Many people spent a career in relative obscurity to deliver a breakthrough result in cancer care today. Hoping this drug gets to all the patients that need it as soon as possible/yesterday.
DEI has been counterproductive generally. By forcing people to focus on race, it has unavoidably heightened racism, not dampened it. If DEI were an antidote to racism, there would exist scores of studies providing the evidence. None exists.
Walk more. Walk for no reason. Walk to think. Walk to reset. Walk to learn. Walk to escape. Walk to get better. Walk to solve problems. It changes everything
rfk jr has been eating a diet of mostly meat and fermented vegetables, and the germans noticed and pointed out that hes essentially recreated the german diet from first principles. hes krautmaxxing.
the biggest lie of the last fifty years, that humans are born as identical blank slates and all difference in outcome is purely environmental, has finally been murdered by people who actually sequenced the genome
this was obvious to anyone with a pair of functioning neurons who didn't have to lie to score political points
so the next time someone tells you intelligence is purely socioeconomic, or that race is just a social construct, or that we are all born with identical cognitive capacity, you can stop engaging with the bad faith
some people are born better. that is in their genes
is that fair? who cares, it's just a fact
“The Supreme Court banned race discrimination in college admissions and Asian acceptance rates increased. I have no idea how these two might be related. I am a reporter.”
Introducing Gemma 4, our series of open weight (Apache 2.0 licensed) models, which are byte for byte the most capable open models in the world!
Gemma 4 is build to run on your hardware: phones, laptops, and desktops.
Frontier intelligence with a 26B MOE and a 31B Dense model!
Pretraining is data-inefficient. This is entirely a consequence of the fact that we throw away the KV cache after every forward-backward step!
If we can integrate efficient KV cache compaction into pretraining, we will unlock human level data efficiency. Neural KV cache compaction makes this possible.
That’s why I usually say that comparing open models with closed-source APIs or products is like comparing apples and oranges. Or comparing an engine with a full car. Or comparing an ingredient with a Michelin dinner (missing ingredients, prep and chef).
There’s a lot of scaffolding and tricks that is done behind APIs that make it unfair to compare to a raw open-source model! However that also means that if you put the work to really try to make open-source models work, they can give you much better results than what benchmark suggests (and that’s without even post-training/fine-tuning)