steve hsu
@hsu_steve
Followers
42K
Following
4K
Media
3K
Statuses
12K
Physicist, AI Founder, Manifold Podcast
Joined June 2010
Richard Feynman and the 19 year old me at my Caltech graduation (Found this in my mom's photo album :-) #feynman
84
164
2K
US spending ~1% GDP to race PRC toward AGI, but ideologues can't be bothered to understand impact of AI/ML and huge genomic datasets on HUMAN REPRODUCTION. 🤡🐒
The Guardian slowly joining the reality-based community in genomics. UK IVF couples use legal loophole to rank embryos based on potential IQ, height and health What is polygenic embryo screening in IVF and does it work? Scientists have developed algorithms that give predictive
0
3
42
This paper from my lab is 5 years old and the results have been replicated many times by other groups. If you dig into this topic you'll find that almost no one whose primary expertise is bioethics or reproductive medicine or even genetics of single gene disorders can read these
1
0
1
You remember COVID era censorship, right? They banned us from their apps. They deplatformed alternate views and tried to silence any form of opposition. That's why we built Unplugged. So when the chips are down, we know we have an independent platform that will have our back.
0
254
991
The Guardian slowly joining the reality-based community in genomics. UK IVF couples use legal loophole to rank embryos based on potential IQ, height and health What is polygenic embryo screening in IVF and does it work? Scientists have developed algorithms that give predictive
7
7
54
DeepSeek Sparse Attention (DSA) in V3.2: "Our Long Context Reasoning benchmark showed no cost to intelligence of the introduction of DSA. DeepSeek reflected this cost advantage of V3.2-Exp by cutting pricing on their first party API from $0.56/$1.68 to $0.28/$0.42 per 1M
DeepSeek V3.2 is the #2 most intelligent open weights model and also ranks ahead of Grok 4 and Claude Sonnet 4.5 (Thinking) - it takes DeepSeek Sparse Attention out of ‘experimental’ status and couples it with a material boost to intelligence @deepseek_ai V3.2 scores 66 on the
2
3
24
Marques has 20M subscribers on YT https://t.co/ftWbTy4tLG He says: $42k Xiaomi SU7 Max competes with $75k cars in the US mkt. Conclusion: Western (incl. Japanese and Korean) carmakers are "COOKED" if Chinese EVs are allowed to be sold in the US and EU. SU7 will be coming to
15
11
127
DS V3.2 (Reasoning) at Pareto frontier of intelligence vs cost. For high-end tasks (like IMO problems or really hard coding or science tasks) it would appear that Speciale is also at the frontier. Speciale might use ~1.5 to 2x more tokens, but the overall cost is still TINY
At DeepSeek's first-party API pricing of $0.28/$0.42 per 1M input/output tokens, V3.2 (Reasoning) sits on the Pareto frontier of the Intelligence vs. Cost to Run Artificial Analysis Intelligence Index chart
1
3
35
What is new: we have (noisy) DNA predictors for cognitive ability. https://t.co/vICjwxxoT9 Some researchers claim ~(5-10) IQ pts expected gain from selection among 10 embryos.
arxiv.org
How do genes affect cognitive ability or other human quantitative traits such as height or disease risk? Progress on this challenging question is likely to be significant in the near future. I...
The Good News Is That One Side Has Definitively Won The Missing Heritability Debate, reports Scott Alexander @slatestarcodex. Actually, the real debate was won long ago: heritability of intelligence is substantially above 0. There can't be an exact "correct" estimate, because
2
2
53
Qwen AI generally won't self-censor on sensitive US political issues. Here is what it says about the Epstein affair. Summary of the Evidence and Theories The Jeffrey Epstein case represents far more than a simple sex trafficking operation. Evidence points to a complex nexus of
Manifold Podcast Episode #100! Jeffrey Epstein, Israel, and Elite Power, with Murtaza Hussain Murtaza Hussain is a reporter for Drop Site News, which has broken important stories based on recently obtained Epstein emails. Hussain reports that Epstein had an “extensive
1
12
69
Bonus option! "... the main purpose of my teaching has not been to prepare you for some examination—it was not even to prepare you to serve industry or the military. I wanted most to give you some appreciation of the wonderful world and the physicist's way of looking at it,
1
1
16
ZTE’s 7B Nebula-GUI model (a vision model trained to understand screens) running locally on the Snapdragon 8 Elite drives the UI like a human: it picks Ctrip app, opens it, fills in dates, finds cheapest rate...
Here you see the cloud + on-device split very cleanly. Doubao handles the semantics: from a single hotel entrance photo it figures out which hotel this is, that I want to book tonight, and that it will need to check the hotel’s pet policy. Then ZTE’s 7B Nebula-GUI model (a
2
9
53
Prototype of ZTE’s Nubia M153 running ByteDance’s Doubao AI agent ... US AI "experts": 😴
Another DeepSeek moment. This is the world’s first actual smart phone. It’s an engineering prototype of ZTE’s Nubia M153 running ByteDance’s Doubao AI agent fused into Android at the OS level. It has complete control over the phone. It can see the UI, choose/download apps,
4
7
52
