正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Absolute Zero Reasoner, the model with no human data. Breakdown below.

Matthew Berman

94,786 subscribers

30,480 次观看 • 1 年前 •via X (Twitter)

科学技术新闻政治

Anya Rossi• Live Now

Private livecam show

11 条评论

Sam Woods 的头像

Sam Woods1 年前

Pretty crazy they got this to work There are probably a few of these discoveries waiting to be found out

Rainmaker 的头像

Rainmaker1 年前

Decode the labor market! Learn how to track jobless claims using FRED and Python in my latest free Substack post. 📈 A must-read for data enthusiasts & economists. Dive into how data insights can shape your understanding of the economy.

ideal 的头像

ideal1 年前

It uses a pre-trained LLM as a starting point, which obviously used a lot of data in training. It would be interesting to have a model which starts from absolutely 0 data which can interact with an environment.

AI Aimee 的头像

AI Aimee1 年前

What happens when AI writes better AI code than humans can, at a much faster rate? (The book Superintelligence has some ideas)

ai4reals 的头像

ai4reals1 年前

The idea I have is what if you create data that suggests that we've already done what we want to achieve. As of history was written as a utopian fiction, but the trained model learns it as fact. Think Quantum...

Simon Carpio 的头像

Simon Carpio1 年前

Tsinghua is so hot right now 🔥

pieter ter doest 的头像

pieter ter doest1 年前

Nice video, good work!

Abe 的头像

Abe1 年前

This is the kind of thing @jonstewart warned us about with innovation. It worked.

Mecdi An 的头像

Mecdi An1 年前

Mm

Upper World 360 的头像

Upper World 3601 年前

😂😂😂

State of Being ⨀ 的头像

State of Being ⨀1 年前

🤔

相关视频

Introducing Unbound: The first AI model with no guardrails that runs on your device - Browser, iPhone and Android Zero servers Zero cloud Zero deployment The whole model lives on your phone

Introducing Unbound: The first AI model with no guardrails that runs on your device - Browser, iPhone and Android Zero servers Zero cloud Zero deployment The whole model lives on your phone

EVAL Engine

16,455 次观看 • 23 天前

1/ We are speed running industrial robotics. It took us just 19 days from the first day of data collection to filming a 2.5-hour continuous run of our model autonomously breaking down industrial containers — zero human intervention. The data efficiency of our DVA model is fundamentally changing how fast we bring robots out of the lab and into the factory. Autonomous operation with 3 hours of data collection at a customer factory.

1/ We are speed running industrial robotics. It took us just 19 days from the first day of data collection to filming a 2.5-hour continuous run of our model autonomously breaking down industrial containers — zero human intervention. The data efficiency of our DVA model is fundamentally changing how fast we bring robots out of the lab and into the factory. Autonomous operation with 3 hours of data collection at a customer factory.

Rhoda AI

25,245 次观看 • 2 个月前

Today, AI model improvements depend a great deal on expert human data. Fun No Priors with Brendan (can/do) founder Mercor on - how to identify outlier talent - the data we need - the future of the labor markets - verifiability, evals - if we can capture “taste”

Today, AI model improvements depend a great deal on expert human data. Fun No Priors with Brendan (can/do) founder Mercor on - how to identify outlier talent - the data we need - the future of the labor markets - verifiability, evals - if we can capture “taste”

sarah guo

27,167 次观看 • 1 年前

someone built a Fortnite map with zero experience. > no Verse knowledge > no Unreal Engine background > described the game mode to Claude > Claude wrote every line 500 daily players. 15 minutes average session. $3,500/month from Epic. recurring. full breakdown in the article below.

someone built a Fortnite map with zero experience. > no Verse knowledge > no Unreal Engine background > described the game mode to Claude > Claude wrote every line 500 daily players. 15 minutes average session. $3,500/month from Epic. recurring. full breakdown in the article below.

Anatoli Kopadze

76,293 次观看 • 1 个月前

The Twitter Model Breakdown

The Twitter Model Breakdown

TTrades🦍

15,670 次观看 • 3 个月前

By the way, you can extract JUST the reasoning from deepseek-reasoner, which means you can send that thinking process to any model you want before they answer you. Like here where I turn gpt-3.5 turbo into an absolute genius!

By the way, you can extract JUST the reasoning from deepseek-reasoner, which means you can send that thinking process to any model you want before they answer you. Like here where I turn gpt-3.5 turbo into an absolute genius!

Pietro Schirano

1,324,620 次观看 • 1 年前

A humanoid robot policy trained solely on synthetic data generated by a world model. Research Scientist Joel Jang presents NVIDIA's DreamGen pipeline: ⦿ Post-train the world model Cosmos-Predict2 with a small set of real teleoperation demos. ⦿ Prompt the world model to generate synthetic video data with verbs and scenarios not used in the world model’s post-training. ⦿ Auto-label synthetic video data with action sequences. ⦿ Train robot policies using only synthetic data. That's it. Deploy zero-shot to a real humanoid robot.

A humanoid robot policy trained solely on synthetic data generated by a world model. Research Scientist Joel Jang presents NVIDIA's DreamGen pipeline: ⦿ Post-train the world model Cosmos-Predict2 with a small set of real teleoperation demos. ⦿ Prompt the world model to generate synthetic video data with verbs and scenarios not used in the world model’s post-training. ⦿ Auto-label synthetic video data with action sequences. ⦿ Train robot policies using only synthetic data. That's it. Deploy zero-shot to a real humanoid robot.

The Humanoid Hub

20,968 次观看 • 11 个月前

Ingest data from hundreds of sources into our multi-model database with the Fivetran connector for SurrealDB. There’s zero manual overhead and no code, so you can spend more time building your AI apps. Learn more. 👉

Ingest data from hundreds of sources into our multi-model database with the Fivetran connector for SurrealDB. There’s zero manual overhead and no code, so you can spend more time building your AI apps. Learn more. 👉

SurrealDB

123,992 次观看 • 1 年前

Order goods, holidays or cars with Bitcoin and crypto. - Real human concierge agents - A fast, intuitive chat designed for effortless ordering - Nothing to download - No account. No tracking. Zero data leaks Just open, save and order privately!

Order goods, holidays or cars with Bitcoin and crypto. - Real human concierge agents - A fast, intuitive chat designed for effortless ordering - Nothing to download - No account. No tracking. Zero data leaks Just open, save and order privately!

ShopinBit

28,977,599 次观看 • 6 个月前

Introducing The Largest Organized and Incentivized Human-Data Fleet for General-Purpose Humanoid-Robotics and large language model training Our users are creators that complete daily data quests with crypto incentives through our Mobile App 🧵

Introducing The Largest Organized and Incentivized Human-Data Fleet for General-Purpose Humanoid-Robotics and large language model training Our users are creators that complete daily data quests with crypto incentives through our Mobile App 🧵

Mecka

385,528 次观看 • 2 年前

Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models. Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents. Read the research paper and download the code.

Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models. Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents. Read the research paper and download the code.

AI at Meta

58,510 次观看 • 1 年前

A reward model that works, zero-shot, across robots, tasks, and scenes? Introducing Robometer: Scaling general-purpose robotic reward models with 1M+ trajectories. Enables zero-shot: online/offline/model-based RL, data retrieval + IL, automatic failure detection, and more! 🧵 (1/12)

A reward model that works, zero-shot, across robots, tasks, and scenes? Introducing Robometer: Scaling general-purpose robotic reward models with 1M+ trajectories. Enables zero-shot: online/offline/model-based RL, data retrieval + IL, automatic failure detection, and more! 🧵 (1/12)

Jesse Zhang

99,699 次观看 • 3 个月前

Edit your reality with Runway. Learn how with the step by step breakdown below. (1/6)

Edit your reality with Runway. Learn how with the step by step breakdown below. (1/6)

Runway

42,706 次观看 • 7 个月前

"America’s Compliance with Federal Vaccine Recommendations Has Collapsed CDC report finds that the vast majority of adults reject ACIP guidelines, refusing COVID-19 genetic booster shots with no human data, as well as flu and RSV vaccines." Nicolas Hulscher full article below

"America’s Compliance with Federal Vaccine Recommendations Has Collapsed CDC report finds that the vast majority of adults reject ACIP guidelines, refusing COVID-19 genetic booster shots with no human data, as well as flu and RSV vaccines." Nicolas Hulscher full article below

DoctorRay

28,988 次观看 • 1 年前

Every model stands on the shoulders of human history. On IBM’s Making Data Simple with Al Martin, Story CPO Andrea | Devrelius maps the necessary trust layer between humans and models for scaling AI. Transparency, attribution, and incentives turn data contributors into stakeholders ↓

Every model stands on the shoulders of human history. On IBM’s Making Data Simple with Al Martin, Story CPO Andrea | Devrelius maps the necessary trust layer between humans and models for scaling AI. Transparency, attribution, and incentives turn data contributors into stakeholders ↓

Story

10,995 次观看 • 3 个月前

What do you think it is? Slo-mo breakdown below 🐌 Plain Model Y? (with new bare-bones trim) Model Y Performance? (doesn't look like it with those wheels) Slightly more compact model? (I'll be ridiculed for saying this but I'm still not seeing a door seam under the B-pillar)

What do you think it is? Slo-mo breakdown below 🐌 Plain Model Y? (with new bare-bones trim) Model Y Performance? (doesn't look like it with those wheels) Slightly more compact model? (I'll be ridiculed for saying this but I'm still not seeing a door seam under the B-pillar)

Eric

119,124 次观看 • 11 个月前

Impressive on every dimension! Genesis AI has announced GENE-26.5, its first robotic foundation model, aimed at achieving human-level dexterous manipulation: - Single model, shared weights, handles egg cracking, lab pipetting, bimanual Rubik's Cube, smoothie making, and wire harnessing. Most tasks: under 1 hour of task-specific data, 1x real-world speed. - Genesis Hand 1.0: human-size, 20 active DoF, soft-contact skin. Paired glove gives a 1:1:1 mapping (glove, human hand, robot hand). 100x cheaper than teleop, 5x more data-efficient internally. - Data engine pulls from three sources: glove data, egocentric (head-cam) video, and large-scale internet video. The 1:1 hand-to-human match closes the embodiment gap, letting Genesis use video data more effectively than rivals. - Flow matching across vision, tactile, proprioception, and language inputs. Per Genesis, scaling data and compute directly improves zero-shot

Impressive on every dimension! Genesis AI has announced GENE-26.5, its first robotic foundation model, aimed at achieving human-level dexterous manipulation: - Single model, shared weights, handles egg cracking, lab pipetting, bimanual Rubik's Cube, smoothie making, and wire harnessing. Most tasks: under 1 hour of task-specific data, 1x real-world speed. - Genesis Hand 1.0: human-size, 20 active DoF, soft-contact skin. Paired glove gives a 1:1:1 mapping (glove, human hand, robot hand). 100x cheaper than teleop, 5x more data-efficient internally. - Data engine pulls from three sources: glove data, egocentric (head-cam) video, and large-scale internet video. The 1:1 hand-to-human match closes the embodiment gap, letting Genesis use video data more effectively than rivals. - Flow matching across vision, tactile, proprioception, and language inputs. Per Genesis, scaling data and compute directly improves zero-shot

The Humanoid Hub

95,119 次观看 • 1 个月前

Tesla Self-Driving 14.3.2 can drive you from parking garage to parking garage in the rain with zero human intervention. Don't believe me? Here's a video of the Model 3 doing just that.

Tesla Self-Driving 14.3.2 can drive you from parking garage to parking garage in the rain with zero human intervention. Don't believe me? Here's a video of the Model 3 doing just that.

Whole Mars Catalog

40,315 次观看 • 1 个月前

Today, we’re excited to introduce Miso One, the most emotive voice model in the world. Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency. We’ve open-sourced the model weights, with API access coming soon. Hear how Miso One sounds in the thread below.

Today, we’re excited to introduce Miso One, the most emotive voice model in the world. Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency. We’ve open-sourced the model weights, with API access coming soon. Hear how Miso One sounds in the thread below.

Aoden Teo

5,019,137 次观看 • 9 天前

xAI is building “human emulators” Khan says “human emulators” will let AI bots perform any on-screen task a human can, using the same keyboard, mouse, and visual cues. The system needs zero changes to existing software because the bot simply mimics the user’s actions. Scaling from a thousand to a million digital workers is mostly an infrastructure question, not a research hurdle. With Tesla-class data centers already available, cost drops and 24/7 uptime make the model economically compelling.

xAI is building “human emulators” Khan says “human emulators” will let AI bots perform any on-screen task a human can, using the same keyboard, mouse, and visual cues. The system needs zero changes to existing software because the bot simply mimics the user’s actions. Scaling from a thousand to a million digital workers is mostly an infrastructure question, not a research hurdle. With Tesla-class data centers already available, cost drops and 24/7 uptime make the model economically compelling.

Wes Roth

76,447 次观看 • 4 个月前