Publishing in the age of AI - a view from the trenches

No posts for a while! I have been overwhelmed with teaching, grant applications, and editorial work. For all the claims that AI is incredible and can automate so much for us, everyone seems busier than ever.

The Physical Review journals are facing a massive increase in submissions - double compared to last year.

A substantial fraction of this increase is due to LLM-assisted (or entirely LLM-generated manuscripts).

To the surprise of some, most of these submissions to Physical Review A are not from "independent researchers," but rather from university-affiliated students and faculty. People who should know better.

The pool of willing referees (already a minority of authors who publish with us) cannot cope with this increase. So editors need to be more selective and reject more papers without external review. Unfortunately for us, LLMs are trained to create content that seems convincing at first glance. And we can't use LLM detectors (which have reliability issues), because the issue is not with the text itself (LLMs are great for polishing, and are now explicitly allowed), but with the (lack of) new and correct physics.

Physical Review A is an archival journal, which means its role is to document substantial research progress, not only the very best results. A researcher may spend months reproducing and then extending a prior result, learning much in the process and laying the foundation for future research breakthroughs.

An LLM model doing the same in a day? No. It will give a result, but you need to understand the result enough to trust it. Which takes time. Publishing unverified material corrupts the scientific record. It is the author's responsibility to be sure what they are publishing is correct.

To be a good writer, read what great authors have written.

To be a good researcher, read about good research. The original material, not the AI summary.

The smoking gun of an AI paper is that the writing is off. Out of place jargon. Unusual article structure. Too many tables when a figure would be much clearer. An overly authoritative tone. Dozens of propositions and theorems in a physics paper where the deepest maths is just diagonalising a small matrix. Seeing these papers, it becomes apparent that the supposed author has not read a single paper in the field in which they are trying to publish. 

Topological Photonics: Fundamentals and Applications

After many years of effort, our edited book "Topological Photonics: Fundamentals and Applications" is close to being published! This was a long and arduous process with many steps: Preparing a proposal, soliciting contributing authors, chasing for overdue drafts, reviewing them, dealing with copyright issues, and carefully checking the final proofs. 

We started this project back in 2022, when large language models were only just starting to emerge. Four years later, many will ask whether books are now obsolete in this age of AI, when any question can be seemingly answered in an instant by a simple prompt.

In my opinion, physical books are more important than ever.

Nowadays, information is cheap but time is limited. With a book, you have the reassurance that someone thought the content was valuable enough to put it into print. You actually own something, not a license that can be revoked without warning. You are also insulated against the temptation to switch to another tab or check your favourite social media app. 

Academic books play an important role in sustaining research. They serve as important and reliable references for graduate students just starting in a field. For mature research fields, they help in distilling the vast primary literature down to the key results that really matter in hindsight. Casual browsing of a well-stocked university library allows one to be exposed to fields and ideas outside one's own narrow area of expertise, the driver of serendipitous discoveries.

The ASEAN Plus Photonics Symposium 2026

I really enjoyed attending The ASEAN Plus Photonics Symposium 2026 earlier in April. I was invited to give a talk on nonlinear topological photonics. My talk slides are available here.

The conference programme covered a broad range of topics, including metasurfaces, exciton-polaritons in microcavities, structured light, topological photonic crystals, and the integration of two-dimensional materials into photonic devices. The full programme is available here.

Beyond the individual talks, a few high-level themes stood out to me:

A recurring topic was the push to move metasurfaces from proofs of concept toward practical devices. This shift raises its own set of challenges: how can fabrication be made faster, cheaper, and scalable? Similar questions arise for device characterisation, which must evolve from carefully controlled, fragile experiments to something closer to a production line approach. Efficient electrical control of devices is increasingly important, as is long-term stability: can one return to a device six months later and expect the same performance?

Another strong theme was the importance of active exchange between theory and experiment. For this to be effective, theorists need to understand which effects are experimentally accessible and which are not, while experimentalists must know when to trust theoretical predictions, and when those predictions need to be revisited in light of conflicting data.

One particularly striking example came from Prof. Feng Li's talk on optical skyrmions. He candidly talked about how early experimental observations in the 2010s were not well understood and lacked an adequate theoretical explanation. This kind of openness about the inevitable challenges of research elevates a good talk into an inspiring one.

Related to this was a recognition of the value of intermediate results. Such results are often essential for understanding the capabilities and limitations of new fabrication schemes, or for training new students. While modest in scope, these studies can sometimes lead to unexpected and rewarding discoveries.

The conference venue itself was memorable: located by the beach and decorated with artworks depicting various Physics Nobel Laureates, it provided a pleasant and stimulating setting for discussion.

There was also noticeable attention on the rise of Vietnam as a manufacturing hub. As costs increase in China, manufacturing activity is shifting, and Vietnam is seeing growing investment in semiconductor fabrication. This trend is driving demand for scientists and engineers with expertise in nanofabrication and related areas, accompanied by increased funding for research into integrated photonics.

I do like smaller conferences such as this. With no parallel sessions, you can better exposed to topics outside one's own direct area of research, broadening knowledge and getting to know colleagues you wouldn't normally interact with.

Looking forward to the next edition of the conference in 2027!

Cusp solitons mediated by a topological nonlinearity - part 2

For an introduction see part 1, which I wrote when our paper was uploaded to arXiv.

After a few rounds of review our paper was published in Chaos last week!

With so much negativity surrounding peer review nowadays, it is important to highlight when it works well. Our original manuscript mentioned a few possible experimental realizations of our new model, without going into specifics - it seemed like a hard problem which could not be solved before Harvey was due to graduate. So we left it for future work.

The referees correctly pointed out that a vague discussion of possible implementations is not very helpful - this hard problem is then delegated to the reader. Could we provide specifics?

We could not at the time of writing the manuscript. But the referee comments inspired (forced?) us to try harder. During our search we stumbled upon a very nice experiment published in Physical Review Letters: Nonlinear Non-Hermitian Skin Effect and Skin Solitons in Temporal Photonic Feedforward Lattices

We realized that the feedforward nonlinearity used in this work could be readily adapted to realize our model. We added the recipe to our manuscript as an appendix, giving strong evidence for the experimental feasibility of observing our results!

Peer review isn't meant to be just a rubber stamp or a hard rejection; it should also be a way to help present the research in the best possible way. While "open science" and "open peer review" are trending topics, anonymity gives referees the safety to ask hard (or annoying) questions. 

Call for Nominations: CSC-SUTD Joint Scholarship

The China Scholarship Council and Singapore University of Technology and Design jointly sponsor outstanding PhD students to visit SUTD for an extended period (6-24 months) to carry out joint projects. The China Scholarship Fund provides for a round-trip international trip and a scholarship within the funding period. Scholarships are used to subsidize the basic study and living expenses while in Singapore.

Interested applicants should first contact a prospective faculty supervisor at SUTD to secure a nomination and letter of invitation before 1st March, 2026. More information is available here.

Haldane on the second quantum revolution

This week I attended a great public lecture by Duncan Haldane: "Quantum Mechanics After One Hundred Years, and the 'Second Quantum Revolution' Today"

Starting from the discovery of quantum mechanics, he explained how the concept of quantum entanglement is fueling today's second quantum revolution and its connection to his Nobel Prize-winning work.

Haldane remarked that his work on quantum spin chains was controversial. He had theorists accosting him at conferences arguing he was wrong. These kinds of disputes among theorists are best settled by experiment. Undoubtedly, Haldane would not have received his Nobel Prize if his predictions had not been validated by experiments. How can you motivate some experimental group to be interested in your theory? If it generates controversy!

Similarly, experiments often are the drive for fresh theoretical advances. For example, the experimental discovery of the quantum Hall and fractional quantum Hall effects came before the theoretical predictions or understanding. 

A good example of this is a second important work by Haldane also cited in his prize: quantum Hall effects in absence of Landau levels. This now-seminal work went largely unnoticed for a decade, because the model based on a two-dimensional honeycomb lattice seemed unfeasible to realize in an experiment. Later, the unanticipated work experimental isolation of graphene drove theorists to this fresh area. Haldane's early theory work was recognised as the foundation for the discovery of time reversal-symmetric topological insulators and the whole "zoo" of topological materials that followed.

Haldane also emphasized the importance of luck in making ground-breaking discoveries. von Klitzing was not the first person to attempt quantum Hall measurements, but previous attempts had used a different experimental setup: varying current with a fixed magnetic field. Imperfections in the current source led to fluctuations in the measured resistivity, which seemed to be consistent with previous approximate theoretical calculations based on perturbation theory. von Klitzing's approach of measuring resistivity as a function of magnetic field strength, with current kept fixed, led to unexpectedly precise quantization which needed new theory to explain. 

Haldane's take-home message was thus: anyone can win a Nobel Prize, but you need luck and the perseverance to defend your work if it is challenged.

An earlier iteration of this talk is available here. A more detailed write-up is available here. 

   

Quantum Computing Summer School at Los Alamos National Laboratory

The Quantum Computing Summer School is an immersive 10-week curriculum that includes tutorials from world-leading experts in quantum computation as well as one-on-one mentoring from Los Alamos National Laboratory staff scientists who are conducting cutting-edge quantum computing research. Summer school fellowship recipients will be exposed to the theoretical foundations of quantum computation and will become skilled at programming commercial quantum computers, such as those developed by IBM, Quera, IonQ, Quantinuum, DWave. All students (undergraduate and graduate) are encouraged to apply.  

In the first 2 weeks, students will attend lectures given by world-leading experts – from academia, industry and national laboratories – in quantum computing research. Following the 2-week lecture period, each student will work on a research project in quantum computing for the remaining 8 weeks. For this research project, each student will be paired with a LANL mentor who will propose project topics and provide guidance. Each project will involve some hands-on programming of a quantum computer (IBM’s, Quantinuum’s, Quera, D-Wave’s, as available). If time permits, the students will begin preparing their results for publication.

This is an fantastic opportunity with so many of the student projects delivering important findings on various hot topics related to quantum computing. This is a testament to the quality of the mentorship provided by the staff scientists involved in the school.

More information including how to apply is available here. The application deadline is January 11th, 2026.