🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Generalists* eager to shape AI policy should work for Lennart Heim ! Lennart Heim has my highest endorsement! (*Especially ML engineers and semiconductor experts, but this is not required - people who aren't that are welcome too)

Many people are asking this. Some starter ideas here: cdn.governance.ai/RFI_Labor_Impa…

Our AI Supercomputers database is now live! Explore specs for 700+ training & inference clusters—compute, chip types, power draw, ownership, location, and more.

People keep asking me ‘Konstantin, where are all the data centers?’ Today, I can finally give you the answer. Explore our new dataset of 750 AI supercomputers, both those that already exist and those planned over the next five years. Some of my own analysis 🧵

The core idea is simple: building a broad and deep understanding, bit by bit, over time, is worthwhile. In particular, it compounds: build up your understanding by answering small questions and you'll be surprised how fast you can start to answer big questions.

Morning. Here are our guest call-ins today: - Bernt Bornich (1X Tech) - EMILY SUNDBERG (Feed Me) - joshua steinman (🇺🇸,🇺🇸) (Galvanick) - John Doyle (Cape) - Lennart Heim (RAND Corp) See you all on the stream.

We're planning to review the first batch of applications next week (Monday 6/23)! Make sure to apply if you're interested in crunching numbers for better AI policy.

NVIDIA's latest chip architecture (Blackwell) achieves 10 TB/s die-to-die bandwidth—nearly 20x faster than the 910C's 540 GB/s. This gap shows the limitations of China's domestic advanced packaging offerings. More: x.com/ohlennart/stat…

TIL about the N-th Country Experiment.

me defending my most recent estimates

Plot twist: Replace "shipbuilding" with "chipbuilding" and swap China for that small island neighbor. Same story, different industry.

We took a look at Chinese government efforts to accelerate AI. TL;DR: They are subsidizing at every layer of the stack, some of it useful, some wasteful. Compute remains a key bottleneck that will be hard to overcome without Western tech.

Read the full piece: "Beyond Export Controls: How Strategic Promotion of American AI Abroad Can Protect the Homeland & Check China" therepublicjournal.com/web-exclusives…

The window of opportunity is currently open. The PRC still lacks the compute necessary to widely export and embed their cloud services and AI systems around the world (see work from Lennart Heim below). x.com/ohlennart/stat…

This is an important and good paper! As someone often seen as an advocate for training compute thresholds, I'm glad to see criticism focused on what matters most: the regulatory target. But some comments: First, I shared similar reflections a few months ago after leaving