPU-hour equivalents in training. - The overarching message: velocity and acceleration matter more than position; xAI asserts it is moving faster than any competitor in multiple arenas. Organizational structure and manpower changes - The company has reorganized as it scales, moving from a startup phase to a more structured organization with four main application areas and supporting infrastructure. - The four areas are GrokMain and Voice, a coding-specific model (Grok Code and related efforts housed under MacroHard for full digital emulation of entire companies), an image and video model (Imagine), and the infrastructure layers. - Some early contributors have departed, and the leadership expresses gratitude for their contributions while welcoming new structure and continued growth. Four application areas and their leaders - GrokMain and Voice: Merged into one team; notable progress includes developing a voice model in six months after lacking an in-house product previously, leading to Grok voice agent API used in more than 2,000,000 Teslas. The aim is for Grok to be genuinely useful across engineering, law, medicine, and more. - Imagine (image and video): Since inception six months ago, Imagine has moved from no internal diffusion code to being integrated across all product surfaces, including X app; users generate close to 50,000,000 videos per day and 6,000,000,000 images in the last 30 days, with Imagine v1 released two weeks prior and multiple releases planned. The team claims to top leaderboards in many areas and envisions transforming imagined content into reality, with rapid iteration (daily product updates, biweekly model updates). - MacroHard: Focused on full digital emulation of companies and high-level automation of tasks that today require human labor; the project aims to build end-to-end digital emulation of human activities across domains like rockets, AI chips, physics, customer service, etc. MacroHard is presented as potentially the most important and lucrative project, with “the words MacroHard” painted on the roof of the training cluster as a symbolic representation of its scope. - Core infrastructure and tooling: Several teams describe their roles, including: - ML infrastructure and tooling (building training, inference, and deployment tooling; solving data center reliability and scale challenges; recounting a major pretraining system rewrite at 30k scale). - Reinforcement learning and inference (scaling to millions of chips, resilience, and hardware-failure handling). - JAX and low-level GPU stack (supporting multi-tenant training, custom optimizations). - Kernels team (low-level GPU optimization, microsecond-scale performance). - Data center and supercomputing infrastructure (Memphis data center; the largest GPU cluster; vertical integration across architecture, mechanical, and electrical disciplines; pursuit of high PUE and efficient power use). - Public-facing platforms and products (X platform, X Chat, X Money), with plans to open-source components of the recommendation algorithm and Grok Chat, plus the launch of a standalone X Chat app designed for general messaging with features like encrypted messaging and multi-user video calls. - Content and outreach: The X platform’s growth is highlighted, with heavy emphasis on engagement, onboarding improvements, and multi-surface enhancements. Key metrics and projections - User and content metrics: nearly 50,000,000 videos generated daily via Imagine and 6,000,000,000 images generated in the last 30 days. The team positions these figures as exceeding all competitors combined. - Computational intensity: a current milestone of 100,000 GPU-hours, with a trajectory toward 1,000,000 GPU-hours; the aim is to sustain unprecedented scale. - Product roadmap: Grok four-point-two (and larger variants) are anticipated to advance within two to three months; Imagine continues to evolve rapidly with ongoing releases; MacroHard is expected to become central to the company’s long-term strategy. - Platform and services: X platform revenue, with subscriptions driving ARR in the hundreds of millions; a standalone X Chat app is planned; X Money is moving from closed beta to external beta and then global launch; the combined strategy includes SpaceX alignment for orbital data centers to accelerate AI training and inference beyond Earth, including plans for moon-based factories, a mass driver, and satellite deployment. Space and future vision - Musk discusses a broader arc: merging xAI with SpaceX to scale AI compute through orbital data centers, with ambitions to launch millions of satellites, mass drivers on the Moon, and expansive solar-system-wide AI infrastructure. The goal is to extend beyond Earth and explore the universe, potentially meeting alien civilizations. Note: The closing promotional content for AG1 is not included in this summary per instructions to omit promotional material.

Speaker 0: I think what a lot of people aren't really familiar with is the bioengineering aspect of this, and we only need to look to this recently published headline from the Daily Mail, which was resurfaced, declassified CIA files that revealed a chilling blueprint to manipulate Americans' minds through covert drugging with vaccines. And it's not just vaccines that was in that blueprint. It's also the food, the water supply, pretty much altering our state of mind and our biology through all of these methods. And this is going back all the way to the fifties. One can only imagine how far they've come now, but you've been digging into this, and you have a bit of an idea as to how far they've come. To us about your latest research. Speaker 1: So you're absolutely right. And this has been, you know, a slow progression. Nothing is just being, you know, introduced new. I mean, it the technology has advanced, but it's been going on for decades decades, hundreds of years. And when you think about pharmaceuticals, the the apparatus of pharmaceuticals, they are all they it is medicinal chemistry, which is synthetic materials, synthetic biology, engineered bacteria, yeasts, molds, and all of those things like you just said. We have we are being assaulted with these these materials, which are now considered devices, you know, with the manipulated EMF and frequencies. And all of those are to exactly what you just said, weaken the system. And really this pro this slow progression of a we're in the midst of a forced evolution to become providers of a synthetic material, hybrid synthetic material. So we'll continue to produce as we do because the humanity's biological systems are by design meant to thrive and recycle and and repurpose themselves, but to survive. And so we accept these synthetic materials, and we and our body slowly begin to make accommodations to those mutations, natural mutations, but also so much of these so much of the synthetic material is coded to go in and trigger a mutation or to forcibly cause a mutation. So we literally are walking around. I mean, all of us, and it goes from the tiny little mushroom that's growing in the woods to, you know, aquatic life to every single biological electrical system, the nervous system, you know, is based on frequency. It's based on electricity. And so that is that's what's being attacked is the nervous system and the immune systems of every living being. Speaker 0: Now you're talking about some very important things here, Lisa. You've sent me this article from Medium titled the synthetic nervous system, a blueprint for physical AI. And in this article, it talks about how for the past decade, AI has lived primarily in a box, but now, our, you know, our interaction with AI has been linguistic and digital. We've cracked the code apparently, completely on generative AI, unlocking the ability to, listen to this, manipulate symbols, pixels, and code at scale, but we're now entering a far more complex epoch, the era of physical AI. And they are talking about the transition from AI that thinks to AI that acts. So they're saying the intelligence behind humanoid robots. They also give, you know, autonomous systems and things of this nature. My concern is that their plan stated goal is that they want humans to integrate with AI. This is something that even Elon Musk itself has said we need to do in order to stay relevant. And your research shows that they're already in the process of doing that. Talk to us a little bit about that. Speaker 1: Yes. And probably have. We and and, you know, I think that life as we know it will fairly stay the same because what the integration is through, and you've heard of this, is the digital twin. You know, assigning each of us a representative in the AI ecosystem, ecosystem, which which is is a a digital twin. But that digital twin is able to function and, perform because it is it is based off of your data, your biological data, your, that they are going in and removing and stealing through the infiltrators and facilitators that is vaccines, bioengineered foods, bioengineered bacteria. The, you know, the pharmaceutical industry is the perfect setup, and it's only one of one setup that goes in, and now these are all synthetic material devices. They work off of Wi Fi. They're software platforms, and they are all digital. And they are being monitored by the Department of Energy, HHS, MITRE now, these private companies and private oligarch, you know, tech companies that all have access to our free our our inner, you know, biological data DNA and and everything. And so that the AI platform, in order for it to succeed and for its longevity, there has to be a cohesive connection between humanity because we are the fuel that is going to feed that AI ecosystem. And it cannot it it's not gonna be one or the other. It has to work cohesively, and and they have to be joined. And how the the joining of those literally is through an infiltration system, which is primarily vaccines and engineered pathogens.

Mike Adams discusses concerns about the global build-out of data centers and presents a multi-part theory about their purpose and implications. He notes that a tweet he posted went viral, drawing responses from figures like Jimmy Dore and Rizwan Virk. He frames his talk as a theory, not a confirmed prediction, and plans to cover it in two parts. Key data and observations - There are about 11,000 existing data centers worldwide. The map and graphics Adams shares focus on 3,000 new or planned/construction sites, showing locations, size, power use, water use, land area, and investment needs. - In Piketon, Ohio, and other U.S. sites (including multiple facilities in Ohio and Texas), as well as Abu Dhabi, Shanghai, Tokyo, Malaysia, and other locations, there are large data centers under construction or announced. The lines in the AI-generated map may mis-point geographically, but the cities and nations listed are accurate. - The aggregate planned/under-construction capacity projects to about 190 gigawatts of power draw once completed. - The projected annual power consumption for these new centers would exceed 1,200 terawatt-hours per year, which Adams compares to about 10% of all power produced by China. - The centers would occupy over 1,000 square kilometers and use about 15+ billion liters of water per year, with some water potentially drawn from neighborhoods or households. Revenue and purpose questions - Adams argues there is not enough AI business, web hosting, data storage, or overall demand to justify the scale of the investment, implying the revenue model may be inadequate to pay back these projects. - He contrasts various high-profile tech figures—Tesla, Sam Altman, and Mark Zuckerberg—suggesting that the motives behind these data center buildups extend beyond serving immediate consumer compute needs, hinting at broader or longer-term strategic aims. Foundational ideas about AI and intelligence - He cites Jan LeCun (referenced as a leading AI researcher) arguing that the current structure of large language models (LLMs) is a dead end for achieving AGI or superintelligence due to gaps in physical-world understanding, memory, and long-term planning. Memory is said to be improving with newer context-handling approaches, but physical-world understanding and planning are highlighted as critical gaps. - LeCun’s idea mentioned is the development of world models and JEPPA architectures that learn from sensory inputs to understand and interact with the physical environment, rather than solely processing language statistics. - Adams suggests that the only viable path to practical superintelligence is to train AI systems in simulated three-dimensional worlds, where physics, gravity, time, light, touch, and other sensory inputs are experienced. He argues that simulated worlds can run at speeds far faster than the real world, limited only by compute and hardware bandwidth. - He mentions NVIDIA’s announced world simulator for training robots as an example of three-dimensional world simulations used for reinforcement learning and rapid iteration. - The concept of digital worlds is tied to the idea of digital evolution or Darwinism: billions of parallel simulated worlds could nurture AI entities that grow and potentially be summoned into our three-dimensional reality. He notes that a simulation-based approach could produce agents whose capabilities enable real-world deployment after learning in fast, rich simulations. - Adams discusses practical applications of three-dimensional simulations beyond AI self-improvement, including autonomous vehicle testing (synthetic data), manufacturing and robotics on factory floors, military scenario planning, surgical robotics, and pilot training. He emphasizes that the more realistic the simulation, the more reliable the results for real-world tasks and decisions. - He invokes the simulation hypothesis, suggesting a link between building simulated worlds and the possibility that our own reality could be a simulation. He plans to address evidence for the simulation hypothesis in part two, along with how simulated beings might be “summoned” into our world. Closing - Adams signals a two-part structure, with Part 1 covering data center build-out, AI constructs, and the simulation framework; Part 2 promising to address the simulation hypothesis with evidence and the idea of summoning advanced AI from simulations into the real world. Note: Promotional content regarding gold and silver investments and Battalion Metals has been omitted from this summary to align with content-avoidance requirements.

Demis Hassabis and Lex Fridman discuss whether classical learning systems can model highly nonlinear dynamical systems, including fluid dynamics, and what this implies for science and AI. - They note that Navier-Stokes dynamics are traditionally intractable for classical systems, yet Vio, a video generation model from DeepMind, can model liquids and specular lighting surprisingly well, suggesting that these systems are reverse engineering underlying structure from data (YouTube videos) and may be learning a lower-dimensional manifold that captures how materials behave. - The conversation pivots to Demis Hassabis’s Nobel Prize lecture conjecture that any pattern generated or found in nature can be efficiently discovered and modeled by a classical learning algorithm. They explore what kinds of patterns or systems might be included: biology, chemistry, physics, cosmology, neuroscience, etc. - AlphaGo and AlphaFold are used as examples of building models of combinatorially high-dimensional spaces to guide search in a tractable way. Hassabis argues that nature’s evolved structures imply learnable patterns, because natural systems have structure shaped by evolutionary processes. This leads to the idea of a potential complexity class for learnable natural systems (LNS) and the possibility that p = NP questions may be reframed as physics questions about information processing in the universe. - They discuss the view that the universe is an informational system, and how that reframes the P vs NP question as a fundamental question about modellability. Hassabis speculates that many natural systems are learnable because they have evolved structure, whereas some abstract problems (like factorizing arbitrary large numbers in a uniform space) may not exhibit exploitable patterns, possibly requiring quantum approaches or brute-force computation. - The dialogue examines whether there could be a broad class of problems that can be solved by polynomial-time classical methods when modeled with the right dynamics and environment—precisely the way AlphaGo and AlphaFold operate. Hassabis emphasizes that classical systems (Turing machines) have already surpassed many expectations by modeling complex biological structures and solving highly challenging tasks, and he believes there is likely more to discover. - They address nonlinear dynamical systems and whether emergent phenomena, such as cellular automata, chaos, or turbulence, might be amenable to efficient classical modeling. Hassabis notes that forward simulation of many emergent systems could be efficient, but chaotic systems with sensitive dependence on initial conditions may be harder to model. He argues that core physics problems, including realistic rendering of physics-like phenomena (e.g., liquids and light interaction), seem tractable with neural networks, suggesting deep structure to nature that can be captured by learning systems. - The conversation shifts to video and world models: Hassabis highlights VOI, video generation, and the hope that future interactive versions could create truly open-ended, dynamically generated game worlds and simulations where players co-create the experience with the environment, beyond current hard-coded or pre-scripted content. They discuss open-world games and the potential for AI to generate content on-the-fly, enabling personalized, ever-changing narratives and experiences. - They discuss Hassabis’s early love of games and his belief that games are a powerful testbed for AI and AGI. He describes the possibility of interactive VO-based experiences that are open-ended and highly responsive to player choices, with emergent behavior that surpasses current procedural generation. - The conversation touches the idea of an open-world world model for AGI: Hassabis imagines a system that can predict and simulate the mechanics of the world, enabling better scientific inquiry and perhaps even a “virtual cell” or virtual biology framework. They discuss AlphaFold as the static prediction of structure and the next step being dynamics and interactions, including protein–protein, protein–RNA, and protein–DNA interactions, and ultimately a model of a whole cell (e.g., yeast). - On the origin of life and origins science: they discuss whether AI could simulate the birth of life from nonliving matter, suggesting a staged approach with a “virtual cell” as a stepping-stone, then moving toward simulating chemical soups and emergent properties that could resemble life. - They consider the nature of consciousness and whether AI systems can or will ever have true consciousness. Hassabis leans toward the view that consciousness (and qualia) may be substrate-dependent and that a classical computer could model the functional aspects of intelligence; but he acknowledges unresolved questions about subjective experience and the potential differences between carbon-based and silicon-based processing. - They discuss the role of AGI in science: the potential for AI to propose new conjectures and hypotheses, to assist in scientific discovery, and perhaps to discover insights that humans might not reach on their own. They acknowledge that “research taste”—the ability to pick the right questions and design experiments meaningfully—is a hard capability for AI to replicate. - They explore the future of video games with AI: Hassabis describes the possibility of open-world, highly interactive experiences that adapt to players’ actions, creating deeply personalized narratives. He compares the future of AI-driven game design to the potential for AI to accelerate scientific progress by modeling complex systems, then translating insights into practical tools and products. - Hassabis discusses the practicalities of running large AI projects at Google DeepMind and Google, noting the balance of startup-like culture with the scale of a large corporation. He emphasizes relentless progress and shipping, while maintaining safety and responsibility, and maintaining collaboration across labs and competitors. - They address data and scaling: Hassabis emphasizes that synthetic data and simulations can help mitigate data scarcity, while real-world data remains essential to guide learning systems. He explains the dynamic between pre-training, post-training, and inference-time compute, noting the importance of balancing improvements across multiple objectives and avoiding overfitting benchmarks. - They discuss governance, safety, and international collaboration: they emphasize the need for shared standards, safety guardrails, and open science where appropriate, while acknowledging the risk of misuse by bad actors and the difficulty of restricting access to powerful AI systems without hampering beneficial applications. Hassabis suggests international cooperation and a CERN-like collaborative model for responsible progress. - They touch on the societal impact of AI: the potential for energy breakthroughs, climate modeling, materials discovery, and fusion, plus the broader economic and political implications. Hassabis anticipates a future where abundant energy reduces scarcity, enabling new levels of human flourishing, but acknowledges distributional concerns and governance challenges. - The dialogue ends with reflections on personal legacies and the human dimension: Hassabis discusses responding to criticism online, his MIT and Drexel affiliations, and the balance between research, podcasting, and public engagement. He emphasizes humility, continuous learning, and openness to collaboration across labs and cultures. Key themes and conclusions preserved from the discussion: - The possibility that many natural patterns are efficiently learnable by classical learning systems if the underlying structure is learned, a view supported by AlphaGo/AlphaFold successes and by phenomena like VOI’s handling of liquids and lighting. - A conjectured link between learnable natural systems and a formal complexity class like LNS, with the broader view that p versus NP is connected to physics and information in the universe. - The potential for classical AI to model complex, nonlinear dynamical systems, including fluid dynamics, with surprising accuracy, given sufficient structure and data. - The idea that nature’s evolutionary processes create patterns that can be reverse-engineered, enabling efficient search and modeling of natural systems. - The role of AI in science as a tool for conjecture generation, hypothesis testing, and accelerating discovery, possibly guiding experiments, reducing wet-lab time, and enabling “virtual cells” and larger-scale simulations. - The interplay between open-world game design, AI-based content creation, and future interactive experiences that adapt to individual players, including the vision of AI-driven world models for AGI. - The practical realities of building and shipping AI products at scale, balancing research breakthroughs with productization, and managing a large organization’s culture and governance to foster safety and innovation. - The ethical and societal questions around AGI: how to ensure safety, how to manage risk from bad actors, the need for international collaboration, governance, and a broad discussion about the role of technology in society. - A hopeful perspective on the long-term future: abundant energy, space exploration, and a transformed civilization driven by AI, with a focus on human values, curiosity, adaptability, and compassion as guiding forces. This summary preserves the essential claims and conclusions of the conversation, including the main positions about learnability, the role of evolution and structure in nature, the potential of classical systems to model complex phenomena, and the broad, multi-domain implications for science, gaming, energy, governance, and society.