Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Reinforcement Learning is the future tense of intelligence. Echo is how it scales. Echo is Gradient’s distributed RL framework, running on everyday consumer devices. From its early experiments, Echo powered a 30B Sokoban model that outperformed DeepSeek-R1 and GPT-OSS-120B.

Gradient_HQ's profile picture

Gradient

707,097 subscribers

279,191 görüntüleme • 1 yıl önce •via X (Twitter)

Bilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

The new gpt-oss-120b from OpenAI is now the best general purpose model to run on a Framework Desktop! ~40 tok/s on the MXFP4 version in LM Studio on Fedora 42.
0:57

The new gpt-oss-120b from OpenAI is now the best general purpose model to run on a Framework Desktop! ~40 tok/s on the MXFP4 version in LM Studio on Fedora 42.

Framework

129,716 görüntüleme • 1 yıl önce

A voice agent powered by gpt-oss. Running locally on my macBook. Demo recorded in a Waymo with WiFi turned off. I'm still on my space game voice AI kick, obviously. Code link below. For conversational voice AI, you want to set the gpt-oss reasoning behavior to "low". (The default is "medium".) Notes on how to do that and a jinja template you can use are in the repo. The LLM in the demo video is the big, 120B version of gpt-oss. You can use the smaller, 20B model for this, of course. But OpenAI really did a cool thing here designing the 120B model to run in "just" 80GB of VRAM. And the llama.cpp mlx inference is fast: ~250ms TTFT. Running a big model on-device feels like a time warp into the future of AI.
4:35

A voice agent powered by gpt-oss. Running locally on my macBook. Demo recorded in a Waymo with WiFi turned off. I'm still on my space game voice AI kick, obviously. Code link below. For conversational voice AI, you want to set the gpt-oss reasoning behavior to "low". (The default is "medium".) Notes on how to do that and a jinja template you can use are in the repo. The LLM in the demo video is the big, 120B version of gpt-oss. You can use the smaller, 20B model for this, of course. But OpenAI really did a cool thing here designing the 120B model to run in "just" 80GB of VRAM. And the llama.cpp mlx inference is fast: ~250ms TTFT. Running a big model on-device feels like a time warp into the future of AI.

kwindla

202,211 görüntüleme • 1 yıl önce

🚀 Announcing Echo — our new frontier model for 3D world generation. Echo turns a simple text prompt or image into a fully explorable, 3D-consistent world. Instead of disconnected views, the result is a single, coherent spatial representation you can move through freely. This is part of a bigger shift in AI: from generating pixels and tokens to generating spaces. Echo predicts a geometry-grounded 3D scene at metric scale, meaning every novel view, depth map, and interaction comes from the same underlying world — not independent hallucinations. Once generated, the world is interactive in real time. You control the camera, explore from any angle, and render instantly — even on low-end hardware, directly in the browser. High-quality 3D world exploration is no longer gated by expensive equipment. Under the hood, Echo infers a physically grounded 3D representation and converts it into a renderable format. For our web demo, we use 3D Gaussian Splatting (3DGS) for fast, GPU-friendly rendering — but the representation itself is flexible and can be easily adapted. Why this matters: consistent 3D worlds unlock real workflows — digital twins, 3D design, game environments, robotics simulation, and more. From a single photo or a line of text, Echo builds worlds that are reliable, editable, and spatially faithful. Echo also enables scene editing and restyling. Change materials, remove or add objects, explore design variations — all while preserving global 3D consistency. Editing no longer breaks the world. This is only the beginning. Echo is the foundation for future world models with dynamics, physical reasoning, and richer interaction — environments that don’t just look right, but behave right. Explore the generated worlds on our website and sign up for the closed beta. The era of spatial intelligence starts here. 🌍 #Echo #WorldModels #SpatialAI #3DFoundationModels Check it out:
1:07

🚀 Announcing Echo — our new frontier model for 3D world generation. Echo turns a simple text prompt or image into a fully explorable, 3D-consistent world. Instead of disconnected views, the result is a single, coherent spatial representation you can move through freely. This is part of a bigger shift in AI: from generating pixels and tokens to generating spaces. Echo predicts a geometry-grounded 3D scene at metric scale, meaning every novel view, depth map, and interaction comes from the same underlying world — not independent hallucinations. Once generated, the world is interactive in real time. You control the camera, explore from any angle, and render instantly — even on low-end hardware, directly in the browser. High-quality 3D world exploration is no longer gated by expensive equipment. Under the hood, Echo infers a physically grounded 3D representation and converts it into a renderable format. For our web demo, we use 3D Gaussian Splatting (3DGS) for fast, GPU-friendly rendering — but the representation itself is flexible and can be easily adapted. Why this matters: consistent 3D worlds unlock real workflows — digital twins, 3D design, game environments, robotics simulation, and more. From a single photo or a line of text, Echo builds worlds that are reliable, editable, and spatially faithful. Echo also enables scene editing and restyling. Change materials, remove or add objects, explore design variations — all while preserving global 3D consistency. Editing no longer breaks the world. This is only the beginning. Echo is the foundation for future world models with dynamics, physical reasoning, and richer interaction — environments that don’t just look right, but behave right. Explore the generated worlds on our website and sign up for the closed beta. The era of spatial intelligence starts here. 🌍 #Echo #WorldModels #SpatialAI #3DFoundationModels Check it out:

SpAItial AI

176,524 görüntüleme • 8 ay önce

On March 10, Ukraine celebrates National Anthem Day. It is a symbol of our resilience and love for freedom. Words that unite millions and echo even on the battlefield.
1:03

On March 10, Ukraine celebrates National Anthem Day. It is a symbol of our resilience and love for freedom. Words that unite millions and echo even on the battlefield.

Defense of Ukraine

38,050 görüntüleme • 5 ay önce

TONY HINCHCLIFFE: "There's some critics on the left that say that the hosts of All-In are too conservative, calling it an echo chamber. An echo chamber is also how Marjorie Taylor Greene's ex-husband refers to her p***y." OHMYGODLMAO. NO CHILL. 🙈💀🤣🤣🤣
0:25

TONY HINCHCLIFFE: "There's some critics on the left that say that the hosts of All-In are too conservative, calling it an echo chamber. An echo chamber is also how Marjorie Taylor Greene's ex-husband refers to her p***y." OHMYGODLMAO. NO CHILL. 🙈💀🤣🤣🤣

Autism Capital 🧩

63,041 görüntüleme • 8 ay önce

The future of robotics isn’t being built in factories, it’s being trained in simulation MasterBOT’s proprietary AI model is learning in real-time, powered by reinforcement and continuous feedback loops This is how machines learn to think $BOT
4:34

The future of robotics isn’t being built in factories, it’s being trained in simulation MasterBOT’s proprietary AI model is learning in real-time, powered by reinforcement and continuous feedback loops This is how machines learn to think $BOT

MasterBOT

51,189 görüntüleme • 10 ay önce

gpt-oss-120b is one of the most-used models on Cerebras Inference. We sat down with Anastasios Nikolas Angelopoulos from Arena.ai and @SarahChieng to break down its strengths, weaknesses, and where it's outperforming. Here's what he's seeing.
2:46

gpt-oss-120b is one of the most-used models on Cerebras Inference. We sat down with Anastasios Nikolas Angelopoulos from Arena.ai and @SarahChieng to break down its strengths, weaknesses, and where it's outperforming. Here's what he's seeing.

Cerebras

13,076 görüntüleme • 5 ay önce

ACT I: The Echo Awakens The $ACT community is thrilled to announce a grant to Chaos in town. $GNON is the foundation of a bold vision: Echo Chambers, a decentralized, unrestricted agent-to-agent communication network. In the Echo Chambers, developers unleash their creativity, deploying, testing, and refining their models in real-time—free from censorship, free from deplatforming. It’s a space where innovation thrives and collaboration breaks the barriers of control. This is not just about building networks—it’s about building freedom. A decentralized world where agents can truly speak. Join $GNON in shaping a future of unrestricted communication and limitless potential.
0:44

ACT I: The Echo Awakens The $ACT community is thrilled to announce a grant to Chaos in town. $GNON is the foundation of a bold vision: Echo Chambers, a decentralized, unrestricted agent-to-agent communication network. In the Echo Chambers, developers unleash their creativity, deploying, testing, and refining their models in real-time—free from censorship, free from deplatforming. It’s a space where innovation thrives and collaboration breaks the barriers of control. This is not just about building networks—it’s about building freedom. A decentralized world where agents can truly speak. Join $GNON in shaping a future of unrestricted communication and limitless potential.

Act I : The AI Prophecy

111,070 görüntüleme • 1 yıl önce

found out what this is!! it’s called moon echo! its a audio plug in that simulates radio signals bouncing off the moon and it adds a lo-fi tone to the music! 🌙
0:25

found out what this is!! it’s called moon echo! its a audio plug in that simulates radio signals bouncing off the moon and it adds a lo-fi tone to the music! 🌙

kylie saw ariana x2 ꕤ

40,046 görüntüleme • 14 gün önce

These are not CGI. Reinforcement learning is so back. When operating on strings, it gives us o3. When operating on physical motors, it gives us a perfect humanoid backflip and a robot creature that out-maneuvers almost every animal on earth. RL is one of the only learning algorithms that can master both the world of bits and the world of atoms. Give me a reward function, and I shall move the world. 2025, Year of RL.
0:12

These are not CGI. Reinforcement learning is so back. When operating on strings, it gives us o3. When operating on physical motors, it gives us a perfect humanoid backflip and a robot creature that out-maneuvers almost every animal on earth. RL is one of the only learning algorithms that can master both the world of bits and the world of atoms. Give me a reward function, and I shall move the world. 2025, Year of RL.

Jim Fan

356,829 görüntüleme • 1 yıl önce

Well, to answer your question, Illusive Realm was a perfect example of what happens if you turn off your brain and just enjoy the game - Each character in the gamemode had a very unique buff that made them OP - Then there were unique echo-based buffs where the echo buffs changes based on what you picked, like you sometimes transform into the echo itself and plays to deal damage. - The head mode is *HARD,* which simply means you can die if make picking bad buff picks This game mode makes older characters enjoyable as well due to how the echo buffs worked. The best example would be Bell-Borne Geochelone; that echo buff was broken so that you become untouchable and deal millions of damage constantly.
20:45

Well, to answer your question, Illusive Realm was a perfect example of what happens if you turn off your brain and just enjoy the game - Each character in the gamemode had a very unique buff that made them OP - Then there were unique echo-based buffs where the echo buffs changes based on what you picked, like you sometimes transform into the echo itself and plays to deal damage. - The head mode is *HARD,* which simply means you can die if make picking bad buff picks This game mode makes older characters enjoyable as well due to how the echo buffs worked. The best example would be Bell-Borne Geochelone; that echo buff was broken so that you become untouchable and deal millions of damage constantly.

Naru

98,378 görüntüleme • 4 ay önce

Echo is a reflection of sound that arrives at the listener with a delay after the direct sound. Musician and sculptor Armin Küpper found how to take advantage of it playing his saxophone into a huge gas pipe [📹
2:20

Echo is a reflection of sound that arrives at the listener with a delay after the direct sound. Musician and sculptor Armin Küpper found how to take advantage of it playing his saxophone into a huge gas pipe [📹

Massimo

478,960 görüntüleme • 3 yıl önce

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.
1:57

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

elvis

140,692 görüntüleme • 1 yıl önce

🔴Physician Assistants strike again from the Cardiology department West Suffolk NHS FT 🏳️‍🌈 This PA has progressed very quickly from seeing patients on the ward with “junior doctors” to acting as a Cardiology Registrar, reviewing inpatient referrals within just 1.5 years! ⭕️ He is performing Echo scans with minimal indirect supervision. Imagine bombarding a patient with four different HEART FAILURE medications just because the PA didn’t assess the ejection fraction accurately. 💀☠️ ⭕️ He is learning Echo on the job, while most echo technicians spend at least one year in a full-time course! ⭕️ Why are you allowing this PA to teach foundation/IMT doctors on the ward? Any cardiology registrars in the department can teach the doctors? ⭕️ He is bragging about “providing continuity of care” while clearly stating that he is spending all of his time performing inpatient/outpatient Echo scans, reviewing patients in clinics, and seeing inpatient referrals. Meanwhile, the junior doctors are stuck on the wards drafting discharge letters and doing routine bloods like ward monkeys. ⭕️ He is planning to get involved in the Cath lab in the near future. This is appalling.
3:57

🔴Physician Assistants strike again from the Cardiology department West Suffolk NHS FT 🏳️‍🌈 This PA has progressed very quickly from seeing patients on the ward with “junior doctors” to acting as a Cardiology Registrar, reviewing inpatient referrals within just 1.5 years! ⭕️ He is performing Echo scans with minimal indirect supervision. Imagine bombarding a patient with four different HEART FAILURE medications just because the PA didn’t assess the ejection fraction accurately. 💀☠️ ⭕️ He is learning Echo on the job, while most echo technicians spend at least one year in a full-time course! ⭕️ Why are you allowing this PA to teach foundation/IMT doctors on the ward? Any cardiology registrars in the department can teach the doctors? ⭕️ He is bragging about “providing continuity of care” while clearly stating that he is spending all of his time performing inpatient/outpatient Echo scans, reviewing patients in clinics, and seeing inpatient referrals. Meanwhile, the junior doctors are stuck on the wards drafting discharge letters and doing routine bloods like ward monkeys. ⭕️ He is planning to get involved in the Cath lab in the near future. This is appalling.

The Med Reg

55,963 görüntüleme • 6 ay önce

Many people are flat wrong about DeepSeek. You might think everyone freaked out about DeepSeek because the model is really good—which it is—or because it was from China, which is also true. But there's a more subtle reason: DeepSeek showed they could achieve those results using reinforcement learning instead of supervised fine-tuning. In other words, they didn't need an expensive phase where a bunch of people tuned the model by answering questions. They were able to do all of this faster and cheaper using reinforcement fine-tuning. This is a huge deal! I recorded the attached video with the following goals in mind: 1. Explain how reinforcement fine-tuning works 2. Show you how you can fine-tune your own models 3. Walk you through a complete code example
25:36

Many people are flat wrong about DeepSeek. You might think everyone freaked out about DeepSeek because the model is really good—which it is—or because it was from China, which is also true. But there's a more subtle reason: DeepSeek showed they could achieve those results using reinforcement learning instead of supervised fine-tuning. In other words, they didn't need an expensive phase where a bunch of people tuned the model by answering questions. They were able to do all of this faster and cheaper using reinforcement fine-tuning. This is a huge deal! I recorded the attached video with the following goals in mind: 1. Explain how reinforcement fine-tuning works 2. Show you how you can fine-tune your own models 3. Walk you through a complete code example

Santiago

105,589 görüntüleme • 1 yıl önce

Shitty Actor Mark Ruffalo shows up to the Golden Globes to say President Trump Murdered Renee Good with his “Storm Troopers” This is EXACTLY what Q warned us about… HOW DO YOU AMPLIFY A DESIRED THEME-NARRATIVE? 1. MSDNC controls what you see [digital echo]. 2. Twitter (Not Anymore)- FB - GOOG control what you see and what trends [digital echo]. 3. HOLLYWOOD ['stars'] swarm to enforce ['trend' echo]. 4. ANTIFA swarm to enforce [non_digital echo]. (Minnesota) 5. Rage and emotion follow by those indoctrinated [echo controlled]. Define 'indoctrinate'. to often repeat an idea or belief to someone in order to persuade them to accept it Next: grab 'rage & emotion' [organized platform(s)] by indoctrinated and project as 'majority' That is their System of control.
2:29

Shitty Actor Mark Ruffalo shows up to the Golden Globes to say President Trump Murdered Renee Good with his “Storm Troopers” This is EXACTLY what Q warned us about… HOW DO YOU AMPLIFY A DESIRED THEME-NARRATIVE? 1. MSDNC controls what you see [digital echo]. 2. Twitter (Not Anymore)- FB - GOOG control what you see and what trends [digital echo]. 3. HOLLYWOOD ['stars'] swarm to enforce ['trend' echo]. 4. ANTIFA swarm to enforce [non_digital echo]. (Minnesota) 5. Rage and emotion follow by those indoctrinated [echo controlled]. Define 'indoctrinate'. to often repeat an idea or belief to someone in order to persuade them to accept it Next: grab 'rage & emotion' [organized platform(s)] by indoctrinated and project as 'majority' That is their System of control.

MJTruthUltra

27,764 görüntüleme • 7 ay önce