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。

Google DeepMind pre-training lead explains two skills with massive demand by AI frontier labs: > Kernel Development > Low Level Performance Engineering

ℏεsam

90,503 subscribers

81,607 次观看 • 3 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

5 条评论

Russian Bot 的头像

Russian Bot3 个月前

If I was running Google I would forbid any “Googler” from doing interviews until Googles position in the AI race was no longer a complete joke.

artyst 的头像

artyst3 个月前

interesting, let's see how it unfolds

David Song 的头像

David Song3 个月前

sw mentioned in the same sentence as low latency feels like a crime hft mastered latency years ago and it's not via writing more layers of sw

Eclipse 🌖 的头像

Eclipse 🌖3 个月前

Makes sense—demand for kernel dev and low-level perf engineering has been climbing with every new training run. Curious how many of those roles are going to ex-FAANG systems folks vs. fresh grads.

GrillePainVert 的头像

GrillePainVert3 个月前

Basically they still need 10 human software developers.

相关视频

Vlad Feinberg (Vlad Feinberg) is Google DeepMind’s pre-training area lead and I asked him all about how to land a job at a frontier lab like Google DeepMind, Anthropic or OpenAI. In this episode: • Skills frontier labs need • Differences between software engineering and AI research • Domains that matter for frontier research • Concrete steps engineers can take to get closer to research • Jeff Dean spot bonus story Where to watch: • YouTube - • Spotify - • Apple Podcasts - • Transcript - Thank you to the sponsor of this episode for supporting my work: • WorkOS: makes your app Enterprise Ready with easy to use APIs to add SSO, SCIM, RBAC, and more in just a few lines of code, check them out at Chapters: 00:00 - Intro 00:33 - Skills frontier labs need 08:45 - The difference between AI research and engineering 21:41 - Domains that matter for the frontier 30:50 - Marketing yourself to frontier labs 35:13 - Concrete steps engineers can take 38:29 - Overview of pre-training areas 47:23 - Jeff Dean spot bonus story 50:14 - Favorite Gemini war story 58:59 - Advice for his younger self 01:03:07 - Outro

Vlad Feinberg (Vlad Feinberg) is Google DeepMind’s pre-training area lead and I asked him all about how to land a job at a frontier lab like Google DeepMind, Anthropic or OpenAI. In this episode: • Skills frontier labs need • Differences between software engineering and AI research • Domains that matter for frontier research • Concrete steps engineers can take to get closer to research • Jeff Dean spot bonus story Where to watch: • YouTube - • Spotify - • Apple Podcasts - • Transcript - Thank you to the sponsor of this episode for supporting my work: • WorkOS: makes your app Enterprise Ready with easy to use APIs to add SSO, SCIM, RBAC, and more in just a few lines of code, check them out at Chapters: 00:00 - Intro 00:33 - Skills frontier labs need 08:45 - The difference between AI research and engineering 21:41 - Domains that matter for the frontier 30:50 - Marketing yourself to frontier labs 35:13 - Concrete steps engineers can take 38:29 - Overview of pre-training areas 47:23 - Jeff Dean spot bonus story 50:14 - Favorite Gemini war story 58:59 - Advice for his younger self 01:03:07 - Outro

Ryan Peterman

191,659 次观看 • 3 个月前

This is the google of AI agents If you're building with AI agents, you need it Search across 164+ agent skills > Filter > Search > Copy install

This is the google of AI agents If you're building with AI agents, you need it Search across 164+ agent skills > Filter > Search > Copy install

0xMarioNawfal

89,655 次观看 • 8 个月前

It’s time to level up your agentic development with Android skills. 🚀 These AI-optimized instructions automate more complex tasks like XML-to-Compose and AGP9 migrations, follow our official best practices and guidance on Android development, and are activated on demand by AI tools. Get started with Android skills →

It’s time to level up your agentic development with Android skills. 🚀 These AI-optimized instructions automate more complex tasks like XML-to-Compose and AGP9 migrations, follow our official best practices and guidance on Android development, and are activated on demand by AI tools. Get started with Android skills →

Android Developers

15,745 次观看 • 5 个月前

Google just did it again. They just launched "Google Skills" a free platform to turn anyone into an AI expert. 3,000+ courses, labs & certificates from Google Cloud, DeepMind, and Grow with Google all in one place. → Learn AI, data, and coding from real Google engineers → Earn credentials that actually get you hired → Build hands-on projects with Gemini Code Assist → Compete on global leaderboards and share your wins No paywalls. No BS. Just world-class AI education free. This might be the biggest AI upskilling move of the decade. 👉

Google just did it again. They just launched "Google Skills" a free platform to turn anyone into an AI expert. 3,000+ courses, labs & certificates from Google Cloud, DeepMind, and Grow with Google all in one place. → Learn AI, data, and coding from real Google engineers → Earn credentials that actually get you hired → Build hands-on projects with Gemini Code Assist → Compete on global leaderboards and share your wins No paywalls. No BS. Just world-class AI education free. This might be the biggest AI upskilling move of the decade. 👉

Alex Veremeyenko

41,961 次观看 • 11 个月前

DeepMind co-founder Shane Legg: AI won't bring sudden disruption; it will quietly take over economic work, domain by domain Within a few years, sectors like software engineering will see massive efficiency gains, with 20 engineers using AI doing the work of 100

DeepMind co-founder Shane Legg: AI won't bring sudden disruption; it will quietly take over economic work, domain by domain Within a few years, sectors like software engineering will see massive efficiency gains, with 20 engineers using AI doing the work of 100

Haider.

61,918 次观看 • 10 个月前

Vibe coding with AI Studio is so addictive! I built an AI Hair Stylist Agent that: > find the best haircut by looking my face > generates the after images using nano banana > lets me edit it live with the live API > find nearby salons using google maps Built in just 5 mins!

Vibe coding with AI Studio is so addictive! I built an AI Hair Stylist Agent that: > find the best haircut by looking my face > generates the after images using nano banana > lets me edit it live with the live API > find nearby salons using google maps Built in just 5 mins!

Shubham Saboo

46,658 次观看 • 11 个月前

here's how to find profitable startup ideas using reddit > go to > select the reddit tool > ask it to find posts where people are asking for ai apps > it’ll summarize the best ideas by category > pick the ones with high demand but few existing products

here's how to find profitable startup ideas using reddit > go to > select the reddit tool > ask it to find posts where people are asking for ai apps > it’ll summarize the best ideas by category > pick the ones with high demand but few existing products

Okara

21,739 次观看 • 11 个月前

5 patterns in Text Arena's price–performance Pareto frontier since 2023: 1. GPT-4-level quality is now ~500x lower cost. - From a ~$50 blended price per million tokens in 2023 to ~$0.10 today. 2. The higher-price end is both better and lower-priced since 2023. - The leading Arena score has climbed ~170 points (1,330 → 1,500). While the price of the higher-end frontier models dropped from ~$50 to ~$20 per million tokens. 3. The low-cost end gained the most. - Under $0.20 per million tokens, the best available model went from ~1,000 Arena score in 2023 to ~1,440 today. 4. The low-cost/top performance gap has nearly closed. - In 2023, sub-$0.20 models trailed the leader by ~350 Arena points. Today, ~60. 5. The cast has rotated quite a bit. - - OpenAI set the 2023–24 benchmark. - AI at Meta strengthened the low-cost end in 2024. - Google DeepMind drove the 2025 jump. - Anthropic holds the peak in 2026. - xAI and Chinese labs like DeepSeek AI, Z.ai, Kimi.ai, Xiaomi MiMo, and Qwen are continuing to push the mid-price frontier.

5 patterns in Text Arena's price–performance Pareto frontier since 2023: 1. GPT-4-level quality is now ~500x lower cost. - From a ~$50 blended price per million tokens in 2023 to ~$0.10 today. 2. The higher-price end is both better and lower-priced since 2023. - The leading Arena score has climbed ~170 points (1,330 → 1,500). While the price of the higher-end frontier models dropped from ~$50 to ~$20 per million tokens. 3. The low-cost end gained the most. - Under $0.20 per million tokens, the best available model went from ~1,000 Arena score in 2023 to ~1,440 today. 4. The low-cost/top performance gap has nearly closed. - In 2023, sub-$0.20 models trailed the leader by ~350 Arena points. Today, ~60. 5. The cast has rotated quite a bit. - - OpenAI set the 2023–24 benchmark. - AI at Meta strengthened the low-cost end in 2024. - Google DeepMind drove the 2025 jump. - Anthropic holds the peak in 2026. - xAI and Chinese labs like DeepSeek AI, Z.ai, Kimi.ai, Xiaomi MiMo, and Qwen are continuing to push the mid-price frontier.

Arena.ai

58,356 次观看 • 4 个月前

Union Alpha lost to DeepSeek V4.1 Flash running locally on 2 DGX Sparks 🐳 DeepSeek v4.1 Flash beat Union Alpha, which is supposed to deliver frontier level performance, though real tasks show a completely different story Run AI models locally ->

Union Alpha lost to DeepSeek V4.1 Flash running locally on 2 DGX Sparks 🐳 DeepSeek v4.1 Flash beat Union Alpha, which is supposed to deliver frontier level performance, though real tasks show a completely different story Run AI models locally ->

atomic.chat

79,581 次观看 • 23 天前

if you’re passionate about ai skills & literacy, this is for you. these programs are free, beginner-friendly, and come with a certificate after completing each course. → google learn ai { free courses by google covering ai fundamentals, prompt writing, and practical use cases for work and business. perfect for beginners. → anthropic courses { learn how to use claude and understand responsible ai, safety, and effective prompting directly from the team building frontier models. → amazon aws ai training { introductory to advanced ai and machine learning courses, including generative ai on aws. great if you want cloud + ai skills. → codecademy { interactive, hands-on coding lessons. you can learn python, data science, and ai-related programming skills step by step in your browser. any of these courses can take you from a complete beginner with zero experience or poor ai skills to a more advanced level. and since they’re completely free, you’ve got no excuses left to become ai literate. thanks for your attention.

if you’re passionate about ai skills & literacy, this is for you. these programs are free, beginner-friendly, and come with a certificate after completing each course. → google learn ai { free courses by google covering ai fundamentals, prompt writing, and practical use cases for work and business. perfect for beginners. → anthropic courses { learn how to use claude and understand responsible ai, safety, and effective prompting directly from the team building frontier models. → amazon aws ai training { introductory to advanced ai and machine learning courses, including generative ai on aws. great if you want cloud + ai skills. → codecademy { interactive, hands-on coding lessons. you can learn python, data science, and ai-related programming skills step by step in your browser. any of these courses can take you from a complete beginner with zero experience or poor ai skills to a more advanced level. and since they’re completely free, you’ve got no excuses left to become ai literate. thanks for your attention.

m0h

19,482 次观看 • 7 个月前

Ayar Labs and Wiwynn are partnering to bring CPO to rack-scale AI systems that support next-generation hyperscale AI workloads. By combining Ayar Labs’ CPO solution with Wiwynn’s rack-level system design and manufacturing capabilities, the two companies are enabling a new class of rack-scale AI infrastructure.

Ayar Labs and Wiwynn are partnering to bring CPO to rack-scale AI systems that support next-generation hyperscale AI workloads. By combining Ayar Labs’ CPO solution with Wiwynn’s rack-level system design and manufacturing capabilities, the two companies are enabling a new class of rack-scale AI infrastructure.

Ayar Labs

137,455 次观看 • 4 个月前

Android 17 root Full chain browser-to-kernel exploit with two 0-day vulnerabilities affecting Firefox before v151.0.2 (CVE-2026-10702) Click on the link -> root Android Discovered by Nebula Security PoC not available. Info:

Android 17 root Full chain browser-to-kernel exploit with two 0-day vulnerabilities affecting Firefox before v151.0.2 (CVE-2026-10702) Click on the link -> root Android Discovered by Nebula Security PoC not available. Info:

Mobile Hacker

53,134 次观看 • 3 个月前

1/ The first big announcement: Google DeepMind has unveiled Genie 3 A new frontier for world models, the first AI that generates interactive 3D worlds in real time at 24 fps. The breakthrough: minutes of consistency at 720p using only text descriptions. No pre-built 3D models required. Let's break it down - with some examples:

1/ The first big announcement: Google DeepMind has unveiled Genie 3 A new frontier for world models, the first AI that generates interactive 3D worlds in real time at 24 fps. The breakthrough: minutes of consistency at 720p using only text descriptions. No pre-built 3D models required. Let's break it down - with some examples:

Chubby♨️

434,629 次观看 • 1 年前

The Nobel Prize winner just said something big. The biggest AI opportunity isn't at Google. It's with you. Demis Hassabis says a kid today could start a multi-billion dollar business by applying AI to something new. The frontier labs can't explore it all. That gap is your opportunity.

The Nobel Prize winner just said something big. The biggest AI opportunity isn't at Google. It's with you. Demis Hassabis says a kid today could start a multi-billion dollar business by applying AI to something new. The frontier labs can't explore it all. That gap is your opportunity.

Alvin Foo

70,773 次观看 • 5 个月前

Gemini 3, scaling laws and the 'finite data' era: my conversation with Sebastian Borgeaud, research engineer at Google DeepMind and a pre-training lead for Gemini 3 00:00 – Cold intro: “We’re ahead of schedule” + AI is now a system 00:58 – Oriol Vinyals's “secret recipe”: better pre- + post-training 02:09 – Why AI progress still isn’t slowing down 03:04 – Are models actually getting smarter? 04:36 – Two–three years out: what changes first? 06:34 – AI doing AI research: faster, not automated 07:45 – Frontier labs: same playbook or different bets? 10:19 – Post-transformers: will a disruption happen? 10:51 – DeepMind’s advantage: research × engineering × infra 12:26 – What a Gemini 3 pre-training lead actually does 13:59 – From Europe to Cambridge to DeepMind 18:06 – Why he left RL for real-world data 20:05 – From Gopher to Chinchilla to RETRO (and why it matters) 20:28 – “Research taste”: integrate or slow everyone down 23:00 – Fixes vs moonshots: how they balance the pipeline 24:37 – Research vs product pressure (and org structure) 26:24 – Gemini 3 under the hood: MoE in plain English 28:30 – Native multimodality: the hidden costs 30:03 – Scaling laws aren’t dead (but scale isn’t everything) 33:07 – Synthetic data: powerful, dangerous? 35:00 – Reasoning traces: what he can’t say (and why) 37:18 – Long context + attention: what’s next 38:40 – Retrieval vs RAG vs long context 41:49 – The real boss fight: evals (and contamination) 42:28 – Alignment: pre-training vs post-training 43:32 – Deep Think + agents + “vibe coding” 46:34 – Continual learning: updating models over time 49:35 – Advice for researchers + founders 53:35 – “No end in sight” for progress + closing

Gemini 3, scaling laws and the 'finite data' era: my conversation with Sebastian Borgeaud, research engineer at Google DeepMind and a pre-training lead for Gemini 3 00:00 – Cold intro: “We’re ahead of schedule” + AI is now a system 00:58 – Oriol Vinyals's “secret recipe”: better pre- + post-training 02:09 – Why AI progress still isn’t slowing down 03:04 – Are models actually getting smarter? 04:36 – Two–three years out: what changes first? 06:34 – AI doing AI research: faster, not automated 07:45 – Frontier labs: same playbook or different bets? 10:19 – Post-transformers: will a disruption happen? 10:51 – DeepMind’s advantage: research × engineering × infra 12:26 – What a Gemini 3 pre-training lead actually does 13:59 – From Europe to Cambridge to DeepMind 18:06 – Why he left RL for real-world data 20:05 – From Gopher to Chinchilla to RETRO (and why it matters) 20:28 – “Research taste”: integrate or slow everyone down 23:00 – Fixes vs moonshots: how they balance the pipeline 24:37 – Research vs product pressure (and org structure) 26:24 – Gemini 3 under the hood: MoE in plain English 28:30 – Native multimodality: the hidden costs 30:03 – Scaling laws aren’t dead (but scale isn’t everything) 33:07 – Synthetic data: powerful, dangerous? 35:00 – Reasoning traces: what he can’t say (and why) 37:18 – Long context + attention: what’s next 38:40 – Retrieval vs RAG vs long context 41:49 – The real boss fight: evals (and contamination) 42:28 – Alignment: pre-training vs post-training 43:32 – Deep Think + agents + “vibe coding” 46:34 – Continual learning: updating models over time 49:35 – Advice for researchers + founders 53:35 – “No end in sight” for progress + closing

Matt Turck

51,484 次观看 • 9 个月前

🚨 Big news: Google just launched Google Skills... giving you access to the same content universities charge $50K+ is now open to everyone. Here’s how it works: - Go to - Pick any AI skill you want to learn - Learn through 700+ hands-on labs with real Google Cloud tools -Earn Google Skill Badges to showcase your expertise ... recruiters actively look for these - Work toward certifications that rank among the highest-paying in IT What’s inside: - 3,000+ AI courses from Google Cloud, DeepMind & Google Edu - $500 in free cloud credits to practice - Built-in Gemini Code Assist - Certificates employers actually recognize (82% hiring preference) Like, REPOST

🚨 Big news: Google just launched Google Skills... giving you access to the same content universities charge $50K+ is now open to everyone. Here’s how it works: - Go to - Pick any AI skill you want to learn - Learn through 700+ hands-on labs with real Google Cloud tools -Earn Google Skill Badges to showcase your expertise ... recruiters actively look for these - Work toward certifications that rank among the highest-paying in IT What’s inside: - 3,000+ AI courses from Google Cloud, DeepMind & Google Edu - $500 in free cloud credits to practice - Built-in Gemini Code Assist - Certificates employers actually recognize (82% hiring preference) Like, REPOST

Beginnersblog

142,634 次观看 • 11 个月前