excited to host the builders and financiers bringing AI into factories and infrastructure next Saturday evening in New York
in conversation with @tryramp lead economist @arakharazian, @lumaril cofounder Mark Elliot, and @JaneStreetGroup’s Nick Wage
request an invitation below
i replicated this and tried a subset of @rikiyatakehi et al's exact experiments on new and old models
its pretty clear that only the latest frontier models have these capabilities
Today, we’re announcing Varda’s $250M Series D raise, bringing the total amount of capital raised to $598 million with a $1.6B valuation.
This funding will help us build more vehicles, grow pharmaceutical partnerships, and advance toward the first medicine manufactured in space for patients on Earth.
Data centers are not only critical to future US innovation, but also being built responsibly despite the noise. Closed loop cooling keeps their water use far below everyday staples like almonds.
our latest research proposes a framework for setting incentives to deter misbehavior in DePINs.
here are the main insights and and an intuitive explanation of the paper's results. https://t.co/6BmZxp5GK0
@helium has had close to ~10% of its online nodes classified as malicious on a denylist, with 25-40M in rewards paid out to cheating nodes. @ionet faced a spoof of over 1.8M fake GPUs in one day
Our new paper tackles this misbehavior by deriving the conditions to set incentives that deter cheaters. We also shipped a widget that calculates the right incentives for your protocol design:
two critiques of dePIN have proven to be the most relevant:
1. way too much inflation
2. quality control
if you look closely enough, the second issue contributes significantly to the first one and should probably be robustly solved first
OpenAI posted a job opening for "Power Trading Lead", responsible for developing and executing hedging strategies for power and gas for data centers.
First job like this I've seen them advertise.
Most robotics datasets are magic soup. Every source, sensor, and quality level poured into one pot. It looks like scale. But if you're training on it, most of it is inconsistent data you can't trace, and that's what quietly degrades your model.