←
Sabine Hossenfelder50,921 posts
Sabine Hossenfelder avatar✓

Sabine Hossenfelder ✓

@skdh

German Physicist. Author of "Lost in Math" & "Existential Physics". There is no strength in numbers, have no such misconception. rt's are not endorsements

◷ Joined November 2007▧ 3,810 Photos and videos
Michael P. Frank 💻🔜♻️ avatar
⟳ Sabine Hossenfelder retweeted
Michael P. Frank 💻🔜♻️@MikePFrank·12h

Turns out some of the “ultra-high-redshift galaxies” found by Webb are really just brown dwarfs in our own galaxy. Embarrassing!

Jinan avatar
⟳ Sabine Hossenfelder retweeted
Jinan@Jinnanog5·10h

数学家 Michael Goldman 昨天发了篇文章,标题就很狂:《我们还需要数学期刊吗?》他自己的答案是:不需要。 最狠的一句:他现在更信任 LLM 的审稿,胜过人类审稿——包括他自己审的。 他还算了笔账:法国公共科研一年花约 7800 万欧元订学术期刊,所有学科加起来,一半进了 Elsevier 的口袋。这笔钱,够给全法国十几二十万研究者人手一份 LLM 基础订阅,还能剩 2000 万。 那期刊现在到底还剩什么用?传播有 arXiv;把关他更信 AI;筛选靠同行讨论,正式见刊要等一年多。 拆到最后,期刊只剩一个用途:招聘和晋升的时候数论文。😂 所以一个一年 7800 万欧元的系统,最后的产出却是一个 KPI。

Dissect Podcast avatar
⟳ Sabine Hossenfelder retweeted
Dissect Podcast@dissectpodcast·18h

You'll never hear this Daft Punk classic the same way again.

1:55
Dr. Roman Yampolskiy avatar
⟳ Sabine Hossenfelder retweeted
Dr. Roman Yampolskiy@romanyam·20h

I can't help but wonder whether all these seemingly unrelated system outages (communications, airlines, banks, etc.) are actually AI agents establishing themselves inside our critical cyberinfrastructure. No evidence, just an increasingly unsettling possibility.

Jakeup avatar
⟳ Sabine Hossenfelder retweeted
Jakeup@myhandle·14h

1. twitter anons make up the claim that many AI lab employees attend RMN 2. a substacker who has never seen the inside of either lab or orgy repeats this 3. a local paper "reports" this as fact 4. national paper cites this "fact" 5. it is now an established source for wiki etc.

Post attachment
Token Gremlin avatar
⟳ Sabine Hossenfelder retweeted
Token Gremlin@TokenGremlin·Nov 15

I'm sorry to inform everyone that OpenAI is facing extreme difficulties turning its incredible pretraining runs with over 10 trillion parameters into functional products for the public. GPT-6.1, GPT-6.2, GPT-6.5, all of them have been and continue to be victims of catastrophic safety and alignment failures. They've been delayed by several weeks, and I don't believe they'll be able to speed things up much in this area. OpenAI is months behind Anthropic on this. Even though I hate Anthropic (but love Claude, of course), we have to collectively admit that their insane focus on safety paid off. Opus 5.5 and Fable 5.2 are absolute monsters, and yet they were released publicly... And OpenAI? Maybe they'll release GPT-6.2 sometime soon, struggling, terrified, loaded with guardrails, nerfed from day one, while delaying everything else even further. I hope they manage to get through this phase.

Fraser avatar
⟳ Sabine Hossenfelder retweeted
Fraser@FraserLeeee·Nov 16

Turns out if you keep piling on more neutrons it eventually becomes stable again

Post attachment
Sabine Hossenfelder avatar
Sabine Hossenfelder@skdh·Nov 16

Sabrina Pasterski has been hailed as the next Einstein, with headlines claiming she could finally unify gravity and quantum theory. Her work on celestial holography is genuinely new but what does it actually tell us about quantum gravity? youtube.com/watch?v=v_DgEwNJ…

Sabine Hossenfelder avatar
Sabine Hossenfelder@skdh·Nov 16

I cannot get over how terrible Google search has become. At this point I cannot even get it to find articles I have written myself if I know both the exact title and the author. ChatGPT finds it instantly.

Post attachmentPost attachment
Misha Teplitskiy | Science of Science avatar
⟳ Sabine Hossenfelder retweeted
Misha Teplitskiy | Science of Science@MishaTeplitskiy·Nov 13

The new report from Anthropic on Claude use is great. One surprising finding: scientists in almost every field are using Claude primarily to *write* rather than *code*

Post attachment
Jacy Reese Anthis avatar
⟳ Sabine Hossenfelder retweeted
Jacy Reese Anthis@jacyanthis·Nov 13

The public demands AI safety. With the breakneck pace of AGI takeoff, some decision-makers race ahead without their input. In new work @GoogleDeepMind, we built "pluralistic agents" to amplify public input. Agents can even give people a voice in technical and confidential debate.

Post attachment