正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

> Natural data is "generated" from a constrained hierarchical / compositional function. > Deep networks learns that hidden structure from polynomially few examples and creatively generate exponentially many valid new ones. > The depth of the network is what's important to overcome the curse of dimensionality, and potentially invalidate... show more

Machine Learning Street Talk

40,098 subscribers

13,826 次观看 • 4 个月前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

6 条评论

Machine Learning Street Talk 的头像

Machine Learning Street Talk4 个月前

> For data with hidden hierarchy, the data gap is exponentially large. > How could this be better? - models that predict their own abstractions/latents. > Naturally, @ylecun agrees!

Machine Learning Street Talk 的头像

Machine Learning Street Talk4 个月前

> Generic high-dimensional data is unlearnable, you need more data than atoms in the universe just to interpolate. > The way out of this is coarse graining, which is often exemplified by word2vec i.e. the concept of a "street" emerges from locality statistics. > But the most abstract concepts are the hardest to learn.

Matthieu wyart 的头像

Matthieu wyart4 个月前

Great fun, Tim. The part I'd underline: the same compositional structure that makes data learnable from polynomially few examples is what lets a model recombine them into exponentially many new valid ones: creativity and sample-efficiency are two sides of one coin.

Lonely nerd. 的头像

Lonely nerd.4 个月前

The one take away from the post AI hype is that LLMs and most researchers are interpolators. Very impressive ones. =-) 👇

Warley Vital 的头像

Warley Vital4 个月前

Interestingly, the previous post in my timeline was this one that points to a similar idea

Mike Pearson 的头像

Mike Pearson4 个月前

Processes can't be represented by single passes through a network. They're picking up the outputs of processes - which do have some stable patterns. Like memorizing times tables as opposed to learning multiplication as an algorithm. Recursion is required.

相关视频

MILEI: THE SO CALL 'INSENSITIVE LIBERTARIANS' LIFTED 10 MILLION PEOPLE OUT OF POVERTY "The disgusting liberals, the politically correct people, the 'cool' leftists, the 'sensitive' ones, those who love the poor so much that all they do is multiply them; they all would tell us that we were going to generate an explosion of poverty. Poverty rate was at 57% (...) however, with a huge work from the Ministry of Human Capital (...) the poverty data today is 33%. (...) It means that the 'insensitive libertarians' lifted more than 10 million people out of poverty." Javier Milei

MILEI: THE SO CALL 'INSENSITIVE LIBERTARIANS' LIFTED 10 MILLION PEOPLE OUT OF POVERTY "The disgusting liberals, the politically correct people, the 'cool' leftists, the 'sensitive' ones, those who love the poor so much that all they do is multiply them; they all would tell us that we were going to generate an explosion of poverty. Poverty rate was at 57% (...) however, with a huge work from the Ministry of Human Capital (...) the poverty data today is 33%. (...) It means that the 'insensitive libertarians' lifted more than 10 million people out of poverty." Javier Milei

Argentina's Milei News 🇦🇷🤝🌎

69,986 次观看 • 1 年前

We're thrilled to present ESM3 in Science Magazine. ESM3 is a generative language model that reasons over the three fundamental properties of proteins: sequence, structure, and function. Today we're making ESM3 available free to researchers worldwide via the public beta of an API for biological intelligence. Trained with over a trillion teraflops of compute, this is the first time a model of this scale has been trained for biology, pushing the frontier of AI for biological discovery and engineering. ESM3 learns to represent the immense complexity of protein biology, learning from billions of natural proteins. From this training it developed the capability to design proteins, responding to complex prompts combining atomic level details and high level instructions to generate new proteins. ESM3 can explore protein space far beyond natural evolution. We prompted ESM3 to generate a fluorescent protein at a far distance from any known fluorescent proteins, searching an unknown region of protein space, to discover a new fluorescent protein. We estimate this is equivalent to simulating five hundred million years of evolution.

We're thrilled to present ESM3 in Science Magazine. ESM3 is a generative language model that reasons over the three fundamental properties of proteins: sequence, structure, and function. Today we're making ESM3 available free to researchers worldwide via the public beta of an API for biological intelligence. Trained with over a trillion teraflops of compute, this is the first time a model of this scale has been trained for biology, pushing the frontier of AI for biological discovery and engineering. ESM3 learns to represent the immense complexity of protein biology, learning from billions of natural proteins. From this training it developed the capability to design proteins, responding to complex prompts combining atomic level details and high level instructions to generate new proteins. ESM3 can explore protein space far beyond natural evolution. We prompted ESM3 to generate a fluorescent protein at a far distance from any known fluorescent proteins, searching an unknown region of protein space, to discover a new fluorescent protein. We estimate this is equivalent to simulating five hundred million years of evolution.

Alex Rives

227,805 次观看 • 1 年前

One of the best papers that combine machine learning and video games. The neural network learns motion from raw human data. The motion in-game feels fluid and natural, and best part is you dont have to manually write motion state-machines.

One of the best papers that combine machine learning and video games. The neural network learns motion from raw human data. The motion in-game feels fluid and natural, and best part is you dont have to manually write motion state-machines.

mansin

198,161 次观看 • 2 年前

The new Johns Hopkins University Bloomberg Center is designed to foster collaboration between policymakers, faculty, and students and brings Johns Hopkins University' world-class research and expertise to the center of Washington. It will serve as a connecting bridge between both ends of Pennsylvania Avenue, and I can't wait to see the ideas and initiatives that emerge from the leaders who will come together in this space.

The new Johns Hopkins University Bloomberg Center is designed to foster collaboration between policymakers, faculty, and students and brings Johns Hopkins University' world-class research and expertise to the center of Washington. It will serve as a connecting bridge between both ends of Pennsylvania Avenue, and I can't wait to see the ideas and initiatives that emerge from the leaders who will come together in this space.

Mike Bloomberg

54,317 次观看 • 3 年前

Seems the general consensus that has settled with the public over the last few days is that Data Republican is a fed and is no longer to be trusted as she is a part of the Con-Inc Sphere. Many of you clocked it from the jump. 🙌🏻

Seems the general consensus that has settled with the public over the last few days is that Data Republican is a fed and is no longer to be trusted as she is a part of the Con-Inc Sphere. Many of you clocked it from the jump. 🙌🏻

Shipwreck

60,883 次观看 • 2 个月前

'The attention should be shifted from what the price of the token is today, to what is the liquidity depth/structure over time.’ Tachi explains how the Arc Forge leverages the DLMM to enable projects to generate a sustainable source of revenue and keep building over time.

'The attention should be shifted from what the price of the token is today, to what is the liquidity depth/structure over time.’ Tachi explains how the Arc Forge leverages the DLMM to enable projects to generate a sustainable source of revenue and keep building over time.

Meteora

27,978 次观看 • 1 年前

Ever wondered what neural networks are and how they work? Systems like ChatGPT use neural networks to work as well as they do. Neural networks are composed of "layers" of neurons, layers with different functions; connections between layers called "weights"; and mathematical functions called "activation functions". If you’re interested in learning about these systems, check the comments. Ultimately, the neural network structure of the model serves to visually demonstrate that it is, in fact, a complex mathematical equation. When companies release the model's weights, they are releasing a key component needed to run the model's complete equation. Without the weights, the equation is incomplete. For the math-minded: the weights of a model are the learned numbers (they are variables during training) that are then used as constants in the mathematical functions that make up the model. Neural networks are ultimately just one big, hyper-complex mathematical function, and when a model is trained, it learns the constants associated with the high-dimensional input.

Ever wondered what neural networks are and how they work? Systems like ChatGPT use neural networks to work as well as they do. Neural networks are composed of "layers" of neurons, layers with different functions; connections between layers called "weights"; and mathematical functions called "activation functions". If you’re interested in learning about these systems, check the comments. Ultimately, the neural network structure of the model serves to visually demonstrate that it is, in fact, a complex mathematical equation. When companies release the model's weights, they are releasing a key component needed to run the model's complete equation. Without the weights, the equation is incomplete. For the math-minded: the weights of a model are the learned numbers (they are variables during training) that are then used as constants in the mathematical functions that make up the model. Neural networks are ultimately just one big, hyper-complex mathematical function, and when a model is trained, it learns the constants associated with the high-dimensional input.

Harper Carroll

50,905 次观看 • 1 年前

Elon Musk: 𝕏 is crucial for democracy and freedom in the U.S. and worldwide. “Twitter is important to the function of democracy, it's important to the function of the United States as a free country and many other countries, and to actually to help freedom in the world more broadly than the US. I think civilizational risk is decreased the more we can increase the trust of Twitter as a public platform.” From: TED Interview, April 14, 2022

Elon Musk: 𝕏 is crucial for democracy and freedom in the U.S. and worldwide. “Twitter is important to the function of democracy, it's important to the function of the United States as a free country and many other countries, and to actually to help freedom in the world more broadly than the US. I think civilizational risk is decreased the more we can increase the trust of Twitter as a public platform.” From: TED Interview, April 14, 2022

ELON CLIPS

15,009 次观看 • 10 个月前

In collaboration with Intel, our Depth Fusion showcases the power of our LDM3D diffusion model in generating 360° views from text prompts provided by the user. The LDM3D diffusion model generates a 2D RGB image and its corresponding relative depth map providing a complete RGBD representation corresponding to the text prompt. The LDM 3D model is a specialized version of the stable diffusion V 1.4 model that has been modified to fit both image and depth map data.The model was then fine tuned on a subset of the Laion400M data set - large scale image caption data set. The depth maps used to fine tune our model were generated by the DPTBeiT large 512 depth estimation model that provides highly accurate relative depth estimates for each pixel. We take the generated 2D RGB image and depth map and use them to compute a 360° projection using touchdesigner. Touchdesigner is a versatile platform that allows for the creation of immersive and interactive multimedia experiences. Our application harnesses the power of touchdesigner to bring the generated 360° views to life, providing users with a unique and engaging way to experience their text prompts, whether it’s a description of a tranquil forest, a noisy cityscape or a futuristic sci fi world. Our depth fusion can bring these concepts to life in a vivid and immersive detail. - Scottie Fox, VP Engineering Blockade Labs ScottieFox #AI #VR #3D #gamedev #stablediffusion

In collaboration with Intel, our Depth Fusion showcases the power of our LDM3D diffusion model in generating 360° views from text prompts provided by the user. The LDM3D diffusion model generates a 2D RGB image and its corresponding relative depth map providing a complete RGBD representation corresponding to the text prompt. The LDM 3D model is a specialized version of the stable diffusion V 1.4 model that has been modified to fit both image and depth map data.The model was then fine tuned on a subset of the Laion400M data set - large scale image caption data set. The depth maps used to fine tune our model were generated by the DPTBeiT large 512 depth estimation model that provides highly accurate relative depth estimates for each pixel. We take the generated 2D RGB image and depth map and use them to compute a 360° projection using touchdesigner. Touchdesigner is a versatile platform that allows for the creation of immersive and interactive multimedia experiences. Our application harnesses the power of touchdesigner to bring the generated 360° views to life, providing users with a unique and engaging way to experience their text prompts, whether it’s a description of a tranquil forest, a noisy cityscape or a futuristic sci fi world. Our depth fusion can bring these concepts to life in a vivid and immersive detail. - Scottie Fox, VP Engineering Blockade Labs ScottieFox #AI #VR #3D #gamedev #stablediffusion

Blockade Labs

11,458 次观看 • 3 年前

From the floor plans to the art on its walls, every detail of the new Johns Hopkins University Bloomberg Center was chosen to foster the creativity, free open exchange of ideas, and interdisciplinary collaboration that we need in our future leaders. Take a look inside to see how:

From the floor plans to the art on its walls, every detail of the new Johns Hopkins University Bloomberg Center was chosen to foster the creativity, free open exchange of ideas, and interdisciplinary collaboration that we need in our future leaders. Take a look inside to see how:

Mike Bloomberg

33,184 次观看 • 3 年前

I just love this elaborate language and grand gestures to convey ideas, sometimes to the point of being verbose is AMaZing! This in a nutshell tells you the hierarchical structure of existence. Natural selection is no joke.

I just love this elaborate language and grand gestures to convey ideas, sometimes to the point of being verbose is AMaZing! This in a nutshell tells you the hierarchical structure of existence. Natural selection is no joke.

Dr STAN

29,141 次观看 • 1 年前

New research from the Johns Hopkins Bloomberg School of Public Health shows a strong link between frequent daytime napping and an elevated risk of developing dementia.

New research from the Johns Hopkins Bloomberg School of Public Health shows a strong link between frequent daytime napping and an elevated risk of developing dementia.

KTLA

273,643 次观看 • 1 个月前

Zohran Mamdani: "When we're talking about data centers, the environmental impacts that come with this, I think one of the things that I really have appreciated from our governor is the implementation of a moratorium on data center construction across the state of New York. That is critically important. And in a country that is so unbelievably polarized, to see data centers as one of the few things that cuts across party lines, where you see Republicans and Democrats who do not want to see these kinds of things being constructed." From our people's town hall with AI Now Institute

Zohran Mamdani: "When we're talking about data centers, the environmental impacts that come with this, I think one of the things that I really have appreciated from our governor is the implementation of a moratorium on data center construction across the state of New York. That is critically important. And in a country that is so unbelievably polarized, to see data centers as one of the few things that cuts across party lines, where you see Republicans and Democrats who do not want to see these kinds of things being constructed." From our people's town hall with AI Now Institute

More Perfect Union

12,781 次观看 • 26 天前

🧠 A Beautiful loop: An active inference theory of consciousness. Many psychological, philosophical and scientific approaches have posited that loops, recursion, and reflective broadcasting are somehow central to the emergence of consciousness in hierarchical complex apparatus such as the nervous system. The geometrodynamics of the experiential reality model itself continuously looping and conforming its own existence. The capacity to reiterate the cognitive process in a variety of states like attention, metacognition, sleep, lucidity, meditation and psychedelic is the cornerstone for the emergence of consciousness. In this scientific study are proposed three conditions necessary to the functionality of self awareness. 1st) Generative World Model Epistemic Field. This model represents the brain's internal representation of the world, which is continuously updated throughout sensory inputs and prior knowledge. 2nd) Inferential Competition Determining Conscious Content. The inferential competition is necessary for determining what becomes conscious and why it’s coherent. This competition arises from the brain's attempt to resolve uncertainty and ambiguity in its internal models. 3d) Epistemic Depth Recursive and Widespread Sharing. The epistemic depth refers to the recursive and widespread sharing of the epistemic field throughout the hierarchical structure, and it’s a fundamental component for consciousness. 🔗

🧠 A Beautiful loop: An active inference theory of consciousness. Many psychological, philosophical and scientific approaches have posited that loops, recursion, and reflective broadcasting are somehow central to the emergence of consciousness in hierarchical complex apparatus such as the nervous system. The geometrodynamics of the experiential reality model itself continuously looping and conforming its own existence. The capacity to reiterate the cognitive process in a variety of states like attention, metacognition, sleep, lucidity, meditation and psychedelic is the cornerstone for the emergence of consciousness. In this scientific study are proposed three conditions necessary to the functionality of self awareness. 1st) Generative World Model Epistemic Field. This model represents the brain's internal representation of the world, which is continuously updated throughout sensory inputs and prior knowledge. 2nd) Inferential Competition Determining Conscious Content. The inferential competition is necessary for determining what becomes conscious and why it’s coherent. This competition arises from the brain's attempt to resolve uncertainty and ambiguity in its internal models. 3d) Epistemic Depth Recursive and Widespread Sharing. The epistemic depth refers to the recursive and widespread sharing of the epistemic field throughout the hierarchical structure, and it’s a fundamental component for consciousness. 🔗

𝓜𝒂𝒖𝒓𝒊𝔃𝒊𝒐 𝗜𝒃𝛼

18,397 次观看 • 1 年前

Elon Musk on why 𝕏 is crucial for democracy and freedom: “It’s important to the function of democracy. It’s important to the function of the United States as a free country and many other countries, and to help freedom in the world more broadly than the US I think that civilizational risk is decreased the more we can increase the trust of 𝕏 as a public platform”

Elon Musk on why 𝕏 is crucial for democracy and freedom: “It’s important to the function of democracy. It’s important to the function of the United States as a free country and many other countries, and to help freedom in the world more broadly than the US I think that civilizational risk is decreased the more we can increase the trust of 𝕏 as a public platform”

X Freeze

19,560 次观看 • 7 个月前

Can you and I overcome the propensities of the natural man or woman through willpower and goal setting and determination and discipline? No. We learn in the Book of Mormon that the Savior is at the center of our transformation: “For the natural man is an enemy to God, and has been from the fall of Adam, and will be, forever and ever, unless he yields to the enticings of the Holy Spirit, and putteth off the natural man and becometh a saint THROUGH the Atonement of Christ the Lord” (Mosiah 3:19).

Can you and I overcome the propensities of the natural man or woman through willpower and goal setting and determination and discipline? No. We learn in the Book of Mormon that the Savior is at the center of our transformation: “For the natural man is an enemy to God, and has been from the fall of Adam, and will be, forever and ever, unless he yields to the enticings of the Holy Spirit, and putteth off the natural man and becometh a saint THROUGH the Atonement of Christ the Lord” (Mosiah 3:19).

David A. Bednar

55,543 次观看 • 6 天前

The Brazilian Amazon is still burning at an alarming rate. We must switch from an economic model that is based on the destruction of the forest and the exploitation of natural resources to a system that values the forest standing.

The Brazilian Amazon is still burning at an alarming rate. We must switch from an economic model that is based on the destruction of the forest and the exploitation of natural resources to a system that values the forest standing.

Greenpeace International

394,974 次观看 • 3 年前