Hendrik Weimer
@hweimer.fediscience.org.ap.brid.gy
Theoretical physicist, professor for condensed matter theory at TU Berlin.
[bridged from https://fediscience.org/@hweimer on the fediverse by https://fed.brid.gy/ ]
[bridged from https://fediscience.org/@hweimer on the fediverse by https://fed.brid.gy/ ]
Reposted by Hendrik Weimer
The highlights. ML can only be used where "its OK to be wrong", which usually means "its OK to be a little more wrong to be a LOT cheaper", which is the edge.
Oh, and LLMs are very niche as they can't be used with "untrusted input + do bad thing". There is no market for the GW datacenters.
Oh, and LLMs are very niche as they can't be used with "untrusted input + do bad thing". There is no market for the GW datacenters.
Nicholas Weaver: The futile future of the gigawatt datacenter
The edge is where it’s at
Blog post by @ncweaver.skerry-tech.com:
pivot-to-ai.com/2025/11/07/t...
Interview with Nick:
www.youtube.com/watch?v=a5rL... - video
pivottoai.libsyn.com/20251107-nic... - podcast
time: 26 min 53 sec
The edge is where it’s at
Blog post by @ncweaver.skerry-tech.com:
pivot-to-ai.com/2025/11/07/t...
Interview with Nick:
www.youtube.com/watch?v=a5rL... - video
pivottoai.libsyn.com/20251107-nic... - podcast
time: 26 min 53 sec
November 7, 2025 at 3:08 PM
The highlights. ML can only be used where "its OK to be wrong", which usually means "its OK to be a little more wrong to be a LOT cheaper", which is the edge.
Oh, and LLMs are very niche as they can't be used with "untrusted input + do bad thing". There is no market for the GW datacenters.
Oh, and LLMs are very niche as they can't be used with "untrusted input + do bad thing". There is no market for the GW datacenters.
In space, no one can hear you violate copyright.
https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/
#ai #copyright #space
https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/
#ai #copyright #space
Exploring a space-based, scalable AI infrastructure system design
Comments
research.google
November 7, 2025 at 1:59 PM
In space, no one can hear you violate copyright.
https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/
#ai #copyright #space
https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/
#ai #copyright #space
Finally, knowing a thing or two about #quantum #physics allows me to earn some bragging rights on the internet...
#pastpuzzle en-168
🟩🟩🟩🟩 (0)
▪️▪️▪️▪️
▪️▪️▪️▪️
▪️▪️▪️▪️
1/4 🥇
https://www.pastpuzzle.de
#pastpuzzle en-168
🟩🟩🟩🟩 (0)
▪️▪️▪️▪️
▪️▪️▪️▪️
▪️▪️▪️▪️
1/4 🥇
https://www.pastpuzzle.de
past puzzle
Errate das gesuchte Jahr mit Hilfe von 4 historischen Ereignissen. Ein von Wordle und Geschichte inspiriertes Spiel.
www.pastpuzzle.de
November 7, 2025 at 7:12 AM
Finally, knowing a thing or two about #quantum #physics allows me to earn some bragging rights on the internet...
#pastpuzzle en-168
🟩🟩🟩🟩 (0)
▪️▪️▪️▪️
▪️▪️▪️▪️
▪️▪️▪️▪️
1/4 🥇
https://www.pastpuzzle.de
#pastpuzzle en-168
🟩🟩🟩🟩 (0)
▪️▪️▪️▪️
▪️▪️▪️▪️
▪️▪️▪️▪️
1/4 🥇
https://www.pastpuzzle.de
Given the drama surrounding the arXiv for its dependence on Google and its recent ban on review articles in computer science, I have a question:
Do we still need the arXiv?
Wouldn't a federated network of institutional repositories be a much better option?
#arxiv #academicchatter #physics […]
Do we still need the arXiv?
Wouldn't a federated network of institutional repositories be a much better option?
#arxiv #academicchatter #physics […]
Original post on fediscience.org
fediscience.org
November 2, 2025 at 7:30 AM
Given the drama surrounding the arXiv for its dependence on Google and its recent ban on review articles in computer science, I have a question:
Do we still need the arXiv?
Wouldn't a federated network of institutional repositories be a much better option?
#arxiv #academicchatter #physics […]
Do we still need the arXiv?
Wouldn't a federated network of institutional repositories be a much better option?
#arxiv #academicchatter #physics […]
The privacy implications of the new Verification of Payee scheme (mandatory since October for SEPA transfers) are ... interesting. I've been able to obtain the name of the second holder of a bank account without even knowing that there was a second person.
#privacy #banking #sepa #eu #gdpr
#privacy #banking #sepa #eu #gdpr
November 1, 2025 at 3:46 PM
Reposted by Hendrik Weimer
Killer quote:
“I’m going back to the Fediverse. Back to Mastodon. To the nerds, the hobbyists, the idealists. The people who don’t talk about reach, but about relevance. To those who understand that decentralization isn’t nostalgic, it’s the future. That digital sovereignty isn’t a gimmick […]
“I’m going back to the Fediverse. Back to Mastodon. To the nerds, the hobbyists, the idealists. The people who don’t talk about reach, but about relevance. To those who understand that decentralization isn’t nostalgic, it’s the future. That digital sovereignty isn’t a gimmick […]
Original post on mastodon.social
mastodon.social
October 30, 2025 at 11:23 AM
Killer quote:
“I’m going back to the Fediverse. Back to Mastodon. To the nerds, the hobbyists, the idealists. The people who don’t talk about reach, but about relevance. To those who understand that decentralization isn’t nostalgic, it’s the future. That digital sovereignty isn’t a gimmick […]
“I’m going back to the Fediverse. Back to Mastodon. To the nerds, the hobbyists, the idealists. The people who don’t talk about reach, but about relevance. To those who understand that decentralization isn’t nostalgic, it’s the future. That digital sovereignty isn’t a gimmick […]
Large language models are quite useful to perform searches through the literature, so you have a good point to start. So, this is what I just got asking ChatGPT for references for the numerical calculation of the replica correlation lengths in topologically ordered systems:
1. Pizorn & […]
1. Pizorn & […]
Original post on fediscience.org
fediscience.org
October 28, 2025 at 1:30 PM
Large language models are quite useful to perform searches through the literature, so you have a good point to start. So, this is what I just got asking ChatGPT for references for the numerical calculation of the replica correlation lengths in topologically ordered systems:
1. Pizorn & […]
1. Pizorn & […]
If you rely on a for-profit supplier, you should be required to carry the transaction costs to switch away from them as a form of debt in your books. Because the financing cost of this debt is what your supplier is eventually going to charge you.
Allowing to hide this debt is creating some of […]
Allowing to hide this debt is creating some of […]
Original post on fediscience.org
fediscience.org
October 24, 2025 at 11:04 AM
If you rely on a for-profit supplier, you should be required to carry the transaction costs to switch away from them as a form of debt in your books. Because the financing cost of this debt is what your supplier is eventually going to charge you.
Allowing to hide this debt is creating some of […]
Allowing to hide this debt is creating some of […]
Reposted by Hendrik Weimer
"We've created a perfect storm: tools that amplify incompetence, used by developers who can't evaluate the output, reviewed by managers who trust the machine more than their people."
https://techtrenches.substack.com/p/the-great-software-quality-collapse
#vibecoding #ai #softwarequality
https://techtrenches.substack.com/p/the-great-software-quality-collapse
#vibecoding #ai #softwarequality
软件质量的重大崩溃，或我们如何将灾难常态化
The great software quality collapse or, how we normalized catastrophe (techtrenches.substack.com)
10-09 ↑ 163 HN Points
techtrenches.substack.com
October 12, 2025 at 9:36 AM
"We've created a perfect storm: tools that amplify incompetence, used by developers who can't evaluate the output, reviewed by managers who trust the machine more than their people."
https://techtrenches.substack.com/p/the-great-software-quality-collapse
#vibecoding #ai #softwarequality
https://techtrenches.substack.com/p/the-great-software-quality-collapse
#vibecoding #ai #softwarequality
The option price betting that Nvidia's shares will crash below $40 (it's above $180 right now) by December 2027 went up by almost 50% last month, despite the underlying stock gaining over that time period.
Looks like some people have started to run for the exit.
#ai #nvidia #business #economy
Looks like some people have started to run for the exit.
#ai #nvidia #business #economy
October 11, 2025 at 4:23 PM
Reposted by Hendrik Weimer
TIL: apparently one of the main technical hurdles for universities to run their own #mastodon instances is to make the instance work with the institution's SSO (most often Shibboleth). It seems there would be a huge opportunity to facilitate the entrance of such major multiplicators into the […]
Original post on mastodon.social
mastodon.social
October 8, 2025 at 7:47 AM
TIL: apparently one of the main technical hurdles for universities to run their own #mastodon instances is to make the instance work with the institution's SSO (most often Shibboleth). It seems there would be a huge opportunity to facilitate the entrance of such major multiplicators into the […]
Reposted by Hendrik Weimer
@dangillmor As Cory Doctorow repeatedly points out, apps are just websites wrapped in DRM. What about if we just start building websites again? It could be the future.
October 5, 2025 at 7:35 AM
@dangillmor As Cory Doctorow repeatedly points out, apps are just websites wrapped in DRM. What about if we just start building websites again? It could be the future.
Large language models are essentially solutions to nonlinear optimization problems to predict the next token in the output text. Mathematically, these problems are quite similar to problems we routinely face in physics, like finding the ground state of a quantum system.
In these physics […]
In these physics […]
Original post on fediscience.org
fediscience.org
September 29, 2025 at 8:57 AM
Large language models are essentially solutions to nonlinear optimization problems to predict the next token in the output text. Mathematically, these problems are quite similar to problems we routinely face in physics, like finding the ground state of a quantum system.
In these physics […]
In these physics […]
Reposted by Hendrik Weimer
It is good IT Security practice to separate data from code. For example, an SQL Injection attack is to get the target to treat data (entered into a form) as code, that is break that barrier.
Now take an "AI agent" doing a task for you like: "Download that podcast that Frank emailed me about" […]
Now take an "AI agent" doing a task for you like: "Download that podcast that Frank emailed me about" […]
Original post on machteburch.social
machteburch.social
September 26, 2025 at 11:01 AM
It is good IT Security practice to separate data from code. For example, an SQL Injection attack is to get the target to treat data (entered into a form) as code, that is break that barrier.
Now take an "AI agent" doing a task for you like: "Download that podcast that Frank emailed me about" […]
Now take an "AI agent" doing a task for you like: "Download that podcast that Frank emailed me about" […]
Corporate fraudsters love AI because LLMs will fake it if necessary. For a while, company figures will look fine, and when the house of cards collapses, nobody will have to go to jail. Just imagine how easy you can build something like Dieselgate with vibe […]
[Original post on fediscience.org]
[Original post on fediscience.org]
September 7, 2025 at 8:57 AM
Corporate fraudsters love AI because LLMs will fake it if necessary. For a while, company figures will look fine, and when the house of cards collapses, nobody will have to go to jail. Just imagine how easy you can build something like Dieselgate with vibe […]
[Original post on fediscience.org]
[Original post on fediscience.org]
Sorry @brembs, but Open Research Europe is a recipe for disaster. It's centered around the idea of big publishing, i,e., there is one huge contract awarded to a big player who runs the entire platform. Currently, it's Taylor & Francis, but if that changes to one of the other big ones, it will […]
Original post on fediscience.org
fediscience.org
September 4, 2025 at 11:15 AM
Sorry @brembs, but Open Research Europe is a recipe for disaster. It's centered around the idea of big publishing, i,e., there is one huge contract awarded to a big player who runs the entire platform. Currently, it's Taylor & Francis, but if that changes to one of the other big ones, it will […]
Q: What is the largest number that has been factorized using Shor's algorithm on a quantum computer?
A: No number has ever been factorized using Shor's algorithm on a quantum computer.
Even for factorizing 15, the smallest possible number, the modular exponentiation part of Shor's algorithm […]
A: No number has ever been factorized using Shor's algorithm on a quantum computer.
Even for factorizing 15, the smallest possible number, the modular exponentiation part of Shor's algorithm […]
Original post on fediscience.org
fediscience.org
July 15, 2025 at 6:42 AM
Q: What is the largest number that has been factorized using Shor's algorithm on a quantum computer?
A: No number has ever been factorized using Shor's algorithm on a quantum computer.
Even for factorizing 15, the smallest possible number, the modular exponentiation part of Shor's algorithm […]
A: No number has ever been factorized using Shor's algorithm on a quantum computer.
Even for factorizing 15, the smallest possible number, the modular exponentiation part of Shor's algorithm […]
The Ising model in a transverse field is one of the most fundamental models in quantum many-body physics. Despite being around for more than 60 years, we still make new discoveries in it. In our latest work, we show the existence of a dynamical phase […]
[Original post on fediscience.org]
[Original post on fediscience.org]
June 18, 2025 at 8:46 AM
The Ising model in a transverse field is one of the most fundamental models in quantum many-body physics. Despite being around for more than 60 years, we still make new discoveries in it. In our latest work, we show the existence of a dynamical phase […]
[Original post on fediscience.org]
[Original post on fediscience.org]
We've recently moved to @OpenSlides for our internal decision-making processes and it is awesome!
#academicchatter
#academicchatter
June 12, 2025 at 12:05 PM
We've recently moved to @OpenSlides for our internal decision-making processes and it is awesome!
#academicchatter
#academicchatter
Reposted by Hendrik Weimer
@brembs "ORE is implemented until 2026 through an EC public procurement contract with the publisher and publishing service provider F1000 Research Ltd., part of the Taylor & Francis publishing companies."
Meet the new boss, same as the old boss.
Meet the new boss, same as the old boss.
June 5, 2025 at 12:48 PM
@brembs "ORE is implemented until 2026 through an EC public procurement contract with the publisher and publishing service provider F1000 Research Ltd., part of the Taylor & Francis publishing companies."
Meet the new boss, same as the old boss.
Meet the new boss, same as the old boss.
Reposted by Hendrik Weimer
Anyons are quasiparticles that differ from the familiar fermions and boson.
Scientists led by Hanns-Christoph Nägerl have observed anyons in a one-dimensional quantum system for the first time.
The results, published in Nature, may contribute to a better […]
[Original post on social.uibk.ac.at]
Scientists led by Hanns-Christoph Nägerl have observed anyons in a one-dimensional quantum system for the first time.
The results, published in Nature, may contribute to a better […]
[Original post on social.uibk.ac.at]
May 28, 2025 at 3:26 PM
Anyons are quasiparticles that differ from the familiar fermions and boson.
Scientists led by Hanns-Christoph Nägerl have observed anyons in a one-dimensional quantum system for the first time.
The results, published in Nature, may contribute to a better […]
[Original post on social.uibk.ac.at]
Scientists led by Hanns-Christoph Nägerl have observed anyons in a one-dimensional quantum system for the first time.
The results, published in Nature, may contribute to a better […]
[Original post on social.uibk.ac.at]
Reposted by Hendrik Weimer
What's going on when some universities jump more than 950% in one year on metrics used in #university #rankings? Are they gaming the metrics?
https://www.biorxiv.org/content/10.1101/2025.05.09.653229v1
"Key findings include publication growth of up to 965%, concentrated in STEM fields; surges […]
https://www.biorxiv.org/content/10.1101/2025.05.09.653229v1
"Key findings include publication growth of up to 965%, concentrated in STEM fields; surges […]
Original post on fediscience.org
fediscience.org
May 18, 2025 at 5:55 PM
What's going on when some universities jump more than 950% in one year on metrics used in #university #rankings? Are they gaming the metrics?
https://www.biorxiv.org/content/10.1101/2025.05.09.653229v1
"Key findings include publication growth of up to 965%, concentrated in STEM fields; surges […]
https://www.biorxiv.org/content/10.1101/2025.05.09.653229v1
"Key findings include publication growth of up to 965%, concentrated in STEM fields; surges […]
Great news! While Trump is busy dismantling the Library of Congress, the German National Library of Science and Technology (@tibhannover) is preparing for more to come: they've downloaded the entire 10 TB of the arXiv preprint server and safeguarded more than 2.6 million scientific articles […]
Original post on fediscience.org
fediscience.org
May 13, 2025 at 2:41 PM
Great news! While Trump is busy dismantling the Library of Congress, the German National Library of Science and Technology (@tibhannover) is preparing for more to come: they've downloaded the entire 10 TB of the arXiv preprint server and safeguarded more than 2.6 million scientific articles […]