Sign in

Dmitry Grinko

@dmitry-grinko.bsky.social
398 followers 323 following 39 posts

quantum stuff @ QuSoft, University of Amsterdam & CWI

PostsRepliesMedia
Reposted by Dmitry Grinko
Maris Ozols @marisozols.bsky.social · 9h
You may finally rest easy at night since SICs exist in all dimensions! github.com/marozols/SIC... SICs are d² unit vectors in d complex dimensions with pairwise equal inner products. Their existences in all dimensions d was conjectured by Zauner in 1999. en.wikipedia.org/wiki/SIC-POVM
Diagram illustrating Lean proof of SIC existence
2103
Reposted by Dmitry Grinko
Thomas Dietterich @tdietterich.bsky.social · 08/10/2026
I suggest that the right attitude to take toward the OpenAI math release is to imagine we have discovered a notebook full of results from a deceased mathematician. Our job is to make sense of the notebook.
1241
Dmitry Grinko @dmitry-grinko.bsky.social · 09/10/2026
Source: youtu.be/iRjasLnigsU?... Reactions of Hugo and others are here: proofsandprompts.com/2026/10/08/1...
proofsandprompts.com
100+ reactions to 100+ solutions
“We are all grappling with the announcement of 6 October. We think that it is useful to have a snapshot of the community’s gut feeling today, and this is what you see below. We will keep it o…
010
Dmitry Grinko @dmitry-grinko.bsky.social · 09/10/2026
This is the definition of MEGASCOOP :(
100
Dmitry Grinko @dmitry-grinko.bsky.social · 06/10/2026
Hilbert’s 12th problem for real quadratic fields is solved and thus SIC-POVM problem (Zauner's conjecture) is resolved! www.danrad.net/papers/stark...
danrad.net
2175
Reposted by Dmitry Grinko
Dulwich Quantum Computing @dulwichquantum.bsky.social · 23/09/2026
The next logical step would be a week-long single-author conference with parallel sessions and 100 AI bots of the same author in the audience. www.washingtonpost.com/education/20...
washingtonpost.com
Using AI, a professor wrote 200 papers this year. Researchers are alarmed.
An online platform removed 257 papers written or co-written by the University of Chicago’s Nicholas Polson after other academics raised questions.
0124
Reposted by Dmitry Grinko
Maris Ozols @marisozols.bsky.social · 22/09/2026
Here are my immediate thoughts I sent to the new Advisory Group on Mathematics and Artificial Intelligence. terrytao.wordpress.com/2026/09/21/a...
terrytao.wordpress.com
Announcing the Advisory Group on Mathematics and Artificial Intelligence
[This is a guest post by the Advisory Group on Mathematics and Artificial Intelligence. This blog post was initially written in a different file format and converted using AI. — T.] We would …
183
Reposted by Dmitry Grinko
Terence Tao @teorth.bsky.social · 11/09/2026
A group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...
mathandai.org
Declaration — Math and AI
Read the declaration and add your name.
422075934
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
But! It seems that OpenAI will train on your data anyway according to this statement from OpenAI exec (am I reading this correctly?): x.com/markchen90/s...
090
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
If you turn it off in "Settings"->"Data controls"->"Improve model for everyone", it is not enough, go to "Learn more" and Privacy center there. It is kinda sneaky from OpenAI, they should not make it that tricky! It can also explain avalanche of concurrent results more generally... 2/3
181
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
One could try to explain Navier-Stokes drama via hypothesis that Levent or Tristan did not properly turn off data-sharing in ChatGPT. It is not enough to turn off "Improve the model for everyone" in Data controls settings, you need to submit it via "Privacy center". 1/3
1153
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
To clarify, by "a vague mathematical idea is not valuable" I meant "a vague mathematical idea does not deserve official credit"
011
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
Similar thing will unavoidably happen with other IP/patents soon. What does it mean for the law? Should we abandon the old principle and make ideas patentable? Or should we simply abandon any form of IP altogether? 3/3
001
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
"A vague mathematical idea is not valuable, until concrete implementation (in the form of the scientific manuscript) is in place" - this principle feels not valid anymore, since vague ideas can be now immediately turned into highly sophisticated manuscripts automatically. 2/3
211
Dmitry Grinko @dmitry-grinko.bsky.social · 08/09/2026
Based on my naive understanding of law, there is a general principle that “ideas are not patentable, only concrete technical implementations of ideas may be.” You can see how the mathematical analogue of this principle is breaking down in the Navier-Stokes drama. 1/3
120
Reposted by Dmitry Grinko
Alexander Stottmeister @stotti-alex.bsky.social · 04/08/2026
Finally, there is a reason to use “we” in a single-author paper.
0282
Dmitry Grinko @dmitry-grinko.bsky.social · 04/08/2026
To be clear, Lee-Yang results are AI-assisted, not AI-generated. Main ideas came from humans
010
Reposted by Dmitry Grinko
Henry Yuen @henryyuen.bsky.social · 01/08/2026
Some initial thoughts, and a complicated mix of feelings. Wow. I mean, Erdos problems are cool (I genuinely mean that), I didn't know about the Jacobian conjecture before it got disproved. But this newest batch from OpenAI hits home in a way the previous announcements did not.
233072
Dmitry Grinko @dmitry-grinko.bsky.social · 01/08/2026
Such an advance was to be expected (especially to those who have been playing with AI tools over the last 10 months), but still it is quite mind-blowing to see this rate of progress happening in "real time": openai.com/index/ten-ad...
openai.com
Ten advances in mathematics and theoretical computer science
OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.
041
Dmitry Grinko @dmitry-grinko.bsky.social · 31/07/2026
If you are curious about details, Ludovico wrote an excellent thread here x.com/LamiLudovico... 2/2
000
Dmitry Grinko @dmitry-grinko.bsky.social · 31/07/2026
We recently strengthened Generalised Quantum Stein's Lemma :) arxiv.org/abs/2607.24712 Thank you, Ludovico @ludovicolami.bsky.social, for such a wonderful collaboration, this was an incredibly fun project to work on! 1/2
scirate.com
140
Dmitry Grinko @dmitry-grinko.bsky.social · 29/07/2026
This is not a good idea, as relatively few people in our community have enough experience with Lean. Even when a proof type-checks, human review is still needed to ensure the theorem formalized accurately reflects what we intend to prove. Reading Lean code is a skill that not everyone has yet.
120
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
Spectral gap of Lee-Yang Hamiltonians: 1) arxiv.org/abs/2607.10765 2) arxiv.org/abs/2607.14401 9/9
120
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
The problem of Unclonable encryption was claimed by these two teams: 1) arxiv.org/abs/2607.21551 2) eprint.iacr.org/2026/1509 8/n
120
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
The problem of 2-copy distillability of Werner states was claimed by these 3 different teams: 1) arxiv.org/abs/2607.21367 2) arxiv.org/abs/2607.23416 3) arxiv.org/abs/2607.24309 4) arxiv.org/abs/2607.24479 7/n
150
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
For context, I had these 3 very recent results in mind: the 2-copy distillability of Werner states, unclonable encryption, spectral gap of Lee-Yang Hamiltonians. I expect many more such cases soon. There is going to be a tsunami of such results. 6/n
120
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
The last point is actually very troubling: there is a huge inequality in the access to frontier models right now. Some people are simply not aware of this AI revolution, some cannot afford paying 200$ per month, especially in less developed countries (and even in developed!). 5/n
261
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
This is very important to resolve adequately, especially for young students. Their work could be in danger of being scooped by quick AI prompting from more experienced researchers, who know more context about given problem, and who have access to more compute/tokens. 4/n
141
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
These are urgent organisational questions now, and we should discuss them openly as community publicly. During this turbulent transitional period in research practices, we must find a good way to separate human-generated technical work from AI-generated for this year QIP. 3/n
220
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
Should PC merge the AI solutions for a given problem from all teams (no matter how many claimed the result) into one talk? Should we have a separate track specifically for the submissions where most of the key work was done by AI? (to align incentives this track e.g. may have higher accept. rate)2/n
120
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
A question for quantum info community and QIP program committee: how would you review AI-generated solutions to important problems in our field? There are already examples of very strong results which several teams claimed simultaneously. 1/n
382
Dmitry Grinko @dmitry-grinko.bsky.social · 28/07/2026
Great news, Felix! Big congratulations!!
030
Dmitry Grinko @dmitry-grinko.bsky.social · 27/07/2026
From this point of view, the first question has an easy answer. However, I am not sure what to say about questions 2 and 3. Any thoughts? (4/4)
000
Dmitry Grinko @dmitry-grinko.bsky.social · 27/07/2026
Who gets credit for Z? One possible answer is that credit assignment should reflect human time spent: if Bob carefully verifies and disseminates the knowledge from LLM, he deserves credit. Of course, Bob has to acknowledge Alice's original question Q, which inspired his efforts. (3/4)
110
Dmitry Grinko @dmitry-grinko.bsky.social · 27/07/2026
The first question is motivated by 3 real-world examples Robert gave earlier. There is the same core question: Alice presents a new scientific question Q (not fully public yet); Bob either submits Q directly or asks LLM a related Q′ and quickly finds new result Z unknown to Alice. (2/4)
100
Dmitry Grinko @dmitry-grinko.bsky.social · 27/07/2026
A very timely post by Antonio Mele (x.com/QuAntonioMel...) about our recent experiences with quantum info research, along with a great talk by Robert Huang (youtube.com/watch?v=vT5b...). Robert asks the following 3 questions: (1/4)
120
Dmitry Grinko @dmitry-grinko.bsky.social · 23/07/2026
Just to be clear, I do not know if it is a good or bad thing ultimately. I am just trying to make the logic of the social contract (or rather my understanding of it) behind science funding more explicit.
230
Dmitry Grinko @dmitry-grinko.bsky.social · 23/07/2026
So if we as society value humanity progress above all, it is inevitable that usage of AI will be encouraged and will practically become mandatory for scientists. I do not see the way out of this logic.
140
Dmitry Grinko @dmitry-grinko.bsky.social · 23/07/2026
By the same logic, scientists should welcome AI tools if they genuinely want to accelerate the production of new knowledge and understanding.
120
Dmitry Grinko @dmitry-grinko.bsky.social · 23/07/2026
I guess, when science is "sold" to the public, the argument is usually that it advances knowledge for society or humanity as a whole. From that perspective, "I enjoy doing science" is a somewhat selfish argument, even though in practice it's one of the main reasons people become scientists.
250
Dmitry Grinko @dmitry-grinko.bsky.social · 24/05/2026
Congratulations, Marco!!
120
Dmitry Grinko @dmitry-grinko.bsky.social · 23/05/2026
This is simply how it works now: hiring committees do pay a lot of attention to citations, h-index etc. Of course it is not a good measure, but it is a kind of proxy
010
Dmitry Grinko @dmitry-grinko.bsky.social · 21/05/2026
After looking at the solution more closely, it is clear the model did not invent new concepts. That gives a little hope for humans (for the time being). AI is incredibly good at connecting existing areas/concepts, but still fails miserably at finding/creating entirely new ones.
050
Dmitry Grinko @dmitry-grinko.bsky.social · 20/05/2026
In light of today’s news, it looks like the only remaining human edge is the ability to pose interesting questions/define problems/set research directions etc. How long will it take for AI systems to capture that ability too?
160
Dmitry Grinko @dmitry-grinko.bsky.social · 20/05/2026
With AI lowering the cost and efforts of generating knowledge, what becomes the main academic currency of the future? If citations and h-indexes lose meaning for hiring committees, what replaces them? Curious to hear what people are thinking. Share your thoughts :)
340
Dmitry Grinko @dmitry-grinko.bsky.social · 10/05/2026
Academic careers are mostly built on the idea of prestige. Solve an old important problem, build a powerful theory, design a new clever experiment, and you earn prestige. It is quite clear now that this currency will collapse. What will replace it, if anything?
181