This report is long but very good.
“With R1, DeepSeek essentially cracked one of the holy grails of AI: getting models to reason step-by-step without relying on massive supervised datasets. Their DeepSeek-R1-Zero experiment showed something remarkable: using pure reinforcement learning with carefully crafted reward functions, they managed to get models to develop sophisticated reasoning capabilities completely autonomously. This wasn't just about solving problems— the model organically learned to generate long chains of thought, self-verify its work, and allocate more computation time to harder problems.
The technical breakthrough here was their novel approach to reward modeling. Rather than using complex neural reward models that can lead to "reward hacking" (where the model finds bogus ways to boost their rewards that don't actually lead to better real-world model performance), they developed a clever rule-based system that combines accuracy rewards (verifying final answers) with format rewards (encouraging structured thinking). This simpler approach turned out to be more robust and scalable than the process-based reward models that others have tried.”
https://t.co/XRqgMNRzg1
Breaking: @krakenfx states officially that the #MtGox#Bitcoin has been successfully distributed to creditors.
#wegoingup ☝️Higher.
https://t.co/yu7SfWUG2U
I had brain #MRI scan last Thursday looking for recurrent #glioblastoma (&/or treatment complications). I found out yesterday that there is still no sign of recurrence. I couldn’t be happier!!!!!
Thank you to the fabulous team looking after me so well especially my wife Katie & wonderful family!
@SydneyLHD@Sydney_Uni
Excited to share our new Nature paper! In this work, we propose a new display design that pairs inverse-designed metasurface waveguides with AI-driven holographic displays to enable full-color 3D augmented reality from a compact eyeglasses-like form factor.
1/8
The two Chinese labs working on replicating LK-99 appear to have found a room-temperature superconductor.
At first blush, here's what's different from last time:
• it's more like "room temperature" than room temperature, the paper says 250K which is -10 F or -23 C. That's still HUGE IF TRUE, because we can get things that cold with liquid nitrogen
• we have actually discovered a superconductor at this temperature before, but it was at high pressure. This paper says it's potentially superconductive AT AMBIENT, NORMAL PRESSURE
• it has *already* replicated, with two separate labs in China confirming the results. Last time the big question was "will it replicate." And the answer this time seems to be "it already has"
Me: Can you draw a very normal image?
ChatGPT: Here is a very normal image depicting a tranquil suburban street scene during the daytime.
Me: Not bad, but can you go more normal than that?
(cont.)
Exactly 17 hours ago Runway dropped amazing improvements to their video generation models...
Don't want to hype it too much but..
ARE you kidding me!
10 - basically HD video - examples:
The discoveries by Katalin Karikó and Drew Weissman – awarded the 2023 #NobelPrize in Physiology or Medicine – were critical for developing effective mRNA vaccines against COVID-19 during the pandemic that began in early 2020.
Through their groundbreaking findings, this year’s laureates have fundamentally changed our understanding of how mRNA interacts with our immune system. They have contributed to the unprecedented rate of vaccine development during one of the greatest threats to human health in modern times.
ChatGPT can now see, hear, and speak. Rolling out over next two weeks, Plus users will be able to have voice conversations with ChatGPT (iOS & Android) and to include images in conversations (all platforms).
https://t.co/uNZjgbR5Bm