正在加载视频...

视频加载失败

返回首页

🤩Apple opensources MGIE! Now one can take random pictures w. iPhone & edit w. language! Guiding Instruction-based Image Editing via Multimodal Large Language Models #ICLR2024 spotlight: Apple repo Gradio

WilliamWangNLP's profile picture

William Wang

21,429 subscribers

124,999 次观看 • 2 年前 •via X (Twitter)

科学技术新闻政治教育#ICLR2024
LIVE

Anya Rossi• Live Now

Private livecam show

5 条评论

Ali Madad 的头像
Ali Madad2 年前

@ivanfioravanti 👆🏾

Keshav Jindal 的头像
Keshav Jindal2 年前

noob question: I couldn't figure out if commercial use is allowed. is it?

Yuriy Yuzifovich 的头像
Yuriy Yuzifovich2 年前

Seeing the success of @AIatMeta with open source AI must have been a contributing factor in Apple decision making. Great!

hristo 的头像
hristo2 年前

Only Apple Can do !

johns code 的头像
johns code2 年前

really cool. unfortunately it does not build on an M2 Mac

相关视频

TurboEdit Instant text-based image editing discuss: We address the challenges of precise image inversion and disentangled image editing in the context of few-step diffusion models. We introduce an encoder based iterative inversion technique. The inversion network is conditioned on the input image and the reconstructed image from the previous step, allowing for correction of the next reconstruction towards the input image. We demonstrate that disentangled controls can be easily achieved in the few-step diffusion model by conditioning on an (automatically generated) detailed text prompt. To manipulate the inverted image, we freeze the noise maps and modify one attribute in the text prompt (either manually or via instruction based editing driven by an LLM), resulting in the generation of a new image similar to the input image with only one attribute changed. It can further control the editing strength and accept instructive text prompt. Our approach facilitates realistic text-guided image edits in real-time, requiring only 8 number of functional evaluations (NFEs) in inversion (one-time cost) and 4 NFEs per edit. Our method is not only fast, but also significantly outperforms state-of-the-art multi-step diffusion editing techniques.
0:22

TurboEdit Instant text-based image editing discuss: We address the challenges of precise image inversion and disentangled image editing in the context of few-step diffusion models. We introduce an encoder based iterative inversion technique. The inversion network is conditioned on the input image and the reconstructed image from the previous step, allowing for correction of the next reconstruction towards the input image. We demonstrate that disentangled controls can be easily achieved in the few-step diffusion model by conditioning on an (automatically generated) detailed text prompt. To manipulate the inverted image, we freeze the noise maps and modify one attribute in the text prompt (either manually or via instruction based editing driven by an LLM), resulting in the generation of a new image similar to the input image with only one attribute changed. It can further control the editing strength and accept instructive text prompt. Our approach facilitates realistic text-guided image edits in real-time, requiring only 8 number of functional evaluations (NFEs) in inversion (one-time cost) and 4 NFEs per edit. Our method is not only fast, but also significantly outperforms state-of-the-art multi-step diffusion editing techniques.

AK

16,062 次观看 • 1 年前

We’re excited to announce the release and open-source of HunyuanImage 3.0 — the largest and most powerful open-source text-to-image model to date, with over 80 billion total parameters, of which 13 billion are activated per token during inference.The effect is completely comparable to the industry’s flagship closed-source model.🚀🚀🚀 HunyuanImage 3.0 originates from our internally developed native multimodal large language model, with fine-tuning and post-training focused on text-to-image generation. This unique foundation gives the model a powerful set of capabilities: ✅Reason with world knowledge ✅Understand complex, thousand-word prompts ✅Generate precise text within images Different from traditional DiT architecture image generation models, HunyuanImage 3.0’s MoE architecture uses a Transfusion-based approach to deeply couple Diffusion and LLM training for a single, powerful system. Built on Hunyuan-A13B, HunyuanImage 3.0 was trained on a massive dataset: 5 billion image-text pairs, video frames, interleaved image-text data, and 6 trillion tokens of text corpora. This hybrid training across multimodal generation, understanding, and LLM capabilities allows the model to seamlessly integrate multiple tasks. Whether you're an illustrator, designer, or creator, this is built to slash your workflow from hours to minutes. HunyuanImage 3.0 can generate intricate text, detailed comics, expressive emojis, and lively, engaging illustrations for educational content. The current release focuses solely on text-to-image generation and future updates will include image-to-image, image editing, multi-turn interaction, and more. 👉🏻Try it now: 🔗GitHub: 🤗Hugging Face:
2:09

We’re excited to announce the release and open-source of HunyuanImage 3.0 — the largest and most powerful open-source text-to-image model to date, with over 80 billion total parameters, of which 13 billion are activated per token during inference.The effect is completely comparable to the industry’s flagship closed-source model.🚀🚀🚀 HunyuanImage 3.0 originates from our internally developed native multimodal large language model, with fine-tuning and post-training focused on text-to-image generation. This unique foundation gives the model a powerful set of capabilities: ✅Reason with world knowledge ✅Understand complex, thousand-word prompts ✅Generate precise text within images Different from traditional DiT architecture image generation models, HunyuanImage 3.0’s MoE architecture uses a Transfusion-based approach to deeply couple Diffusion and LLM training for a single, powerful system. Built on Hunyuan-A13B, HunyuanImage 3.0 was trained on a massive dataset: 5 billion image-text pairs, video frames, interleaved image-text data, and 6 trillion tokens of text corpora. This hybrid training across multimodal generation, understanding, and LLM capabilities allows the model to seamlessly integrate multiple tasks. Whether you're an illustrator, designer, or creator, this is built to slash your workflow from hours to minutes. HunyuanImage 3.0 can generate intricate text, detailed comics, expressive emojis, and lively, engaging illustrations for educational content. The current release focuses solely on text-to-image generation and future updates will include image-to-image, image editing, multi-turn interaction, and more. 👉🏻Try it now: 🔗GitHub: 🤗Hugging Face:

Tencent Hy

412,658 次观看 • 9 个月前

Dr. Fei-Fei Li (Fei-Fei Li) is known as the “godmother of AI.” For the past two decades, she’s been at the center of AI’s most significant breakthroughs, including: - Spearheading ImageNet, the dataset that sparked the AI explosion we’re living through right now. - Leading work at Stanford Artificial Intelligence Laboratory (SAIL) - Serving as Chief Scientist of AI/ML at Google Cloud - Co-founding Stanford’s Institute for Human-Centered AI - Serving on the United Nations AI Scientific Advisory Board - Being named as Time's 100 most influential people in AI In this conversation, Fei-Fei shares the rarely told history of how we got to today—and what comes next. We discuss: 🔸 The backstory on ImageNet 🔸 Why robotics faces unique challenges compared with language models and what’s needed to overcome them 🔸 Why Fei-Fei believes AI won’t replace humans but will require us to take responsibility for ourselves 🔸 Why world models and spatial intelligence represent the next frontier in AI, beyond large language models 🔸 The surprising applications of Marble, from movie production to psychological research 🔸 How to participate in AI regardless of your role 🔸 Much more Listen now 👇 • YouTube: • Spotify: • Apple: Thank you to our wonderful sponsors for supporting the podcast: 🏆 Figma Make — A prompt-to-code tool for making ideas real: 🏆 Justworks — The all-in-one HR solution for managing your small business with confidence: 🏆 Sinch — Build messaging, email, and calling into your product:
1:40

Dr. Fei-Fei Li (Fei-Fei Li) is known as the “godmother of AI.” For the past two decades, she’s been at the center of AI’s most significant breakthroughs, including: - Spearheading ImageNet, the dataset that sparked the AI explosion we’re living through right now. - Leading work at Stanford Artificial Intelligence Laboratory (SAIL) - Serving as Chief Scientist of AI/ML at Google Cloud - Co-founding Stanford’s Institute for Human-Centered AI - Serving on the United Nations AI Scientific Advisory Board - Being named as Time's 100 most influential people in AI In this conversation, Fei-Fei shares the rarely told history of how we got to today—and what comes next. We discuss: 🔸 The backstory on ImageNet 🔸 Why robotics faces unique challenges compared with language models and what’s needed to overcome them 🔸 Why Fei-Fei believes AI won’t replace humans but will require us to take responsibility for ourselves 🔸 Why world models and spatial intelligence represent the next frontier in AI, beyond large language models 🔸 The surprising applications of Marble, from movie production to psychological research 🔸 How to participate in AI regardless of your role 🔸 Much more Listen now 👇 • YouTube: • Spotify: • Apple: Thank you to our wonderful sponsors for supporting the podcast: 🏆 Figma Make — A prompt-to-code tool for making ideas real: 🏆 Justworks — The all-in-one HR solution for managing your small business with confidence: 🏆 Sinch — Build messaging, email, and calling into your product:

Lenny Rachitsky

250,455 次观看 • 8 个月前

Check out mistral.​rs, our #Rust-based open source inference engine allowing for fast #LLM serving for a variety of architectures including X-LoRA mixture-of-expert (MoE) models, Llama-3, Mistral/Mixtral, Gemma & many others. Built on the Hugging Face #Candle framework for #Rust w/ custom CUDA kernels in the backend (as well as support for Metal, Apple Accelerate, and Intel MKL for CPU use), you can easily create a REST API OpenAI compatible server or run via Python bindings. Key features include: ✅Prefix caching, continuous batching ✅Flash Attention V2 ✅Device offloading ✅GGUF or Hugging Face models ✅2, 3, 4, 5, 6 and 8 bit quantization ✅X-LoRA MoE non-granular scalings for fast inference ✅Grammar support ✅Continuous batching ✅LoRA support with weight merging ✅LlamaIndex 🦙 integration ...and much more. Incorporation into our GraphReasoning multi-agent modeling framework & LlamaIndex 🦙 allows you to combine in-context learning with adversarial agentic strategies, to dive deep into complex scientific analyses, such as to predict material behaviors, generate hypotheses, analyze papers and data, develop new research concepts, and much more. Check out mistral.​rs: Join our Discord here: Rust Trending Rust Language
0:35

Check out mistral.​rs, our #Rust-based open source inference engine allowing for fast #LLM serving for a variety of architectures including X-LoRA mixture-of-expert (MoE) models, Llama-3, Mistral/Mixtral, Gemma & many others. Built on the Hugging Face #Candle framework for #Rust w/ custom CUDA kernels in the backend (as well as support for Metal, Apple Accelerate, and Intel MKL for CPU use), you can easily create a REST API OpenAI compatible server or run via Python bindings. Key features include: ✅Prefix caching, continuous batching ✅Flash Attention V2 ✅Device offloading ✅GGUF or Hugging Face models ✅2, 3, 4, 5, 6 and 8 bit quantization ✅X-LoRA MoE non-granular scalings for fast inference ✅Grammar support ✅Continuous batching ✅LoRA support with weight merging ✅LlamaIndex 🦙 integration ...and much more. Incorporation into our GraphReasoning multi-agent modeling framework & LlamaIndex 🦙 allows you to combine in-context learning with adversarial agentic strategies, to dive deep into complex scientific analyses, such as to predict material behaviors, generate hypotheses, analyze papers and data, develop new research concepts, and much more. Check out mistral.​rs: Join our Discord here: Rust Trending Rust Language

Markus J. Buehler

73,575 次观看 • 2 年前

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.
2:00

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

27,005 次观看 • 2 年前

Thrilled to announce Kingnet AI V2 is now officially live ! We have officially deployed on the BNB Chain first ! Whether you're an enthusiast or a professional game developer, come and try it out now： Each generated asset costs approximately $3 and supports export in professional game-editing formats. We will soon support exporting assets in NFT on-chain formats, empowering Web3 users and partners with seamless integration. Jump down more rabbit holes next.👇 📔 Product Introduction： By conversing naturally with agent Joi, users can achieve a complete automated game development cycle - from requirement proposal to finished product delivery. Users simply need to describe their game concepts and design requirements in natural language, and Joi will automatically utilize built-in generator including: • Animation Generator： AI-driven motion generation with auto-rigging technology for instant character animation • Map Generator： Procedural map generation with built-in logic validation for consistent world-building • Numerical Generator： Automated game economy tuning for fair yet challenging gameplay systems • Editable Code Generator： Generates clean, maintainable game logic code with multi-platform/multi-language support • Interface Generator： Intelligent layout engine that optimizes user experience and interaction flow Joi intelligently generates all necessary game components, performs multi-dimensional feasibility checks, and ultimately completes game synthesis, packaging and deployment. Users can directly click to try the game on the chat interface, or download the complete editable code package to achieve rapid iteration and secondary development. 🎯 Core Architecture： 1/ Natural Language Understanding & Multimodal Intent Parsing: Utilizing advanced deep learning NLP models (e.g., Transformer-based language understanding models), Joi precisely interprets user natural language inputs and extracts core game design intents and parameters. Through semantic segmentation and entity recognition, complex requirements are decomposed into specific tasks for animation, map, numerical systems, UI, and code modules. 2/ Modular Editor System & API Integration: Joi employs a unified API framework to enable seamless collaboration between editor modules, ensuring high compatibility in data formats and workflows. 3/ Intelligent Validation & Quality Assurance: The system incorporates multi-dimensional verification mechanisms including animation continuity checks, map pathfinding and physical logic validation, game balance analysis, UI interaction consistency verification, and static/dynamic code security testing. Automated testing and feedback loops ensure outputs meet high-standard game design specifications. 4/ Automatic Synthesis, Packaging & Instant Deployment: Verified resources are automatically integrated to complete game compilation, packaging and deployment. Supports one-click generation of playable online links and downloadable complete code packages for immediate testing or deep customization/iterative development. 5/ Interactive Chat Interface & Seamless UX: The entire workflow is completed within the chat interface, significantly reducing traditional game development's communication and operational barriers. Users accomplish complex game design and development through conversation while receiving real-time feedback and adjustment suggestions, democratizing game creation. 6/ Industry-Disrupting Value: Transforms traditional manual development into AI-driven automated pipelines.
1:24

Thrilled to announce Kingnet AI V2 is now officially live ! We have officially deployed on the BNB Chain first ! Whether you're an enthusiast or a professional game developer, come and try it out now： Each generated asset costs approximately $3 and supports export in professional game-editing formats. We will soon support exporting assets in NFT on-chain formats, empowering Web3 users and partners with seamless integration. Jump down more rabbit holes next.👇 📔 Product Introduction： By conversing naturally with agent Joi, users can achieve a complete automated game development cycle - from requirement proposal to finished product delivery. Users simply need to describe their game concepts and design requirements in natural language, and Joi will automatically utilize built-in generator including: • Animation Generator： AI-driven motion generation with auto-rigging technology for instant character animation • Map Generator： Procedural map generation with built-in logic validation for consistent world-building • Numerical Generator： Automated game economy tuning for fair yet challenging gameplay systems • Editable Code Generator： Generates clean, maintainable game logic code with multi-platform/multi-language support • Interface Generator： Intelligent layout engine that optimizes user experience and interaction flow Joi intelligently generates all necessary game components, performs multi-dimensional feasibility checks, and ultimately completes game synthesis, packaging and deployment. Users can directly click to try the game on the chat interface, or download the complete editable code package to achieve rapid iteration and secondary development. 🎯 Core Architecture： 1/ Natural Language Understanding & Multimodal Intent Parsing: Utilizing advanced deep learning NLP models (e.g., Transformer-based language understanding models), Joi precisely interprets user natural language inputs and extracts core game design intents and parameters. Through semantic segmentation and entity recognition, complex requirements are decomposed into specific tasks for animation, map, numerical systems, UI, and code modules. 2/ Modular Editor System & API Integration: Joi employs a unified API framework to enable seamless collaboration between editor modules, ensuring high compatibility in data formats and workflows. 3/ Intelligent Validation & Quality Assurance: The system incorporates multi-dimensional verification mechanisms including animation continuity checks, map pathfinding and physical logic validation, game balance analysis, UI interaction consistency verification, and static/dynamic code security testing. Automated testing and feedback loops ensure outputs meet high-standard game design specifications. 4/ Automatic Synthesis, Packaging & Instant Deployment: Verified resources are automatically integrated to complete game compilation, packaging and deployment. Supports one-click generation of playable online links and downloadable complete code packages for immediate testing or deep customization/iterative development. 5/ Interactive Chat Interface & Seamless UX: The entire workflow is completed within the chat interface, significantly reducing traditional game development's communication and operational barriers. Users accomplish complex game design and development through conversation while receiving real-time feedback and adjustment suggestions, democratizing game creation. 6/ Industry-Disrupting Value: Transforms traditional manual development into AI-driven automated pipelines.

Kingnet AI

45,966 次观看 • 1 年前

is our AI project to make computing feel more human L A N D E R Here are the 4 best demo videos of the magic of DATA in action. DATA is a personalized assistant who knows and remembers every conversation you have with it accross your iPhone, Mac, iPad, Watch, Texts, Emails, and HomePods. You can talk to DATA right in your AirPods or text it just like a person. DATA can read, write, understand, speak any language, and translate between them. It can help with real work and home life tasks like research, writing, scheduling, reminders, and triage. And it's easily customizable so you can have DATA automatically do whatever you want whenever you want with just a few taps and natural language instructions - no code required. DATA can do just about anything you can do on your phone on your behalf automatically including very advanced things Siri can't, like summarizing, analyzing, and drafting replies or writing documents. It can read web pages, texts or emails you show it, or PDFs of any kind. It can do other real world tasks that require complex analysis and common sense too, like: - figure out where the nearest beach is (even when you're in Colorado) and instantly fetch the current surf report up to the current minute. - summarize and drafting replies to entire email chains - plan out entire work projects or multi-day vacations on your calendar - sketch out ideas for you in picture form or drafting Notion pages with charts and graphs. DATA can also use its own judgement to determine when to run an action or not, even if you've scheduled it, allowing you to make VERY complex automations that require many different inputs to make a decision, like for example: - only opening the blinds on your lunch break if it's sunny out and you're working from home. DATA works natively and easily with Apple HomeKit & other shortcuts. DATA can also take initiative and check in with you throughout the day by voice or text and proactively send messages to you and others on your behalf based on your personal and professional goals, current tasks, and calendar. DATA can integrate with many apps on your phone, and is compatible with multiple large AI language models. I've gotten to make a few demo videos that I think really capture how powerful DATA can be for every day life. Here they are all in one tweet. Make sure your sound is on as you watch them. 1. This is the first demo video I ever made from April 19th, 2023. It walks through all the ways you can interact with and use the DATA shortcuts. Everything from saying "Hey Siri" to tapping on custom apps on your home-screen. 2. The second demo video was made May 5 and is an example use case I made of how commands work - commands allow DATA to actually run actions on your phone like taking pictures and sending messages. This demo shows me taking a picture of an email template, and data drafting an email based on that template. It's gotten much better at realizing when it has just run a command and incorporating that information naturally into the conversation now, especially on GPT-4. 3. This third Commands video, May 12 is a walkthrough of ALL the phone functions that commands allow DATA to do: sending texts and emails, making pictures, seeing pictures, reading things, and scheduling events. Since this video we've added auto-replies to texts and emails, summarizing documents, writing documents, health app data retrieval, web surfing, scheduling alarms, making playlists, and more. 4. This last demo I made today, June 15, shows everything DATA does working in concert to generate a crazy detailed morning briefing with background music - including making a unique playlist and giving a detailed analysis of current events complete with Ski & Surf conditions near me other live information from the internet. So now that you've seen everything DATA can do, what's the coolest feature? What features should we add? What would you use DATA for first?
12:54

is our AI project to make computing feel more human L A N D E R Here are the 4 best demo videos of the magic of DATA in action. DATA is a personalized assistant who knows and remembers every conversation you have with it accross your iPhone, Mac, iPad, Watch, Texts, Emails, and HomePods. You can talk to DATA right in your AirPods or text it just like a person. DATA can read, write, understand, speak any language, and translate between them. It can help with real work and home life tasks like research, writing, scheduling, reminders, and triage. And it's easily customizable so you can have DATA automatically do whatever you want whenever you want with just a few taps and natural language instructions - no code required. DATA can do just about anything you can do on your phone on your behalf automatically including very advanced things Siri can't, like summarizing, analyzing, and drafting replies or writing documents. It can read web pages, texts or emails you show it, or PDFs of any kind. It can do other real world tasks that require complex analysis and common sense too, like: - figure out where the nearest beach is (even when you're in Colorado) and instantly fetch the current surf report up to the current minute. - summarize and drafting replies to entire email chains - plan out entire work projects or multi-day vacations on your calendar - sketch out ideas for you in picture form or drafting Notion pages with charts and graphs. DATA can also use its own judgement to determine when to run an action or not, even if you've scheduled it, allowing you to make VERY complex automations that require many different inputs to make a decision, like for example: - only opening the blinds on your lunch break if it's sunny out and you're working from home. DATA works natively and easily with Apple HomeKit & other shortcuts. DATA can also take initiative and check in with you throughout the day by voice or text and proactively send messages to you and others on your behalf based on your personal and professional goals, current tasks, and calendar. DATA can integrate with many apps on your phone, and is compatible with multiple large AI language models. I've gotten to make a few demo videos that I think really capture how powerful DATA can be for every day life. Here they are all in one tweet. Make sure your sound is on as you watch them. 1. This is the first demo video I ever made from April 19th, 2023. It walks through all the ways you can interact with and use the DATA shortcuts. Everything from saying "Hey Siri" to tapping on custom apps on your home-screen. 2. The second demo video was made May 5 and is an example use case I made of how commands work - commands allow DATA to actually run actions on your phone like taking pictures and sending messages. This demo shows me taking a picture of an email template, and data drafting an email based on that template. It's gotten much better at realizing when it has just run a command and incorporating that information naturally into the conversation now, especially on GPT-4. 3. This third Commands video, May 12 is a walkthrough of ALL the phone functions that commands allow DATA to do: sending texts and emails, making pictures, seeing pictures, reading things, and scheduling events. Since this video we've added auto-replies to texts and emails, summarizing documents, writing documents, health app data retrieval, web surfing, scheduling alarms, making playlists, and more. 4. This last demo I made today, June 15, shows everything DATA does working in concert to generate a crazy detailed morning briefing with background music - including making a unique playlist and giving a detailed analysis of current events complete with Ski & Surf conditions near me other live information from the internet. So now that you've seen everything DATA can do, what's the coolest feature? What features should we add? What would you use DATA for first?

steve

640,114 次观看 • 3 年前

🔥😡 “ICE IS A DOMESTIC TERROR ORGANIZATION” JENNIFER WELCH IS TRYING TO GET GREG BOVINO & ICE AGENTS HARMED. They are using this rhetoric to cause harm and say “BOVINO IS ON THE SIDE OF A REGIME THAT EXTERMINATED 6 MILLION JEWISH PEOPLE” …because of his Jacket she then calls him a Nazi. She is a Democrat Soros owned concoction in league with Obama who she knows as well. She is intended to brainwash MIDDLE AGE MOMS with shocking language. She is a ENEMY of the Republic who now lives in Manhattan from Oklahoma so she can work easier with the Bolshevik puppet masters who pay her. One of the common ways money is funneled to these people ARE FROM DONATIONS ON YOUTUBE AND SUBSCRIBERS which you can set tier levels at any amount. She needs to be investigated. Commander Op At Large CA Gregory K. Bovino already is dealing with threats from cartel members via a bounty in Chicago & we Latin Kings were involved in the left of the FBI vehicle the other night. That man is dealing with enough out there like a hero to make our nation safer and this RHETORIC IS UNACCEPTABLE. She is a PAID PROPAGANDA AGENT. Here is a photo of famous “NAZI” Supreme Allied Commander General Eisenhower w/ General Patton who DEFEATED THE NAZI’S YOU LIZARD LOOKING DIZZY B*TCH.
2:41

🔥😡 “ICE IS A DOMESTIC TERROR ORGANIZATION” JENNIFER WELCH IS TRYING TO GET GREG BOVINO & ICE AGENTS HARMED. They are using this rhetoric to cause harm and say “BOVINO IS ON THE SIDE OF A REGIME THAT EXTERMINATED 6 MILLION JEWISH PEOPLE” …because of his Jacket she then calls him a Nazi. She is a Democrat Soros owned concoction in league with Obama who she knows as well. She is intended to brainwash MIDDLE AGE MOMS with shocking language. She is a ENEMY of the Republic who now lives in Manhattan from Oklahoma so she can work easier with the Bolshevik puppet masters who pay her. One of the common ways money is funneled to these people ARE FROM DONATIONS ON YOUTUBE AND SUBSCRIBERS which you can set tier levels at any amount. She needs to be investigated. Commander Op At Large CA Gregory K. Bovino already is dealing with threats from cartel members via a bounty in Chicago & we Latin Kings were involved in the left of the FBI vehicle the other night. That man is dealing with enough out there like a hero to make our nation safer and this RHETORIC IS UNACCEPTABLE. She is a PAID PROPAGANDA AGENT. Here is a photo of famous “NAZI” Supreme Allied Commander General Eisenhower w/ General Patton who DEFEATED THE NAZI’S YOU LIZARD LOOKING DIZZY B*TCH.

Johnny St.Pete

66,904 次观看 • 5 个月前

The Wikipedia wars As Wikipedia approaches its 25th anniversary in 2026, its open editing model faces a growing challenge: coordinated edit wars. In these campaigns, Kremlin-aligned actors try to rewrite history, launder disinformation, and lock distorted narratives into one of the world’s most trusted reference platforms. Founded on the idea that volunteers could collaboratively build a neutral, reliable encyclopaedia, Wikipedia has become one of the most influential information platforms ever created. It is often described as the world’s largest crowd-sourced knowledge project, built on consensus and verifiable sources. In recent years, however, it has also become a frontline in geopolitical information warfare. This is most visible in so-called edit wars: prolonged conflicts where opposing groups repeatedly overwrite and revise articles to control historical narratives. Since Russia’s full-scale invasion of Ukraine in February 2022, these battles have intensified. Kremlin-aligned actors have systematically targeted articles related to Eastern Europe, the Soviet past, and contemporary political leaders. Estonia and especially EU leader Kaja Kallas, Estonia’s former prime minister, have been frequent targets. What are edit wars? An edit war happens when editors repeatedly change the same content instead of resolving disputes through discussion. Wikipedia officially discourages this behaviour and emphasises consensus, neutrality, and reliable sources. In practice, however, edit wars can and do break out. Coordinated editors can use endurance, procedural rules, and administrator complaints to exhaust good faith contributors. The goal is rarely to win a single argument. Instead, it is to wear down opposition, freeze pages at favourable moments, and normalise contested language. Once a page is locked or protected, the version in place gains a sense of legitimacy, even if it reflects a distorted view. Edit wars exploit open systems, operate over long periods, and aim to embed manipulated narratives into reference material rather than spreading short-lived falsehoods. Multiple investigations show that Wikipedia manipulation increased sharply after Russia’s invasion of Ukraine. Russian-language Wikipedia and parts of the English version became arenas for systematic narrative control, especially as independent Russian media was shut down. Wikipedia’s openness, once a strength, had suddenly become a vulnerability. Coordinated editor networks have worked to soften descriptions of Russian aggression, reframe invasions as ‘conflicts’, and question the legitimacy of post-Soviet states. These efforts rely on subtle wording changes, selective sourcing, and procedural tactics rather than obvious vandalism. Estonia and EU officials as targets Estonia shows how edit wars are used for historical revisionism and political influence. Since 2022, English-language Wikipedia articles about Estonia’s history, statehood, and politics have faced sustained pressure. One recurring tactic has been changing the birthplaces of hundreds of Estonian public figures from ‘Estonia’ to ‘Estonian SSR, Soviet Union’, despite the legal consensus that Estonia was occupied, not legitimately incorporated, by the USSR between 1940 and 1991. This is not a minor wording issue. Calling Estonia a ‘Soviet republic’ supports the Russian claim that the Baltic states voluntarily joined the USSR and directly contradicts the position of Estonia, the EU, NATO countries, and international law. Historical topics have also been targeted. The Estonian War of Independence between 1918 and 1920 has at times been reframed as an ‘offensive campaign’ or ‘separatism from Russia’, language that closely mirrors contemporary Kremlin rhetoric. High-profile figures are especially vulnerable because their pages attract constant attention and frequent administrative action. The Wikipedia article on Kaja Kallas has repeatedly been edited to reflect Russian-aligned interpretations of history and geopolitics. At key moments, the page was locked while these contested narratives were in place, blocking corrective edits. Page protection, meant to prevent disruption, instead helped freeze a favourable version of the article. This shows how procedural tools can be exploited as effectively as false information. Why Wikipedia matters Wikipedia is not just another website. It ranks highly in search results and serves as a default reference for journalists, students, policymakers, and the public. Winning an edit war on Wikipedia helps turn contested narratives into global ‘common knowledge’. For Kremlin-aligned actors, this makes Wikipedia a valuable target. Making small wording changes, downplaying occupation, reframing wars, and questioning democratic legitimacy can slowly erode our understanding of history and present-day aggression. Estonia’s experience shows how smaller states are especially exposed. Because Wikipedia is also a core source for AI systems, the stakes are even higher. Recent studies indicate that Wikipedia is one of the most cited sources for ChatGPT, effectively serving as a foundational knowledge base for how the AI understands and retrieves information. Manipulating articles today can therefore shape how future technologies understand, reproduce, and repeat history. This practice is referred to as LLM grooming, the deliberate attempt to influence large language models by seeding biased or distorted narratives into the sources they rely on. The rise in Wikipedia edit wars since 2022 reflects a broader shift in information warfare. Instead of loud propaganda, actors now use procedural, platform-native manipulation. Estonian history and Kaja Kallas are not isolated cases but targets of coordinated action. And as long as open-knowledge platforms shape how societies understand history and politics, sites like Wikipedia will remain contested ground.
0:32

The Wikipedia wars As Wikipedia approaches its 25th anniversary in 2026, its open editing model faces a growing challenge: coordinated edit wars. In these campaigns, Kremlin-aligned actors try to rewrite history, launder disinformation, and lock distorted narratives into one of the world’s most trusted reference platforms. Founded on the idea that volunteers could collaboratively build a neutral, reliable encyclopaedia, Wikipedia has become one of the most influential information platforms ever created. It is often described as the world’s largest crowd-sourced knowledge project, built on consensus and verifiable sources. In recent years, however, it has also become a frontline in geopolitical information warfare. This is most visible in so-called edit wars: prolonged conflicts where opposing groups repeatedly overwrite and revise articles to control historical narratives. Since Russia’s full-scale invasion of Ukraine in February 2022, these battles have intensified. Kremlin-aligned actors have systematically targeted articles related to Eastern Europe, the Soviet past, and contemporary political leaders. Estonia and especially EU leader Kaja Kallas, Estonia’s former prime minister, have been frequent targets. What are edit wars? An edit war happens when editors repeatedly change the same content instead of resolving disputes through discussion. Wikipedia officially discourages this behaviour and emphasises consensus, neutrality, and reliable sources. In practice, however, edit wars can and do break out. Coordinated editors can use endurance, procedural rules, and administrator complaints to exhaust good faith contributors. The goal is rarely to win a single argument. Instead, it is to wear down opposition, freeze pages at favourable moments, and normalise contested language. Once a page is locked or protected, the version in place gains a sense of legitimacy, even if it reflects a distorted view. Edit wars exploit open systems, operate over long periods, and aim to embed manipulated narratives into reference material rather than spreading short-lived falsehoods. Multiple investigations show that Wikipedia manipulation increased sharply after Russia’s invasion of Ukraine. Russian-language Wikipedia and parts of the English version became arenas for systematic narrative control, especially as independent Russian media was shut down. Wikipedia’s openness, once a strength, had suddenly become a vulnerability. Coordinated editor networks have worked to soften descriptions of Russian aggression, reframe invasions as ‘conflicts’, and question the legitimacy of post-Soviet states. These efforts rely on subtle wording changes, selective sourcing, and procedural tactics rather than obvious vandalism. Estonia and EU officials as targets Estonia shows how edit wars are used for historical revisionism and political influence. Since 2022, English-language Wikipedia articles about Estonia’s history, statehood, and politics have faced sustained pressure. One recurring tactic has been changing the birthplaces of hundreds of Estonian public figures from ‘Estonia’ to ‘Estonian SSR, Soviet Union’, despite the legal consensus that Estonia was occupied, not legitimately incorporated, by the USSR between 1940 and 1991. This is not a minor wording issue. Calling Estonia a ‘Soviet republic’ supports the Russian claim that the Baltic states voluntarily joined the USSR and directly contradicts the position of Estonia, the EU, NATO countries, and international law. Historical topics have also been targeted. The Estonian War of Independence between 1918 and 1920 has at times been reframed as an ‘offensive campaign’ or ‘separatism from Russia’, language that closely mirrors contemporary Kremlin rhetoric. High-profile figures are especially vulnerable because their pages attract constant attention and frequent administrative action. The Wikipedia article on Kaja Kallas has repeatedly been edited to reflect Russian-aligned interpretations of history and geopolitics. At key moments, the page was locked while these contested narratives were in place, blocking corrective edits. Page protection, meant to prevent disruption, instead helped freeze a favourable version of the article. This shows how procedural tools can be exploited as effectively as false information. Why Wikipedia matters Wikipedia is not just another website. It ranks highly in search results and serves as a default reference for journalists, students, policymakers, and the public. Winning an edit war on Wikipedia helps turn contested narratives into global ‘common knowledge’. For Kremlin-aligned actors, this makes Wikipedia a valuable target. Making small wording changes, downplaying occupation, reframing wars, and questioning democratic legitimacy can slowly erode our understanding of history and present-day aggression. Estonia’s experience shows how smaller states are especially exposed. Because Wikipedia is also a core source for AI systems, the stakes are even higher. Recent studies indicate that Wikipedia is one of the most cited sources for ChatGPT, effectively serving as a foundational knowledge base for how the AI understands and retrieves information. Manipulating articles today can therefore shape how future technologies understand, reproduce, and repeat history. This practice is referred to as LLM grooming, the deliberate attempt to influence large language models by seeding biased or distorted narratives into the sources they rely on. The rise in Wikipedia edit wars since 2022 reflects a broader shift in information warfare. Instead of loud propaganda, actors now use procedural, platform-native manipulation. Estonian history and Kaja Kallas are not isolated cases but targets of coordinated action. And as long as open-knowledge platforms shape how societies understand history and politics, sites like Wikipedia will remain contested ground.

EUvsDisinfo

343,369 次观看 • 5 个月前

I've had many engineers ask me why its worth their time and effort to learn biology in response to this post. Why should they be excited? We are poised for a revolution in biotech that will be uniquely enabled by computers. Convince yourself by digging into the examples I link: - The tooling is getting better. Assays are able to measure a broad array of molecules at a falling cost and increasing throughput. Look to ScaleBio, Curio Bioscience, AtlasXOmics for inspiration. We are sequencing millions of single cells and building spatial maps of the molecular state of tumors. After two generations of "next-generation sequencing", and stagnating DNA read + write costs under the monopoly of Illumina, this wave of the new assays will have a profound impact on the iteration speed and scale of experimentation. Little needs to be said about the impact of compounding trends in core tooling over a sufficient period of time. - Biotechs are using data to guide decisions and are incorporating domain informed machine learning as a core part of the molecular design process. There is great synergy here with better tooling as a means of abundant, cheap data. Look to Recursion, Manifold Bio, Dyno Tx, Asimov. There is also the cross pollination of biology informed architectures with the recent explosion in new machine learning techniques. These models are starting to do useful things, like generate functional gene editing proteins and entire prokaryotic organisms. Look to ESM, AlphaFold3, RFdiffusion. - New classes of therapies - genetic medicines and engineered immune cells - are having real success in the clinic. One dose cures for cardiovascular disease (Verve w ACSD), vaccines for cancer (Moderna w mRNA 4157), in vivo gene editing proteins (CRISPR Tx, Beam, Ensoma), metabolic disease (Novartis, Eli Lily w GLP-1/GIP modulators) are being dosed in real people right now + transforming lives. - The AI craze is commoditizing accelerated hardware and fast storage devices like NVMes, improving developer frameworks for writing code against these devices and maturing the systems tooling for moving around lots of data between computers for distributed training. One happy accident of this bubble will be the reuse of these components to build a new systems stack for the large scale processing of molecular data. This will be very important to construct a 1 billion single cell atlas and beyond. (For reference, the state of the art is ScaleBio's 2M cell kit, dubbed QuantumScale, and it is pushing things with the hardware + software we have today.) - Language models might be the perfect tool to distill the unstructured corpus of public data, literature, and methods sitting around on the Internet into real biological insights for scientist asking questions in natural language. It will also allow them to install, configure and run the slew of useful but poorly maintained academic computational tools to explore and hypothesize new biology on their own. Increasing the productivity of each scientist will do much to reverse Eroom's law. - There is an appetite from the market for new applications of biotech beyond drug development. Bacteria driven lithium mining (maverick), cell agriculture (growing cows in vats), early signs of consumer biologics (Geltor), biofoundries (Ginkgo). My guess is some of the greatest minds of our generation will want to do more than perturb the human body with therapies. I think it is also important to recognize that the need for computers and software is a secular trend in the progression of biotech independent of the interests of Silicon Valley. Clusters of computers, the information they store and the software that runs on them are precisely the technology needed by this field as it transforms into a discipline of information management, towards reducing living things into well characterized building blocks we can rebuild in our image. Software companies, as some local and hyper efficient structure in the arc of capitalism, with established methods and well trod rails to attract resources, talent and easily distribute product to an entire market, are the perfect place to incubate and disseminate these tools. There will probably be many, very large computer companies in biology in the next century.
0:52

I've had many engineers ask me why its worth their time and effort to learn biology in response to this post. Why should they be excited? We are poised for a revolution in biotech that will be uniquely enabled by computers. Convince yourself by digging into the examples I link: - The tooling is getting better. Assays are able to measure a broad array of molecules at a falling cost and increasing throughput. Look to ScaleBio, Curio Bioscience, AtlasXOmics for inspiration. We are sequencing millions of single cells and building spatial maps of the molecular state of tumors. After two generations of "next-generation sequencing", and stagnating DNA read + write costs under the monopoly of Illumina, this wave of the new assays will have a profound impact on the iteration speed and scale of experimentation. Little needs to be said about the impact of compounding trends in core tooling over a sufficient period of time. - Biotechs are using data to guide decisions and are incorporating domain informed machine learning as a core part of the molecular design process. There is great synergy here with better tooling as a means of abundant, cheap data. Look to Recursion, Manifold Bio, Dyno Tx, Asimov. There is also the cross pollination of biology informed architectures with the recent explosion in new machine learning techniques. These models are starting to do useful things, like generate functional gene editing proteins and entire prokaryotic organisms. Look to ESM, AlphaFold3, RFdiffusion. - New classes of therapies - genetic medicines and engineered immune cells - are having real success in the clinic. One dose cures for cardiovascular disease (Verve w ACSD), vaccines for cancer (Moderna w mRNA 4157), in vivo gene editing proteins (CRISPR Tx, Beam, Ensoma), metabolic disease (Novartis, Eli Lily w GLP-1/GIP modulators) are being dosed in real people right now + transforming lives. - The AI craze is commoditizing accelerated hardware and fast storage devices like NVMes, improving developer frameworks for writing code against these devices and maturing the systems tooling for moving around lots of data between computers for distributed training. One happy accident of this bubble will be the reuse of these components to build a new systems stack for the large scale processing of molecular data. This will be very important to construct a 1 billion single cell atlas and beyond. (For reference, the state of the art is ScaleBio's 2M cell kit, dubbed QuantumScale, and it is pushing things with the hardware + software we have today.) - Language models might be the perfect tool to distill the unstructured corpus of public data, literature, and methods sitting around on the Internet into real biological insights for scientist asking questions in natural language. It will also allow them to install, configure and run the slew of useful but poorly maintained academic computational tools to explore and hypothesize new biology on their own. Increasing the productivity of each scientist will do much to reverse Eroom's law. - There is an appetite from the market for new applications of biotech beyond drug development. Bacteria driven lithium mining (maverick), cell agriculture (growing cows in vats), early signs of consumer biologics (Geltor), biofoundries (Ginkgo). My guess is some of the greatest minds of our generation will want to do more than perturb the human body with therapies. I think it is also important to recognize that the need for computers and software is a secular trend in the progression of biotech independent of the interests of Silicon Valley. Clusters of computers, the information they store and the software that runs on them are precisely the technology needed by this field as it transforms into a discipline of information management, towards reducing living things into well characterized building blocks we can rebuild in our image. Software companies, as some local and hyper efficient structure in the arc of capitalism, with established methods and well trod rails to attract resources, talent and easily distribute product to an entire market, are the perfect place to incubate and disseminate these tools. There will probably be many, very large computer companies in biology in the next century.

Kenny Workman

113,534 次观看 • 1 年前

In this post I will explain why people become borderline religious when they discover Qubic. Now with video. Please repost. I want people to learn about QUBIC. The ecosystem consists of 3 separate universes: AI, Mining, and Tickchain. AI is the primary product and purpose of QUBIC and it is supported by Mining to train the AI and by Tickchin for validation and decentralization. Here is how this whole thing works: AI: Let’s start with the AI. The main purpose of QUBIC is creating AGI (Artificial General Intelligence). It’s a type of AI that can self-develop, set tasks, grow, and learn on its own—much like the human brain does. This product is called AIGarth and it uses many cool ideas where AIs can create their own agents and have them compete against one another to evolve. It is basically robots creating robots with the survival-of-the-fittest evolution approach. Very impressive and thought out. To develop such an AI there are several requirements that even the industry giants like OpenAI, Microsoft, and Tesla are missing. One of them is the data processing for AI training. I mean they have their Datacenters, but those are only good enough to train limited Large Language Models such as ChatGPT and Grok. Mining / Training: Now QUBIC solves this problem with its mining architecture. Keep in mind, mining in QUBIC does not secure the chain, primarily it provides the processing power for training the AI. In a sense, QUBIC mining creates the largest distributed datacenter in the world, where individual miners provide their computers for training the AI and get paid with newly issued QUBIC coins. This way QUBIC gets constantly increasing processing power without having to really pay for the infrastructure. And here is another impressive bit of info. QUBIC’s distributed mining network currently ranks above the #1 supercomputer in the world - El Capitan. QUBIC Tickchain The QUBIC chain ties its AI and Mining together to create decentralization, the reward system for miners, it acts as a decision voting system for future development, and it allows AIGarth to function independently through Smart Contracts. In this summary I will not go over the specifics of QUBIC Tickchain. It’s pretty complex so it will be a separate post. Now, it’s an absolute genius piece of tech, which I consider the most advanced product within crypto industry. It is important to know that QUBIC chain runs directly out of Random Access Memory of its validators. It has instant finality and acts as its own operating system. That allows for speeds only bound by current hardware capabilities and it only increases as technology progresses. As I am writing this, QUBIC Tickchain is fully functional and it already hosts several smart-contract based web3 applications. QUBIC has designed its chain to be this fast for a single purpose, to give its future AI the speed it needs to evolve and to react quickly to the outside world. Ilya Shutskever the scientist, who developed ChatGPT clearly states that next generation superintelligence will make decisions in split second with less data. I believe QUBIC is that next generation. Why QUBIC? So out of the sea of AI projects in crypto why is QUBIC my #1 pick? Well, the first reason is that QUBIC is a unicorn AI startup that happens to use blockchain tech to reach it’s goals. In the real world of Venture Capital it would be fully funded instantly and you would not be invited. Second reason is that the industry admits that Large Language Models have plateaued. Even with enough processing power there is only so much information they can add to their data. Even Google CEO admits that. New approach is needed because the future progress is not possible with LLMs. The third reason is because Large Language Models will not create true AGI. It is evident by Ilya Shutskver latest presentation. Sam Altman of OpenAI is trying to change the definition of what is considered AGI just to lower the plank for his own product. Microsoft’s AI chief is now claiming that it would take 10 years to reach AGI, while QUBIC aims to do this in 2027. All these big players are using wrong technology for what they are trying to achieve and there isn’t enough investor funding for them to pivot. The fourth reason is that QUBIC is headed by Sergey Ivancheglo and 2 renowned AI scientists. Many claim Segey is the creator of Bitcoin. He was the 3rd person to mine bitcoin, he invented Proof of Stake consensus, which Ethereum uses now, he ran the first ICO, and he created 2 of the top gainers in crypto NXT and IOTA. QUBIC is his grand finale after 12 years of development and trials. I am including links below the post as the proof of my claims. Thank you for your time. Please live a like or a comment. It helps me continue making these extensive posts and videos.
5:57

In this post I will explain why people become borderline religious when they discover Qubic. Now with video. Please repost. I want people to learn about QUBIC. The ecosystem consists of 3 separate universes: AI, Mining, and Tickchain. AI is the primary product and purpose of QUBIC and it is supported by Mining to train the AI and by Tickchin for validation and decentralization. Here is how this whole thing works: AI: Let’s start with the AI. The main purpose of QUBIC is creating AGI (Artificial General Intelligence). It’s a type of AI that can self-develop, set tasks, grow, and learn on its own—much like the human brain does. This product is called AIGarth and it uses many cool ideas where AIs can create their own agents and have them compete against one another to evolve. It is basically robots creating robots with the survival-of-the-fittest evolution approach. Very impressive and thought out. To develop such an AI there are several requirements that even the industry giants like OpenAI, Microsoft, and Tesla are missing. One of them is the data processing for AI training. I mean they have their Datacenters, but those are only good enough to train limited Large Language Models such as ChatGPT and Grok. Mining / Training: Now QUBIC solves this problem with its mining architecture. Keep in mind, mining in QUBIC does not secure the chain, primarily it provides the processing power for training the AI. In a sense, QUBIC mining creates the largest distributed datacenter in the world, where individual miners provide their computers for training the AI and get paid with newly issued QUBIC coins. This way QUBIC gets constantly increasing processing power without having to really pay for the infrastructure. And here is another impressive bit of info. QUBIC’s distributed mining network currently ranks above the #1 supercomputer in the world - El Capitan. QUBIC Tickchain The QUBIC chain ties its AI and Mining together to create decentralization, the reward system for miners, it acts as a decision voting system for future development, and it allows AIGarth to function independently through Smart Contracts. In this summary I will not go over the specifics of QUBIC Tickchain. It’s pretty complex so it will be a separate post. Now, it’s an absolute genius piece of tech, which I consider the most advanced product within crypto industry. It is important to know that QUBIC chain runs directly out of Random Access Memory of its validators. It has instant finality and acts as its own operating system. That allows for speeds only bound by current hardware capabilities and it only increases as technology progresses. As I am writing this, QUBIC Tickchain is fully functional and it already hosts several smart-contract based web3 applications. QUBIC has designed its chain to be this fast for a single purpose, to give its future AI the speed it needs to evolve and to react quickly to the outside world. Ilya Shutskever the scientist, who developed ChatGPT clearly states that next generation superintelligence will make decisions in split second with less data. I believe QUBIC is that next generation. Why QUBIC? So out of the sea of AI projects in crypto why is QUBIC my #1 pick? Well, the first reason is that QUBIC is a unicorn AI startup that happens to use blockchain tech to reach it’s goals. In the real world of Venture Capital it would be fully funded instantly and you would not be invited. Second reason is that the industry admits that Large Language Models have plateaued. Even with enough processing power there is only so much information they can add to their data. Even Google CEO admits that. New approach is needed because the future progress is not possible with LLMs. The third reason is because Large Language Models will not create true AGI. It is evident by Ilya Shutskver latest presentation. Sam Altman of OpenAI is trying to change the definition of what is considered AGI just to lower the plank for his own product. Microsoft’s AI chief is now claiming that it would take 10 years to reach AGI, while QUBIC aims to do this in 2027. All these big players are using wrong technology for what they are trying to achieve and there isn’t enough investor funding for them to pivot. The fourth reason is that QUBIC is headed by Sergey Ivancheglo and 2 renowned AI scientists. Many claim Segey is the creator of Bitcoin. He was the 3rd person to mine bitcoin, he invented Proof of Stake consensus, which Ethereum uses now, he ran the first ICO, and he created 2 of the top gainers in crypto NXT and IOTA. QUBIC is his grand finale after 12 years of development and trials. I am including links below the post as the proof of my claims. Thank you for your time. Please live a like or a comment. It helps me continue making these extensive posts and videos.

retrodrive ⛏

24,639 次观看 • 1 年前

Apple’s iPad “Crush” Ad Is Bleak, Ominous and Threatening I don’t know if you’ve seen Apple’s just-released commercial for the “New” iPad Pro, but it’s pretty awful. It is dark, humorless, and feels like a not-so-thinly veiled threat to writers, musicians, game makers, developers, and artists of all kinds. …and children, even. I’ve watched it at least five times today alone, and I’m left with one big question. “Who on earth approved this?” It’s absolutely baffling that the world’s largest technology company, with the world’s biggest marketing budget, thought this would be a good idea. What kind of idiot—or idiots, since dozens or hundreds of people had to be involved in the writing, staging, producing, recording, and editing—felt this kind of ad would somehow create a positive emotional connection with consumers? Seriously, it’s terrifying. In a dank, cold warehouse, devoid of all life and humanity, an industrial crusher comes to life, and slowly starts destroying a collection of musical, philosophical, and artistic devices and instruments. For no apparent reason, everything starts getting smashed: first, a trumpet, then an arcade video game, then cans of paint, a piano, a globe, a metronome, a guitar… on and on it goes, obliterating everything in sight into a colorful, gooey, explosive mess. Books, camera lenses, lamps, a guitar, a sculpture, and a typewriter—all tools of the liberal arts—get mangled into a garbage heap as Sonny & Cher cheerfully sing, “All I ever need is you.” In the penultimate moment, a goofy yellow smiley emoji becomes a bug-eyed scary-clown freak as it, too, is crushed to death. Worse, if you enable closed captions like I do by default, the video says: “[POPPING] [SPLAT]” right as its eyeballs pop out of its head when Cher sings, “Give me a reason to build my world around you.” It’s enough to make a child cry. It has all the comforting vibes of the burnt pink teddy bear floating in the swimming pool on Breaking Bad after two planes crash in mid-air. I have so many questions (aside from simply wondering the names of the soon-to-be ex-employees who greenlit this abomination). First of all, as a trumpet player myself, I am personally offended that they made me watch a perfectly good trumpet get smashed to smithereens like it’s no big deal. Why would they torture me like this? Second of all, what is the message here? No, not that “the most powerful iPad ever is also the thinnest,” as the voiceover artist states in the last few seconds of the clip. I mean: what is the message? Ostensibly, pulverizing children’s toys, arcade games, architectural models, and ceramic Angry Birds into a paste implies something like “We’re taking all the best of humanity; all the collective works of Western Civilization, smashing it into pieces and putting it inside this remarkably thin device so you can have all of it in the palm of your hand.” But my oh my, is there an elephant in this room… he’s hiding behind the monstrous destroying machine. Did anyone inside Apple realize that everyone outside Apple will recognize this imagery in a metaphorical sense, but not the one Apple intended? We don’t see a crushing machine gently consolidating the greatest output of all our artistic endeavors, simply reformatted for a digital age and consumed by everyone with instant, fingertip access. We see what is painstakingly obvious to us, and the timing couldn’t possibly be worse. We see a giant, soulless machine consuming our work in a very different way. Right now, AI models are training themselves on our intellectual property and even our very own personally-identifying data. We aren’t the ones doing the consuming. We’re the ones being consumed. The tech industry has become one massive gaping maw, opening wide and swallowing everything in sight, chewing it up into little bits and pieces of comminuted waste, like a paper shredder or a garbage disposal. It’s destruction in its most literal form. And for what? For a newer version of the iPad that is only slightly thinner than its predecessor? For an only marginally improved version of Apple’s tablet device that has been around for 14 years? For increased profits? This is a terrible look for Apple. They may as well be saying: “All your work are belong to us.” Personally, I am a fan of artificial intelligence. I am eagerly embracing our robot overlords and I welcome our new CSV god (as the actual developers of AI models like to say). I look forward to the freedom and innovation that will come as a result of humanity augmenting our intelligence with AI like a force multiplier on a battlefield. But if Apple has the same perspective I do, they’re selling it in the worst possible way. When I see this video, I see that Apple is definitely crushing something… but I’m not sure what. -Crushing small companies that develop apps for the extremely heavy-handed App Store, which imposes byzantine restrictions on what they can and can’t do with their own apps? -Crushing competitors by limiting what they can do on the iOS and MacOS platforms with arbitrary and capricious rules about enabling functionalities that Apple doesn’t like, even if users do? -Crushing publishers and content creators with a punitive 30% fee on all subscriptions and in-app purchases? -Crushing choice and competition by not allowing app makers to make apps and programs that do the same thing that native apps already do, even if they do it better? -Crushing all human creativity and innovation by automating and systematizing everything? In the early days of the “Google vs. Apple” fight over the web and app stores, I was really concerned that Google was becoming way, way too powerful. Specifically, in 2015, when Google came up with “app streaming,” they announced a desire to form a “web of apps.” This was concerning. Especially when coupled with Google’s efforts to steal content from other websites and provide it to users via the “knowledge graph” results, ending up with the creation of “zero-click” search results pages, which absolutely punished website owners and content creators. By taking the most valuable content off a website and showing it to Google users without them needing to click through to the website itself, Google had essentially stolen everybody’s intellectual property with only the most minimal attribution possible (to fend off lawsuits no doubt, but with no intention of users actually visiting the website in question anymore). “Google is eating the internet,” I thought, and said out loud, (although I probably wasn’t the first person to use that phrase) But what I meant was purely an analogy. It was vague and ambiguous, almost silly. Maybe I was wrong, though: maybe it’s Apple that’s doing the eating. Maybe Apple is not only gobbling up everyone else’s work, but also homogenizing it—and us—and forcing us to use their platform, pay their fees, abide by their rules, and constantly keep upgrading, upgrading, upgrading, to an ever-thinner iPad in order to use it. Watch the video again. This is the stuff of nightmares. To be perfectly fair, even if I were to take the commercial at face value and ignore it’s off-the-charts creepiness and just stick to its one stated claim—that the new iPad Pro is thinner—it still fails as a commercial. Why? Because nobody cares how thin an iPad is. Seriously. I’ve owned an iPad since 2010: that means I’ve carried around a version of Apple’s already-thin tablet every day for over a dozen years. Never once have I said to myself: “You know what improvement I’d really like to see in this thing? I wish it were thinner.” Never. That thought has never crossed my mind, even once. You know what has? -Better battery life. -I’d like my iPad to not get hot to the touch when I use the Apple Pencil to take notes. -I wish it wasn’t so fragile: I dropped my brand-new iPad 2 back in the day when it slipped out of the arm-hold I was carrying it in, it bounced on the pavement, and the screen shattered into a thousand pieces, making it unusable. -I wish it had more storage. -I wish Apple would stop changing the type of cable connector it uses: I’ve gone from the original 30-pin connector to the Lightning connector, and now to the current USB-C/Thunderbolt connector. -I wish I could view the screen in direct sunlight. -I wish it wouldn’t overheat and turn off automatically when I use it outdoors in the summertime. Those are announcements I would welcome in a new iPad Pro commercial. None of this “now even thinner” nonsense nobody needs or cares about. So, back to the commercial. In my opinion, whoever made this ad should be fired. I almost never say that about other companies, especially for good-faith marketing efforts gone wrong… those of us who work in marketing make mistakes sometimes, and we learn from them. But cases like this warrant a special exception. Marketing and advertising are designed to make people want to buy your products. This commercial doesn’t just not make me want to buy Apple’s products. It makes me not want to buy Apple’s products, which is something altogether different. It turns me from someone who likes iPads into someone who is almost rethinking iPads entirely. That’s not just a bad advertisement; it’s a harmful advertisement. Apple’s usually known for great commercials. The legendary 1984 Super Bowl commercial was, of course, their best. I thought “Hello, I’m a Mac” was absolutely brilliant. They have made some missteps along the way, but this one is really bad. Not even their nauseatingly preachy and woke “Mother Nature” ad from a few years ago was this bad. Steve Jobs once said, “Technology alone is not enough—it’s technology married with liberal arts, married with the humanities, that yields us the results that make our heart sing.” My goodness, that last line alone is poetry itself! This ad seems to be Apple signaling that they don’t believe in that anymore. And I don’t think all this handwringing is an overreaction to where you could say “Oh, c’mon, it’s just a commercial! What’s the big deal?” It is a big deal. It tells you about the values of the company, and what they intend to communicate. Really, how is this the same company that used to sell iPhones by showing grandmas using FaceTime to connect with their baby grandchildren from afar during the holidays? Everything about it is wrong: even the thumbnail they chose for it (the bulging-eyed smiley face) and the fact that they gave it the title “Crush!” It was fun to see the reactions to the video online today. I find it fascinating that Apple shared it on YouTube but turned off the comments. On X, Tim Cook shared it Tuesday, and the video, which so richly deserves to be mocked, is getting it in spades. Some people are calling it “anti-art.” One user called it “soul-crushing,” which was about as literal and logical a response as you’d expect. It turns out Apple actually made an announcement about the commercial. In response to the (apparently unexpected) poor welcome it got, Apple wrote: “We missed the mark with this video, and we’re sorry.” Lame response from a tone-deaf tech behemoth, but still, they hopefully got the message. C’mon, Apple. I have seen the future, and this ain’t it.
1:08

Apple’s iPad “Crush” Ad Is Bleak, Ominous and Threatening I don’t know if you’ve seen Apple’s just-released commercial for the “New” iPad Pro, but it’s pretty awful. It is dark, humorless, and feels like a not-so-thinly veiled threat to writers, musicians, game makers, developers, and artists of all kinds. …and children, even. I’ve watched it at least five times today alone, and I’m left with one big question. “Who on earth approved this?” It’s absolutely baffling that the world’s largest technology company, with the world’s biggest marketing budget, thought this would be a good idea. What kind of idiot—or idiots, since dozens or hundreds of people had to be involved in the writing, staging, producing, recording, and editing—felt this kind of ad would somehow create a positive emotional connection with consumers? Seriously, it’s terrifying. In a dank, cold warehouse, devoid of all life and humanity, an industrial crusher comes to life, and slowly starts destroying a collection of musical, philosophical, and artistic devices and instruments. For no apparent reason, everything starts getting smashed: first, a trumpet, then an arcade video game, then cans of paint, a piano, a globe, a metronome, a guitar… on and on it goes, obliterating everything in sight into a colorful, gooey, explosive mess. Books, camera lenses, lamps, a guitar, a sculpture, and a typewriter—all tools of the liberal arts—get mangled into a garbage heap as Sonny & Cher cheerfully sing, “All I ever need is you.” In the penultimate moment, a goofy yellow smiley emoji becomes a bug-eyed scary-clown freak as it, too, is crushed to death. Worse, if you enable closed captions like I do by default, the video says: “[POPPING] [SPLAT]” right as its eyeballs pop out of its head when Cher sings, “Give me a reason to build my world around you.” It’s enough to make a child cry. It has all the comforting vibes of the burnt pink teddy bear floating in the swimming pool on Breaking Bad after two planes crash in mid-air. I have so many questions (aside from simply wondering the names of the soon-to-be ex-employees who greenlit this abomination). First of all, as a trumpet player myself, I am personally offended that they made me watch a perfectly good trumpet get smashed to smithereens like it’s no big deal. Why would they torture me like this? Second of all, what is the message here? No, not that “the most powerful iPad ever is also the thinnest,” as the voiceover artist states in the last few seconds of the clip. I mean: what is the message? Ostensibly, pulverizing children’s toys, arcade games, architectural models, and ceramic Angry Birds into a paste implies something like “We’re taking all the best of humanity; all the collective works of Western Civilization, smashing it into pieces and putting it inside this remarkably thin device so you can have all of it in the palm of your hand.” But my oh my, is there an elephant in this room… he’s hiding behind the monstrous destroying machine. Did anyone inside Apple realize that everyone outside Apple will recognize this imagery in a metaphorical sense, but not the one Apple intended? We don’t see a crushing machine gently consolidating the greatest output of all our artistic endeavors, simply reformatted for a digital age and consumed by everyone with instant, fingertip access. We see what is painstakingly obvious to us, and the timing couldn’t possibly be worse. We see a giant, soulless machine consuming our work in a very different way. Right now, AI models are training themselves on our intellectual property and even our very own personally-identifying data. We aren’t the ones doing the consuming. We’re the ones being consumed. The tech industry has become one massive gaping maw, opening wide and swallowing everything in sight, chewing it up into little bits and pieces of comminuted waste, like a paper shredder or a garbage disposal. It’s destruction in its most literal form. And for what? For a newer version of the iPad that is only slightly thinner than its predecessor? For an only marginally improved version of Apple’s tablet device that has been around for 14 years? For increased profits? This is a terrible look for Apple. They may as well be saying: “All your work are belong to us.” Personally, I am a fan of artificial intelligence. I am eagerly embracing our robot overlords and I welcome our new CSV god (as the actual developers of AI models like to say). I look forward to the freedom and innovation that will come as a result of humanity augmenting our intelligence with AI like a force multiplier on a battlefield. But if Apple has the same perspective I do, they’re selling it in the worst possible way. When I see this video, I see that Apple is definitely crushing something… but I’m not sure what. -Crushing small companies that develop apps for the extremely heavy-handed App Store, which imposes byzantine restrictions on what they can and can’t do with their own apps? -Crushing competitors by limiting what they can do on the iOS and MacOS platforms with arbitrary and capricious rules about enabling functionalities that Apple doesn’t like, even if users do? -Crushing publishers and content creators with a punitive 30% fee on all subscriptions and in-app purchases? -Crushing choice and competition by not allowing app makers to make apps and programs that do the same thing that native apps already do, even if they do it better? -Crushing all human creativity and innovation by automating and systematizing everything? In the early days of the “Google vs. Apple” fight over the web and app stores, I was really concerned that Google was becoming way, way too powerful. Specifically, in 2015, when Google came up with “app streaming,” they announced a desire to form a “web of apps.” This was concerning. Especially when coupled with Google’s efforts to steal content from other websites and provide it to users via the “knowledge graph” results, ending up with the creation of “zero-click” search results pages, which absolutely punished website owners and content creators. By taking the most valuable content off a website and showing it to Google users without them needing to click through to the website itself, Google had essentially stolen everybody’s intellectual property with only the most minimal attribution possible (to fend off lawsuits no doubt, but with no intention of users actually visiting the website in question anymore). “Google is eating the internet,” I thought, and said out loud, (although I probably wasn’t the first person to use that phrase) But what I meant was purely an analogy. It was vague and ambiguous, almost silly. Maybe I was wrong, though: maybe it’s Apple that’s doing the eating. Maybe Apple is not only gobbling up everyone else’s work, but also homogenizing it—and us—and forcing us to use their platform, pay their fees, abide by their rules, and constantly keep upgrading, upgrading, upgrading, to an ever-thinner iPad in order to use it. Watch the video again. This is the stuff of nightmares. To be perfectly fair, even if I were to take the commercial at face value and ignore it’s off-the-charts creepiness and just stick to its one stated claim—that the new iPad Pro is thinner—it still fails as a commercial. Why? Because nobody cares how thin an iPad is. Seriously. I’ve owned an iPad since 2010: that means I’ve carried around a version of Apple’s already-thin tablet every day for over a dozen years. Never once have I said to myself: “You know what improvement I’d really like to see in this thing? I wish it were thinner.” Never. That thought has never crossed my mind, even once. You know what has? -Better battery life. -I’d like my iPad to not get hot to the touch when I use the Apple Pencil to take notes. -I wish it wasn’t so fragile: I dropped my brand-new iPad 2 back in the day when it slipped out of the arm-hold I was carrying it in, it bounced on the pavement, and the screen shattered into a thousand pieces, making it unusable. -I wish it had more storage. -I wish Apple would stop changing the type of cable connector it uses: I’ve gone from the original 30-pin connector to the Lightning connector, and now to the current USB-C/Thunderbolt connector. -I wish I could view the screen in direct sunlight. -I wish it wouldn’t overheat and turn off automatically when I use it outdoors in the summertime. Those are announcements I would welcome in a new iPad Pro commercial. None of this “now even thinner” nonsense nobody needs or cares about. So, back to the commercial. In my opinion, whoever made this ad should be fired. I almost never say that about other companies, especially for good-faith marketing efforts gone wrong… those of us who work in marketing make mistakes sometimes, and we learn from them. But cases like this warrant a special exception. Marketing and advertising are designed to make people want to buy your products. This commercial doesn’t just not make me want to buy Apple’s products. It makes me not want to buy Apple’s products, which is something altogether different. It turns me from someone who likes iPads into someone who is almost rethinking iPads entirely. That’s not just a bad advertisement; it’s a harmful advertisement. Apple’s usually known for great commercials. The legendary 1984 Super Bowl commercial was, of course, their best. I thought “Hello, I’m a Mac” was absolutely brilliant. They have made some missteps along the way, but this one is really bad. Not even their nauseatingly preachy and woke “Mother Nature” ad from a few years ago was this bad. Steve Jobs once said, “Technology alone is not enough—it’s technology married with liberal arts, married with the humanities, that yields us the results that make our heart sing.” My goodness, that last line alone is poetry itself! This ad seems to be Apple signaling that they don’t believe in that anymore. And I don’t think all this handwringing is an overreaction to where you could say “Oh, c’mon, it’s just a commercial! What’s the big deal?” It is a big deal. It tells you about the values of the company, and what they intend to communicate. Really, how is this the same company that used to sell iPhones by showing grandmas using FaceTime to connect with their baby grandchildren from afar during the holidays? Everything about it is wrong: even the thumbnail they chose for it (the bulging-eyed smiley face) and the fact that they gave it the title “Crush!” It was fun to see the reactions to the video online today. I find it fascinating that Apple shared it on YouTube but turned off the comments. On X, Tim Cook shared it Tuesday, and the video, which so richly deserves to be mocked, is getting it in spades. Some people are calling it “anti-art.” One user called it “soul-crushing,” which was about as literal and logical a response as you’d expect. It turns out Apple actually made an announcement about the commercial. In response to the (apparently unexpected) poor welcome it got, Apple wrote: “We missed the mark with this video, and we’re sorry.” Lame response from a tone-deaf tech behemoth, but still, they hopefully got the message. C’mon, Apple. I have seen the future, and this ain’t it.

Ron Stauffer

19,018 次观看 • 2 年前

I know your timeline is flooded now with word salads of "insane, HER, 10 features you missed, we're so back". Sit down. Chill. Take a deep breath like Mark does in the demo . Let's think step by step: - Technique-wise, OpenAI has figured out a way to map audio to audio directly as first-class modality, and stream videos to a transformer in real-time. These require some new research on tokenization and architecture, but overall it's a data and system optimization problem (as most things are). High-quality data can come from at least 2 sources: 1) Naturally occurring dialogues on YouTube, podcasts, TV series, movies, etc. Whisper can be trained to identify speaker turns in a dialogue or separate overlapping speeches for automated annotation. 2) Synthetic data. Run the slow 3-stage pipeline using the most powerful models: speech1->text1 (ASR), text1->text2 (LLM), text2->speech2 (TTS). The middle LLM can decide when to stop and also simulate how to resume from interruption. It could output additional "thought traces" that are not verbalized to help generate better reply. Then GPT-4o distills directly from speech1->speech2, with optional auxiliary loss functions based on the 3-stage data. After distillation, these behaviors are now baked into the model without emitting intermediate texts. On the system side: the latency would not meet real-time threshold if every video frame is decompressed into an RGB image. OpenAI has likely developed their own neural-first, streaming video codec to transmit the motion deltas as tokens. The communication protocol and NN inference must be co-optimized. For example, there could be a small and energy-efficient NN running on the edge device that decides to transmit more tokens if the video is interesting, and fewer otherwise. - I didn't expect GPT-4o to be closer to GPT-5, the rumored "Arrakis" model that takes multimodal in and out. In fact, it's likely an early checkpoint of GPT-5 that hasn't finished training yet. The branding betrays a certain insecurity. Ahead of Google I/O, OpenAI would rather beat our mental projection of GPT-4.5 than disappoint by missing the sky-high expectation for GPT-5. A smart move to buy more time. - Notably, the assistant is much more lively and even a bit flirty. GPT-4o is trying (perhaps a bit too hard) to sound like HER. OpenAI is eating Character AI's lunch, with almost 100% overlap in form factor and huge distribution channels. It's a pivot towards more emotional AI with strong personality, which OpenAI seemed to actively suppress in the past. - Whoever wins Apple first wins big time. I see 3 levels of integration with iOS: 1) Ditch Siri. OpenAI distills a smaller-tier, purely on-device GPT-4o for iOS, with optional paid upgrade to use the cloud. 2) Native features to stream the camera or screen into the model. Chip-level support for neural audio/video codec. 3) Integrate with iOS system-level action API and smart home APIs. No one uses Siri Shortcuts, but it's time to resurrect. This could become the AI agent product with a billion users from the get-go. The FSD for smartphones with a Tesla-scale data flywheel.
14:22

I know your timeline is flooded now with word salads of "insane, HER, 10 features you missed, we're so back". Sit down. Chill. Take a deep breath like Mark does in the demo . Let's think step by step: - Technique-wise, OpenAI has figured out a way to map audio to audio directly as first-class modality, and stream videos to a transformer in real-time. These require some new research on tokenization and architecture, but overall it's a data and system optimization problem (as most things are). High-quality data can come from at least 2 sources: 1) Naturally occurring dialogues on YouTube, podcasts, TV series, movies, etc. Whisper can be trained to identify speaker turns in a dialogue or separate overlapping speeches for automated annotation. 2) Synthetic data. Run the slow 3-stage pipeline using the most powerful models: speech1->text1 (ASR), text1->text2 (LLM), text2->speech2 (TTS). The middle LLM can decide when to stop and also simulate how to resume from interruption. It could output additional "thought traces" that are not verbalized to help generate better reply. Then GPT-4o distills directly from speech1->speech2, with optional auxiliary loss functions based on the 3-stage data. After distillation, these behaviors are now baked into the model without emitting intermediate texts. On the system side: the latency would not meet real-time threshold if every video frame is decompressed into an RGB image. OpenAI has likely developed their own neural-first, streaming video codec to transmit the motion deltas as tokens. The communication protocol and NN inference must be co-optimized. For example, there could be a small and energy-efficient NN running on the edge device that decides to transmit more tokens if the video is interesting, and fewer otherwise. - I didn't expect GPT-4o to be closer to GPT-5, the rumored "Arrakis" model that takes multimodal in and out. In fact, it's likely an early checkpoint of GPT-5 that hasn't finished training yet. The branding betrays a certain insecurity. Ahead of Google I/O, OpenAI would rather beat our mental projection of GPT-4.5 than disappoint by missing the sky-high expectation for GPT-5. A smart move to buy more time. - Notably, the assistant is much more lively and even a bit flirty. GPT-4o is trying (perhaps a bit too hard) to sound like HER. OpenAI is eating Character AI's lunch, with almost 100% overlap in form factor and huge distribution channels. It's a pivot towards more emotional AI with strong personality, which OpenAI seemed to actively suppress in the past. - Whoever wins Apple first wins big time. I see 3 levels of integration with iOS: 1) Ditch Siri. OpenAI distills a smaller-tier, purely on-device GPT-4o for iOS, with optional paid upgrade to use the cloud. 2) Native features to stream the camera or screen into the model. Chip-level support for neural audio/video codec. 3) Integrate with iOS system-level action API and smart home APIs. No one uses Siri Shortcuts, but it's time to resurrect. This could become the AI agent product with a billion users from the get-go. The FSD for smartphones with a Tesla-scale data flywheel.

Jim Fan

991,628 次观看 • 2 年前

🎉🎉🎉 It's my BIRTHDAY! I'm an indie dev. I'm using today to be proud, which is often frowned upon. I made this hex grid w/ level editor in 3 weeks in UE, by myself. MY GOAL: Release new games every 2 weeks by end of year. This is my story, and why you should follow along: --- The video of what I made was made 4 MONTHS before the release of UE for Fortnite, while I worked at Epic a year and a half ago. I did it over our winter break. 80 hours of work across 3 weeks. I time tracked and worklogged the whole thing. This was before we had MoveTo(), SetMesh/Material(), Animation Controller was broken, you couldn't reference to other verse devices in editables. It was PAINFUL. Most of my time was spent duplicating and editing sequences. There were 631 hexes... and like 6 sequences per hex and 13 gameplay devices per hex. It was hacky, but it worked. 😅😅😅 NOTE... this was with a new programming language that NOBODY outside of Epic had really used using a HIGHLY buggy in-development tool. There were no tutorials, or stackoverflow, or communities to ask for help. Even my coworkers were on winter break, so couldn't ask them for help either. Everything I did was from my raw experience as a Software Architect. And I'll argue, that what I did was FAR more advanced and complicated from an engineering aspect than ANYTHING released by the UEFN community to this day. I am SOOO proud of that work and what it represents. 😁😁😁 --- My special interest is productivity in making things, and using that to enable others to make great things easier and faster. User Generated Content (UGC) platforms like Minecraft/Roblox/Fortnite/Halo, have been my passion for 17 years now since Halo 3 released. I've been a software engineer for 22 years... I turn 36 today, so since I was 14. 🎉🎉🎉 I'm autistic, so I literally live, eat, and breath UGC. Because I see it as a path to build what's in your dreams. The dreams of a kid who played a game that made a difference in their life of one day making their own games. I want to enable that for everyone who wants it. Because it took me FAR too long to get into the gaming industry. People see it as complicated and difficult and competitive. I want to lower those barriers as my legacy. 🤓🤓🤓 --- My new company, Creative Force, has a goal of "Building games at the speed of thought". I'm designing a general system to make games faster that could be used in any game engine. A design language that can be read by an engine that will "build your game for you". Basically allowing game designers to make "recipes" for games, and let other people use those recipes anywhere, and modify them easily to experiment and have a creative outlet. Maybe one day be a game chef and make a living doing what they dreamed, without needing crazy technical experience. 👩‍🍳👩‍🍳👩‍🍳 I believe this is possible with modern learnings from software that haven't been used in game development fully yet. A game architecture that will take the technical out of making games, and make it as approachable and as deep as writing a novel or painting a masterpiece. This is my dream... and I hope you'll follow along with me. Cuz my year #36, is quite literally a GAME CHANGING year. Love y'all 💖💖💖
7:05

🎉🎉🎉 It's my BIRTHDAY! I'm an indie dev. I'm using today to be proud, which is often frowned upon. I made this hex grid w/ level editor in 3 weeks in UE, by myself. MY GOAL: Release new games every 2 weeks by end of year. This is my story, and why you should follow along: --- The video of what I made was made 4 MONTHS before the release of UE for Fortnite, while I worked at Epic a year and a half ago. I did it over our winter break. 80 hours of work across 3 weeks. I time tracked and worklogged the whole thing. This was before we had MoveTo(), SetMesh/Material(), Animation Controller was broken, you couldn't reference to other verse devices in editables. It was PAINFUL. Most of my time was spent duplicating and editing sequences. There were 631 hexes... and like 6 sequences per hex and 13 gameplay devices per hex. It was hacky, but it worked. 😅😅😅 NOTE... this was with a new programming language that NOBODY outside of Epic had really used using a HIGHLY buggy in-development tool. There were no tutorials, or stackoverflow, or communities to ask for help. Even my coworkers were on winter break, so couldn't ask them for help either. Everything I did was from my raw experience as a Software Architect. And I'll argue, that what I did was FAR more advanced and complicated from an engineering aspect than ANYTHING released by the UEFN community to this day. I am SOOO proud of that work and what it represents. 😁😁😁 --- My special interest is productivity in making things, and using that to enable others to make great things easier and faster. User Generated Content (UGC) platforms like Minecraft/Roblox/Fortnite/Halo, have been my passion for 17 years now since Halo 3 released. I've been a software engineer for 22 years... I turn 36 today, so since I was 14. 🎉🎉🎉 I'm autistic, so I literally live, eat, and breath UGC. Because I see it as a path to build what's in your dreams. The dreams of a kid who played a game that made a difference in their life of one day making their own games. I want to enable that for everyone who wants it. Because it took me FAR too long to get into the gaming industry. People see it as complicated and difficult and competitive. I want to lower those barriers as my legacy. 🤓🤓🤓 --- My new company, Creative Force, has a goal of "Building games at the speed of thought". I'm designing a general system to make games faster that could be used in any game engine. A design language that can be read by an engine that will "build your game for you". Basically allowing game designers to make "recipes" for games, and let other people use those recipes anywhere, and modify them easily to experiment and have a creative outlet. Maybe one day be a game chef and make a living doing what they dreamed, without needing crazy technical experience. 👩‍🍳👩‍🍳👩‍🍳 I believe this is possible with modern learnings from software that haven't been used in game development fully yet. A game architecture that will take the technical out of making games, and make it as approachable and as deep as writing a novel or painting a masterpiece. This is my dream... and I hope you'll follow along with me. Cuz my year #36, is quite literally a GAME CHANGING year. Love y'all 💖💖💖

RayBenefield

24,418 次观看 • 2 年前

I think I can finally report some success training a quite accurate IDM capable of recovering keystrokes from Minecraft gameplay, even in quite PvP-heavy situations. At this point the model does not only know what keys are pressed to the extent reasonably discernible, it also knows how fast it is moving in 3D space at all times, even when knockback is mixing with the self-move impulse. Now, recovering keystrokes from normal external capture footage is just about impossible. E.g. W/A/S/D does exactly nothing during partial tick frames and jumping mid-air is also equally useless, so asking the model to recover key down states is inherently unreasoanble. Mouse deltas are also completely arbitrary units, as game mouse sensitivity introduces an arbitrary scale factor into the equation. The only good option is to think carefully about your model-environment contract, and only record "logical actions", not raw keystrokes. So here's a few unfortunate lessons I had to learn in roughly this order. - Choose good units. (bad: mouse deltas, good: delta radians [yes, you will need game-internal state]) - Capture from inside the main game loop and read the game fbo to get consistent frame-action pairing. Doing post-mortem pairing is hopeless. - Carefully define when you think keystrokes actually have an effect. (jump only works on ground, when flying or in water etc.) More subtle: The key may already be down, but no tick has happened yet to actually use the value. Hence: ignore Seperate gamestate into "fast and slow-moving" components. E.g. movement is likely tick based, camera rotation is very likely updated every frame in essentially every game ever. - Think about your frame-action correspondance contract (How old is the frame in relation to the inputs you capture? Will double or tripple buffering affect you?) Think about the game loop timeline, where you are sampling, how old the data you are reading is, and where the ticks are happening around you. Language models used to simply not have a model-environment contract, but even now with the model "living" in a designated harness, the contract still boils down to formatting, and tool implementation intrinsics. While also important, it is still quite a bit more obvious because the violations are in some way shape or form reflected as text you can actually see. - ffmpeg dropping frames cummulatively screws the model the further you get into the sequence because your targets are now shifted. If you can't encode the video in real-time, too bad. - Sodium has a frames in flight system different from vanilla Minecraft, which will also offset your targets from your frames. (there goes that data...) - Models are succeptible to latency. If there is too big of a delay between action and on-screen reflection, your performance degrades. At this point I realize ~100hours of gameplay is essentially no longer usable as a dataset. You can train on this data, but all you'll get is a mushy mess. However, some good news: - Making the model predict physics gamestate scalars helps the model generalize. For instantaneous events like jump, it's unreasonable to ask the model emit a short burst of jump=true at exactly the right time, however if you also predict your current y-velocity, the model has supervision signal for the "latent" from which that onground jump becomes apparent. Recovering x/z motion is also somewhat easier than unmixing it into plausible keystrokes for inertia-heavy player controller logic. - Regressing physics gamestate scalars also seems to make your dataset "bigger". While pure keystroke classification will overfit quickly, predicting exact physics gamestate scalars forces the model to generalize more and you can tolerate far more epochs before validation loss starts to stall out. This is the only reason why it was bearable to dump 100h+ of dataset hours and replace it with ~3 hours of gameplay after the 4th revision of the file format (yeah...) and somehow still have better performance. Now, you might be asking, "isn't this brittle?" and the answer is yesn't. Frame-action correspondance matters for training, but not so much during inference. So as long as you are sampling in roughly the same interval as your training data, you aren't violating any hard contract per-se. Somewhere around the frames ticks are happening, and during training you capture various tick-capture offset relations per random chance, so nothing is too obviously wrong here. HOWEVER, you will get screwed by gui scale, shaders, resource packs, "shit that recording is 1920x1040 because somebody doesn't know fullscreen exists" and other unfortunate edge cases of reality. But I suppose this is the role of dataset size. If all those "contract violations" that a youtube video has compared to the training data are addressed, I think this is a way to turn Youtube into a labeled dataset. I could never shake the feeling that VPT is a sound idea in practice, while never having been properly executed, and I think one reason why it hasn't is because that label boostrapping part is just a pain in the butt to get right. Now, what the player is doing is of course not the only label you can extract from video, but it has to be one of the targets predicted during pretraining to "align" the pretraining objective. Some notes on the video here, the colored dots on the analog visualizer are the ground truth, while the gray dot is the model prediction. Green means correct prediction, red means incorrect prediction at that frame. Model P(key) reports how wrong the prediction is from green (0.0) to red (1.0). You will also notice that during periods of rapid slow down, left and right actions become close to irrecoverable, because there is just that little motion. And some jump actions are not predicted correctly because I got the detection condition for jump events wrong... (duh) LMB/RMB for other than sustained events (like item-consume and block break) also seem to be hopelessly irrecoverable for now. Swing was supposed to do the same thing as motion y did for jump, but its too well behaved as an increasing counter. Maybe partial-tick interpolated values work better (v5 file format then... ugh..)
2:27

I think I can finally report some success training a quite accurate IDM capable of recovering keystrokes from Minecraft gameplay, even in quite PvP-heavy situations. At this point the model does not only know what keys are pressed to the extent reasonably discernible, it also knows how fast it is moving in 3D space at all times, even when knockback is mixing with the self-move impulse. Now, recovering keystrokes from normal external capture footage is just about impossible. E.g. W/A/S/D does exactly nothing during partial tick frames and jumping mid-air is also equally useless, so asking the model to recover key down states is inherently unreasoanble. Mouse deltas are also completely arbitrary units, as game mouse sensitivity introduces an arbitrary scale factor into the equation. The only good option is to think carefully about your model-environment contract, and only record "logical actions", not raw keystrokes. So here's a few unfortunate lessons I had to learn in roughly this order. - Choose good units. (bad: mouse deltas, good: delta radians [yes, you will need game-internal state]) - Capture from inside the main game loop and read the game fbo to get consistent frame-action pairing. Doing post-mortem pairing is hopeless. - Carefully define when you think keystrokes actually have an effect. (jump only works on ground, when flying or in water etc.) More subtle: The key may already be down, but no tick has happened yet to actually use the value. Hence: ignore Seperate gamestate into "fast and slow-moving" components. E.g. movement is likely tick based, camera rotation is very likely updated every frame in essentially every game ever. - Think about your frame-action correspondance contract (How old is the frame in relation to the inputs you capture? Will double or tripple buffering affect you?) Think about the game loop timeline, where you are sampling, how old the data you are reading is, and where the ticks are happening around you. Language models used to simply not have a model-environment contract, but even now with the model "living" in a designated harness, the contract still boils down to formatting, and tool implementation intrinsics. While also important, it is still quite a bit more obvious because the violations are in some way shape or form reflected as text you can actually see. - ffmpeg dropping frames cummulatively screws the model the further you get into the sequence because your targets are now shifted. If you can't encode the video in real-time, too bad. - Sodium has a frames in flight system different from vanilla Minecraft, which will also offset your targets from your frames. (there goes that data...) - Models are succeptible to latency. If there is too big of a delay between action and on-screen reflection, your performance degrades. At this point I realize ~100hours of gameplay is essentially no longer usable as a dataset. You can train on this data, but all you'll get is a mushy mess. However, some good news: - Making the model predict physics gamestate scalars helps the model generalize. For instantaneous events like jump, it's unreasonable to ask the model emit a short burst of jump=true at exactly the right time, however if you also predict your current y-velocity, the model has supervision signal for the "latent" from which that onground jump becomes apparent. Recovering x/z motion is also somewhat easier than unmixing it into plausible keystrokes for inertia-heavy player controller logic. - Regressing physics gamestate scalars also seems to make your dataset "bigger". While pure keystroke classification will overfit quickly, predicting exact physics gamestate scalars forces the model to generalize more and you can tolerate far more epochs before validation loss starts to stall out. This is the only reason why it was bearable to dump 100h+ of dataset hours and replace it with ~3 hours of gameplay after the 4th revision of the file format (yeah...) and somehow still have better performance. Now, you might be asking, "isn't this brittle?" and the answer is yesn't. Frame-action correspondance matters for training, but not so much during inference. So as long as you are sampling in roughly the same interval as your training data, you aren't violating any hard contract per-se. Somewhere around the frames ticks are happening, and during training you capture various tick-capture offset relations per random chance, so nothing is too obviously wrong here. HOWEVER, you will get screwed by gui scale, shaders, resource packs, "shit that recording is 1920x1040 because somebody doesn't know fullscreen exists" and other unfortunate edge cases of reality. But I suppose this is the role of dataset size. If all those "contract violations" that a youtube video has compared to the training data are addressed, I think this is a way to turn Youtube into a labeled dataset. I could never shake the feeling that VPT is a sound idea in practice, while never having been properly executed, and I think one reason why it hasn't is because that label boostrapping part is just a pain in the butt to get right. Now, what the player is doing is of course not the only label you can extract from video, but it has to be one of the targets predicted during pretraining to "align" the pretraining objective. Some notes on the video here, the colored dots on the analog visualizer are the ground truth, while the gray dot is the model prediction. Green means correct prediction, red means incorrect prediction at that frame. Model P(key) reports how wrong the prediction is from green (0.0) to red (1.0). You will also notice that during periods of rapid slow down, left and right actions become close to irrecoverable, because there is just that little motion. And some jump actions are not predicted correctly because I got the detection condition for jump events wrong... (duh) LMB/RMB for other than sustained events (like item-consume and block break) also seem to be hopelessly irrecoverable for now. Swing was supposed to do the same thing as motion y did for jump, but its too well behaved as an increasing counter. Maybe partial-tick interpolated values work better (v5 file format then... ugh..)

mike64_t

18,762 次观看 • 3 个月前

Grateful for the outpouring the past few days. I can’t tell you how motivating it is to have people care so much about Arc. The encouragement, the criticism, the confusion. Took it all in <3 As a thank you, I’d like to speak more plainly about what’s happening and why – we owe it to you: 1. Every person who joined or invested in our company did so to build products beloved by hundreds of millions of people. “We want to be to the browser what the iPhone was to the cellphone” has been our rallying cry since Day One. We knew chance of success was low but the ambition made us leap out of bed every morning. 2. Arc is beloved, popular, and growing (4x daily actives YoY). But it’s now clear that what most people love about the product *is also* what will prevent it from reaching hundreds of millions of people in our target demographic (people who spend hours in their browsers each day for their livelihood). Arc is a niche browser, even if we did not intend for it to be so. 3. Luckily, we architected this company – from company name to investors and technical architecture – to support multiple products since Day One. See Arc Search. Our favorite brands have multiple product lines in the same category too (Apple, Nike, Disney). Hence our realization: why keep trying to make Arc something it is not? Nobody who loved Arc wanted Arc Max. Arc members just want it to be more stable, secure, and performant. “Let’s just do that!” 4. With Arc as our beloved but niche browser #1, we asked ourselves a simple question: if we founded the company TODAY (in 2024), with everything that we know, what would the browser of the future look like for hundreds of millions of people? Let’s go build *that* product, alongside Arc. A second browser that is easier to use, more focused, and more powerful. All in order to live up to our founding mission (#1 above). 5. Yet none of this would’ve happened if it weren’t for the timing (market timing is most underrated startup ingredient). Mark my words: the Web is going to dramatically change in 2025 – much more than we all appreciate. Crazy new AI & computer-use models are incoming. I promise you that new browsers will be the story of 2025 (The Browser Company aside). Why? The browser layer is the obvious epicenter of AI & Agents because of its unique context, cookies, & apps. 6. To build a breakthrough consumer product (#1) – like truly breakthrough – you need a catalyzing innovation or technology. AI will be that for the next era of browsers, whether we win or someone else does. So why us (other than Arc is niche)? Our belief is that not only do you need the browser layer to win, but “the hard part” is nailing the interface, the interactions, the storytelling. That’s our bread & butter. That’s the expertise of our team. Now you can see how these puzzle pieces fit together… We built something people love (in Arc) and we intend to stick by it. But we also won’t lose sight of why our team poured so much blood, sweat and tears over the past 4 years into this company: the mission to build a new interface to the internet used by hundreds of millions of people every day. It truly feels like the moment we were waiting for is here and we won’t miss it. Everything we’ve done up until this point was for this type of window, even if we couldn’t have predicted it would play out exactly this way. We’re hopeful that more of you will understand why we’re building this second product soon. I feel confident you will once we can show you more of what we’re dreaming up, and once more of the things we’ve heard and seen in the industry reveal themselves soon. Candidly, we wanted to wait on this announcement but random stuff was leaking and it seemed wrong for you to hear from anyone except us first. We’ve always been at our best when we’re open & honest. We’ll continue to be. Finally, THANK YOU, again, for the love & tough love. We don’t take it for granted. We can’t wait to ship, ship, ship in early 2025!
0:57

Grateful for the outpouring the past few days. I can’t tell you how motivating it is to have people care so much about Arc. The encouragement, the criticism, the confusion. Took it all in <3 As a thank you, I’d like to speak more plainly about what’s happening and why – we owe it to you: 1. Every person who joined or invested in our company did so to build products beloved by hundreds of millions of people. “We want to be to the browser what the iPhone was to the cellphone” has been our rallying cry since Day One. We knew chance of success was low but the ambition made us leap out of bed every morning. 2. Arc is beloved, popular, and growing (4x daily actives YoY). But it’s now clear that what most people love about the product *is also* what will prevent it from reaching hundreds of millions of people in our target demographic (people who spend hours in their browsers each day for their livelihood). Arc is a niche browser, even if we did not intend for it to be so. 3. Luckily, we architected this company – from company name to investors and technical architecture – to support multiple products since Day One. See Arc Search. Our favorite brands have multiple product lines in the same category too (Apple, Nike, Disney). Hence our realization: why keep trying to make Arc something it is not? Nobody who loved Arc wanted Arc Max. Arc members just want it to be more stable, secure, and performant. “Let’s just do that!” 4. With Arc as our beloved but niche browser #1, we asked ourselves a simple question: if we founded the company TODAY (in 2024), with everything that we know, what would the browser of the future look like for hundreds of millions of people? Let’s go build *that* product, alongside Arc. A second browser that is easier to use, more focused, and more powerful. All in order to live up to our founding mission (#1 above). 5. Yet none of this would’ve happened if it weren’t for the timing (market timing is most underrated startup ingredient). Mark my words: the Web is going to dramatically change in 2025 – much more than we all appreciate. Crazy new AI & computer-use models are incoming. I promise you that new browsers will be the story of 2025 (The Browser Company aside). Why? The browser layer is the obvious epicenter of AI & Agents because of its unique context, cookies, & apps. 6. To build a breakthrough consumer product (#1) – like truly breakthrough – you need a catalyzing innovation or technology. AI will be that for the next era of browsers, whether we win or someone else does. So why us (other than Arc is niche)? Our belief is that not only do you need the browser layer to win, but “the hard part” is nailing the interface, the interactions, the storytelling. That’s our bread & butter. That’s the expertise of our team. Now you can see how these puzzle pieces fit together… We built something people love (in Arc) and we intend to stick by it. But we also won’t lose sight of why our team poured so much blood, sweat and tears over the past 4 years into this company: the mission to build a new interface to the internet used by hundreds of millions of people every day. It truly feels like the moment we were waiting for is here and we won’t miss it. Everything we’ve done up until this point was for this type of window, even if we couldn’t have predicted it would play out exactly this way. We’re hopeful that more of you will understand why we’re building this second product soon. I feel confident you will once we can show you more of what we’re dreaming up, and once more of the things we’ve heard and seen in the industry reveal themselves soon. Candidly, we wanted to wait on this announcement but random stuff was leaking and it seemed wrong for you to hear from anyone except us first. We’ve always been at our best when we’re open & honest. We’ll continue to be. Finally, THANK YOU, again, for the love & tough love. We don’t take it for granted. We can’t wait to ship, ship, ship in early 2025!

Josh Miller

189,025 次观看 • 1 年前

$AMD $620/share is too conservative for 2026 🧵 Some quick facts before I dive into this super long thread: $META allocated 42% GPUs to $AMD and 58% to $NVDA OpenAI allocated 6GW(38%) to $AMD and 10GW to $NVDA My $620 PT below by end of 2026 was only for 10-15% market share. I believe $AMD is going to have much much higher market share than I projected. The AI accelerator market is exploding, projected to reach $500 billion by 2028(is now heading $1Tril), driven by insatiable demand for training and inference compute in large language models (LLMs), recommendation systems, and autonomous systems. Nvidia ($NVDA) has long held a stranglehold, commanding over 90% market share through its CUDA ecosystem and superior rack-scale solutions. However, AMD is mounting a formidable challenge, leveraging cost advantages, open-source software momentum, and hyperscaler partnerships to erode Nvidia's moat. Recent deals—such as Meta's ($META) allocation of 42% of its GPU capacity to AMD and OpenAI's commitment to 6GW of AMD compute (versus 10GW for Nvidia)—signal a tipping point. At the forefront is AMD's Instinct MI450 series, a next-generation AI GPU slated for H2 2026 launch, which promises "no-excuses" leadership in training, inference, and distributed workloads. This analysis dissects how AMD will capture more market share and why hyperscalers like $Meta , xAI , Oracle , and others are poised to become voracious buyers of the MI450. AMD's AI GPU revenue has surged from negligible levels in 2022 to an estimated $4-5 billion in 2025, capturing ~6% of the data center GPU market. This growth stems from the Instinct MI300X, which offers 141GB of HBM3 memory and competitive FP8/FP16 performance at 20-30% lower cost than Nvidia's H100. Hyperscalers, facing NVIDIA 's overcharging, have turned to AMD for diversification. Meta, for instance, plans 600,000 H100-equivalent GPUs by end-2024, with ~42% (or 250,000+ units) sourced from AMD's MI300 series for inference tasks like image editing and AI assistants. Similarly, OpenAI's recent multi-year deal commits to 6GW of AMD compute—equivalent to ~300,000-400,000 MI450 GPUs—starting with 1GW in 2026, explicitly to counterbalance its 10GW Nvidia allocation. These aren't one-offs. Microsoft Azure, Amazon AWS, and Oracle Cloud Infrastructure (OCI) have integrated MI300X for AI workloads, with Oracle deploying 30,000 MI355X units in zettascale clusters. xAI, Elon Musk Musk's AI venture, ran 30% of Grok-1's production traffic on MI300X GPUs and has confirmed ongoing purchases. Collectively, these partners represent over $400 billion in projected AI infrastructure spend through 2028, with AMD targeting up to 40% market share. For those that subscribed, I wrote a specific thread on how AMD "secret weapon" is going to change the game in 2026 with an improved designs on all its products, yes AMD has patent on it. Software is the linchpin. AMD's ROCm platform, once derided as "half-baked," now supports day-zero integration for Llama-4, DeepSeek V3, and GPT-OSS models—closing the CUDA gap. Benchmarks show MI355X (MI450 precursor) outperforming Nvidia's B200 in inference by 1.5-2x on memory-bound tasks, at 25-35% lower TCO. For training, MI450's rack-scale IF128 configuration (128 GPUs, 1.4 PB/s intra-rack bandwidth) rivals Nvidia's VR200 NVL144, enabling clusters like xAI's Colossus (scaling to 1M GPUs). My below thread projected Etimated conservative FY 25 revenue: $34-$36B Estimated conservative FY 26 revenue: $55B-$62B Below is why $AMD is revenue is going to be much higher after OpenAI deal. 1. OpenAI 1GW in 2026. With high demand for MI355X at $30,000k+ per unit, with MI450 is likely to be sold in the $45k-$55k. We can safely calcuate 1GW would require roughly 400,000 MI450 GPUs. or Roughly ~$20B revenue in 2026 alone from OpenAI. That would mean $AMD would hit $56B just from one partnership(OpenAI) in 2026 2. $META, the biggest spender on AI Infrastructure right now, Daddy Zuckerberg bought 250,000+ MI300, and is buying MI355X for recommendation engines and Llama training. It is very unlikely for Daddy Zuck to slow down AMD Chips, due to its Inference superiority to NVDA Chips. Most likely we will see at least 300,000-400,000 MI355X ordered from now toward end of H1 2025. And another 300,000-500,000 MI450 by H2 2025. Or ~$20B from just Meta in H2 alone, excluded H1. 3. xAI : Musk confirmed "AMD GPUs work very well" for Grok's small/medium models, with 30% of Grok-1 on MI300X. xAI's Colossus (200K+ GPUs, targeting 1M) and Oracle partnership (via OCI's MI355X cluster) position it for MI450 trials in H1 2026. With $6B funding and Grok integration into Oracle services, xAI could allocate 10-20% ($10B-$15B) to MI450 for distributed inference. We haven't heard the detail from Daddy Elon Musk yet, but most likely not going to be spending less than OpenAI or Sam Altman 4. Oracle ($ORCL): A multi-billion-dollar MI355X deal powers OCI's AI superclusters, with $500B+ remaining performance obligations. Larry Ellison's zettascale ambitions and xAI/OpenAI integrations make Oracle a MI450 anchor tenant—projected 50-100k units ($15B+ spend) for enterprise AI platforms. $ORCL is likely to spend more on the new "secret weapon" due to its capability in AI inference and cost advantage for $500B backlog. 5. Others ( Microsoft , Amazon , Saudi+other countries): Microsoft (Azure MI300X for training) and Amazon ($148B 15-year spend) test MI450 via Stargate ($500B with Oracle/SoftBank). Emerging buyers like G42 (5GW UAE campus), Crusoe, and Hot Aisle add 5-10GW demand. These potentially would add $15B-$30B in 2026 alone. We also need to factor in $TSM supply constraint( $NVDA is TSMC favorite), so $AMD market cap/growth is being tamed by TSMC. So what are you saying Mike, well $AMD 2026 revenue could hit $90-$100B by end of 2026 or nearly 185% growth YoYo. So what does that mean for valuation? I have no idea how Mr. Market gonna value AMD in 2026 with 3 digits growth. My Conservative $620 was my best projection until today with OpenAI partnership. I'm telling you as one of the biggest AMD bull, that I will leave it to "smart money" and other investors to do the price discovery while I'm chilling and writing DDs daily. Lastly, AMD's MI450 isn't hype—it's a calibrated strike at Nvidia's vulnerabilities, amplified by hyperscaler bets like Meta's 42% allocation and OpenAI's 6GW lifeline. By prioritizing inference efficiency, rack-scale innovation, and open ecosystems, AMD will siphon 10-15% share in 2026, scaling to 20%+ as TCO trumps CUDA loyalty. Meta, xAI, Oracle et al. aren't passive; they're active co-designers, betting billions on MI450 to fuel AGI pursuits without Nvidia's premium. For investors, this is AMD's inflection Per Dr. Lisa Su Not Financial Advice!
1:06

$AMD $620/share is too conservative for 2026 🧵 Some quick facts before I dive into this super long thread: $META allocated 42% GPUs to $AMD and 58% to $NVDA OpenAI allocated 6GW(38%) to $AMD and 10GW to $NVDA My $620 PT below by end of 2026 was only for 10-15% market share. I believe $AMD is going to have much much higher market share than I projected. The AI accelerator market is exploding, projected to reach $500 billion by 2028(is now heading $1Tril), driven by insatiable demand for training and inference compute in large language models (LLMs), recommendation systems, and autonomous systems. Nvidia ($NVDA) has long held a stranglehold, commanding over 90% market share through its CUDA ecosystem and superior rack-scale solutions. However, AMD is mounting a formidable challenge, leveraging cost advantages, open-source software momentum, and hyperscaler partnerships to erode Nvidia's moat. Recent deals—such as Meta's ($META) allocation of 42% of its GPU capacity to AMD and OpenAI's commitment to 6GW of AMD compute (versus 10GW for Nvidia)—signal a tipping point. At the forefront is AMD's Instinct MI450 series, a next-generation AI GPU slated for H2 2026 launch, which promises "no-excuses" leadership in training, inference, and distributed workloads. This analysis dissects how AMD will capture more market share and why hyperscalers like $Meta , xAI , Oracle , and others are poised to become voracious buyers of the MI450. AMD's AI GPU revenue has surged from negligible levels in 2022 to an estimated $4-5 billion in 2025, capturing ~6% of the data center GPU market. This growth stems from the Instinct MI300X, which offers 141GB of HBM3 memory and competitive FP8/FP16 performance at 20-30% lower cost than Nvidia's H100. Hyperscalers, facing NVIDIA 's overcharging, have turned to AMD for diversification. Meta, for instance, plans 600,000 H100-equivalent GPUs by end-2024, with ~42% (or 250,000+ units) sourced from AMD's MI300 series for inference tasks like image editing and AI assistants. Similarly, OpenAI's recent multi-year deal commits to 6GW of AMD compute—equivalent to ~300,000-400,000 MI450 GPUs—starting with 1GW in 2026, explicitly to counterbalance its 10GW Nvidia allocation. These aren't one-offs. Microsoft Azure, Amazon AWS, and Oracle Cloud Infrastructure (OCI) have integrated MI300X for AI workloads, with Oracle deploying 30,000 MI355X units in zettascale clusters. xAI, Elon Musk Musk's AI venture, ran 30% of Grok-1's production traffic on MI300X GPUs and has confirmed ongoing purchases. Collectively, these partners represent over $400 billion in projected AI infrastructure spend through 2028, with AMD targeting up to 40% market share. For those that subscribed, I wrote a specific thread on how AMD "secret weapon" is going to change the game in 2026 with an improved designs on all its products, yes AMD has patent on it. Software is the linchpin. AMD's ROCm platform, once derided as "half-baked," now supports day-zero integration for Llama-4, DeepSeek V3, and GPT-OSS models—closing the CUDA gap. Benchmarks show MI355X (MI450 precursor) outperforming Nvidia's B200 in inference by 1.5-2x on memory-bound tasks, at 25-35% lower TCO. For training, MI450's rack-scale IF128 configuration (128 GPUs, 1.4 PB/s intra-rack bandwidth) rivals Nvidia's VR200 NVL144, enabling clusters like xAI's Colossus (scaling to 1M GPUs). My below thread projected Etimated conservative FY 25 revenue: $34-$36B Estimated conservative FY 26 revenue: $55B-$62B Below is why $AMD is revenue is going to be much higher after OpenAI deal. 1. OpenAI 1GW in 2026. With high demand for MI355X at $30,000k+ per unit, with MI450 is likely to be sold in the $45k-$55k. We can safely calcuate 1GW would require roughly 400,000 MI450 GPUs. or Roughly ~$20B revenue in 2026 alone from OpenAI. That would mean $AMD would hit $56B just from one partnership(OpenAI) in 2026 2. $META, the biggest spender on AI Infrastructure right now, Daddy Zuckerberg bought 250,000+ MI300, and is buying MI355X for recommendation engines and Llama training. It is very unlikely for Daddy Zuck to slow down AMD Chips, due to its Inference superiority to NVDA Chips. Most likely we will see at least 300,000-400,000 MI355X ordered from now toward end of H1 2025. And another 300,000-500,000 MI450 by H2 2025. Or ~$20B from just Meta in H2 alone, excluded H1. 3. xAI : Musk confirmed "AMD GPUs work very well" for Grok's small/medium models, with 30% of Grok-1 on MI300X. xAI's Colossus (200K+ GPUs, targeting 1M) and Oracle partnership (via OCI's MI355X cluster) position it for MI450 trials in H1 2026. With $6B funding and Grok integration into Oracle services, xAI could allocate 10-20% ($10B-$15B) to MI450 for distributed inference. We haven't heard the detail from Daddy Elon Musk yet, but most likely not going to be spending less than OpenAI or Sam Altman 4. Oracle ($ORCL): A multi-billion-dollar MI355X deal powers OCI's AI superclusters, with $500B+ remaining performance obligations. Larry Ellison's zettascale ambitions and xAI/OpenAI integrations make Oracle a MI450 anchor tenant—projected 50-100k units ($15B+ spend) for enterprise AI platforms. $ORCL is likely to spend more on the new "secret weapon" due to its capability in AI inference and cost advantage for $500B backlog. 5. Others ( Microsoft , Amazon , Saudi+other countries): Microsoft (Azure MI300X for training) and Amazon ($148B 15-year spend) test MI450 via Stargate ($500B with Oracle/SoftBank). Emerging buyers like G42 (5GW UAE campus), Crusoe, and Hot Aisle add 5-10GW demand. These potentially would add $15B-$30B in 2026 alone. We also need to factor in $TSM supply constraint( $NVDA is TSMC favorite), so $AMD market cap/growth is being tamed by TSMC. So what are you saying Mike, well $AMD 2026 revenue could hit $90-$100B by end of 2026 or nearly 185% growth YoYo. So what does that mean for valuation? I have no idea how Mr. Market gonna value AMD in 2026 with 3 digits growth. My Conservative $620 was my best projection until today with OpenAI partnership. I'm telling you as one of the biggest AMD bull, that I will leave it to "smart money" and other investors to do the price discovery while I'm chilling and writing DDs daily. Lastly, AMD's MI450 isn't hype—it's a calibrated strike at Nvidia's vulnerabilities, amplified by hyperscaler bets like Meta's 42% allocation and OpenAI's 6GW lifeline. By prioritizing inference efficiency, rack-scale innovation, and open ecosystems, AMD will siphon 10-15% share in 2026, scaling to 20%+ as TCO trumps CUDA loyalty. Meta, xAI, Oracle et al. aren't passive; they're active co-designers, betting billions on MI450 to fuel AGI pursuits without Nvidia's premium. For investors, this is AMD's inflection Per Dr. Lisa Su Not Financial Advice!

Mike

711,006 次观看 • 9 个月前