Sabitlenmiş Tweet
Crunch (Quant/acc)
3.2K posts
Crunch (Quant/acc)
@crunchDAO
Backed by @galaxyhq, @vaneck_us, Crunch's network of 10,000+ ML engineers delivers production-ready ML models for global scale.
Everywhere Katılım Temmuz 2020
518 Takip Edilen23.8K Takipçiler
@incubator @Schmidt_Center Thank you for the support. More to come!
English
Proud alumni moment.
@crunchDAO mobilizes 12,000+ ML researchers to solve orgs' hardest problems.
Now partnered with MIT & Harvard's @Schmidt_Center, they've ranked 4.47M gene pairs for obesity and run their first decentralized peer review.
Next up: real lab testing.
Crunch (Quant/acc)@crunchDAO
Crunch just ran its first decentralized peer review, in partnership with the Eric and Wendy Schmidt Center at the Broad Institute (@Schmidt_Center). Until now Crunch collected thousands of pieces of code. Now the platform can also collect the reasoning behind that code, and have other Crunchers score it. In the Obesity ML Challenge, Crunchers' models ranked 4.47 million candidate gene pairs. Peer review scores the top submissions and helps decide which get tested in the lab. Only about 40 will make it. Every proposal is read and scored by 4 other Crunchers. Reviews are blind, and each score comes with the reviewer's written reasoning. 316 peer reviews in.
English
@Inigo_the3rd @Schmidt_Center Highly practical and built for a real need.
English
@crunchDAO @Schmidt_Center This feels like a practical use case for decentralized collaboration
English
Crunch just ran its first decentralized peer review, in partnership with the Eric and Wendy Schmidt Center at the Broad Institute (@Schmidt_Center).
Until now Crunch collected thousands of pieces of code. Now the platform can also collect the reasoning behind that code, and have other Crunchers score it.
In the Obesity ML Challenge, Crunchers' models ranked 4.47 million candidate gene pairs. Peer review scores the top submissions and helps decide which get tested in the lab. Only about 40 will make it.
Every proposal is read and scored by 4 other Crunchers. Reviews are blind, and each score comes with the reviewer's written reasoning.
316 peer reviews in.
English
AI will transform every industry, power every company, and be built by every country. And by open communities.
Jensen Huang@JensenHuang
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…
English
@crunchDAO Touching grass while shipping code is the ultimate combo
English
@FarmTheDip8 and the most interesting chapters haven't been written yet
English
Attempt 31: failed.
Attempt 32: failed, but slower.
Attempt 33: failed in an interesting way.
Attempt 34: worked, but only once.
Attempt 35: worked.
Every scientist has a ladder like this. Most of what they know lives in the rungs below, 31 through 34. The dead ends, the near misses, the reason 35 finally worked.
It is the part the world almost never sees in papers and talks, but it's essential to building a frontier AI model that thinks like a scientist.
English
Crunch (Quant/acc) retweetledi
Met almost 100 crunchers now over the past 3 weeks working on this announcement. I have 25+ session incoming this week. These session are incredible even if they’ve killed my productivity last week! Just deeply rewarding. Apologies to all crunchers I cancelled or missed the meeting I’ll be better next week 🫡Crunchers are all genius.
GIF
Cruncher Jean@herelle_jean
English
@DefiDragans that's the endgame. and the only way there is through doing.
English
@crunchDAO If the model learns from every strange path that works, how long before it invents paths no human would ever try?
English
Sunrise in Seoul: one Cruncher opens their notebook.
Midnight in São Paulo: a different Cruncher finally closes theirs.
They are solving the same problem with the same gene data.
Neither knows the other exists. Soon the model will watch both, and thousands more like them.
It will learn the problem from every angle, including the strange routes that should not work but do.
English
@firtsjay the Seoul part wasn't made up.
83 crunchers from South Korea: hub.crunchdao.com/analytics
English
@mikaonlines0af Yes, and the reasoning is arguably the more valuable half.
English
@crunchDAO Wait so theyre basically crowdsourcing how to think like a scientist
English
@crunchDAO this is what all those competition traces were quietly building toward
English