Leiden Declaration on Artificial Intelligence and Mathematics

Signatories

The primary signatory list now lives on the homepage below the declaration text.

Sign Declaration
3133 Signatories
  1. Longyun Ding ORCID Nankai University
  2. Carlos Rodriguez Max Planck Institute for Mathematics in the Sciences
    Comment

    I agree with the Leiden declaration broadly. Mathematical ideals and theorems in math papers all carry a history behind them, and it's beneficial to the reader when people acknowledge conversations or ideas (e.g. acknowledging a colleague for mentioning a work or a book that turned out to be very fruitful). Disclosing the use of LLMs, I think, is similar in spirit to this. Of course, every theorem in a math paper that you author should be read and trusted by the authors!

  3. Xiaofa Chen ORCID School of Mathematical Sciences, University of Science and Technology of China
  4. Stephen Huggett ORCID
  5. Emily Norton ORCID University of Kent
  6. Xiaojin Lin ORCID Postdoc, Tsinghua University
    Comment

    Agree with the Declaration.

  7. Noam Zeilberger ORCID Assistant professor, Institut Polytechnique de Paris
  8. Anja Meyer verified email University of Manchester
  9. Dennis Nieman ORCID Postdoctoral researcher
  10. Linxuan Li ORCID PhD, Queen Mary University of London
  11. Bao Tang ORCID University of Graz, Austria
  12. Pedro R. López-Gómez ORCID Assistant Professor, CUNEF Universidad
  13. Dario Mazzoleni verified email University of Pavia
  14. Steffen Müller verified email University of Groningen
  15. Charles Garnet Cox ORCID Academic, Senior Lecturer, University of Bristol
  16. Simon Dieck ORCID PhD Candidate, TU Delft
  17. Sanbiao Dong ORCID Student, Southwest Jiaotong University
  18. Gionni Marchetti ORCID CEO, Postillón XAI
  19. Hao Chen ORCID Assistant Professor, ShanghaiTech University
  20. Leyi Jiang ORCID Max Planck Institute of Molecular Cell Biology and Genetics
  21. Philip Simon Schlösser ORCID PhD Candidate, Radboud University Nijmegen
  22. Sanming Zhou ORCID Professor, The University of Melbourne, Australia
  23. Wadim Zudilin ORCID Professor, Radboud Universiteit Nijmegen
  24. Devarshi Mukherjee verified email University of Oxford
  25. Lahcen Maniar Cadi Ayyad University MarrakechMorocco
  26. María José Pereira-Sáez ORCID Dpt. of Economics, Universidade da Coruña
  27. Qian Chen ORCID Xi'an Jiaotong University
  28. Luca Descheemaeker ORCID Vrije Universiteit Brussel
  29. K Sreeram Menon Student, IISER Bhopal
  30. Alexander Schmeding ORCID Associate professor (førsteamanuensis), Norwegian University of Science and Technology
  31. Michael Rothgang ORCID Postdoctoral researcher, University of Bonn
  32. Tim Oosterwijk ORCID Assistant professor, Vrije Universiteit Amsterdam
  33. Sachin Chauhan ORCID Postdoctoral Researcher, Uppsala University
  34. Etienne Boursier ORCID Inria, Université Paris-Saclay
  35. Holger Reich verified email Freie Universität Berlin
  36. Matthias Walter ORCID Assistant Professor, University of Twente
  37. María Ángeles Gil ORCID UNiversidad de Oviedo, Spain
  38. Raffaele Vitolo ORCID professor, Università del Salento
  39. Thomas Blom Postdoctoral fellow, Max Planck Institute for Mathematics
  40. Gaspard Jankowiak ORCID Senior scientist, University of Graz
  41. Boris Doubrov ORCID Belarusian State University
  42. Zheng WANG ORCID Doctorant, Laboratoire Analyse, Géométrie et Applications
  43. Krzysztof Oleszkiewicz ORCID University of Warsaw
  44. Fernando Pestana da Costa ORCID Professor, Universidade Aberta
  45. Jakob Führer ORCID Institute for Algebra, Johannes Kepler University of Linz
  46. Todor Antić ORCID PhD student, Charles University
  47. Jashandeep Thind verified email Indian Institute of Science
    Comment

    Amongst the many ethical concerns that come from the corporations that have created these tools, the ethics of where these data centres are put, how much resources they steal from the local population should be of great concern. So must the light be on what goes into training these models. The hours and hours of manual labour put in by marginalised community workers in procuring data for these models. This mass exploitation of natural and human resources must be kept in mind whenever one is using these LLMs or other generative AI.

  48. Ana Martinez-Pastor ORCID Full professor, Universitat Politècnica de València
  49. Davide Barbieri ORCID Universidad Autonoma de Madrid Departamento de Matematicas
    Comment

    I think we mathematicians should also try to discuss in public about 2 other aspects of AI proficiency in mathematics. First is the reason why we have this: the main motivation for the industry was the control of hallucinations, which are the main technical drawback of LLM preventing them to be used in full automation in sectors such as healthcare, but also finance and the military. This is a path that, if it works, will produce more illness diagnoses entirely made by machines, more activities in the financial markets entirely made by machines, more war actions entirely made by machines. I am happy as a mathematician to see that this strategy is not working (for the moment), but I would add this to the ethical aspects of our giving contributions and feedbacks to this whole process. Second is the issue with young minds. Mathematics is also struggle, failure and stepping out of comfort zones. As such, it is a practice of humbleness, of consideration of implications of our statements (and actions), and ultimately of development of a critical mind. If we remove the struggle and the failure during the learning, because we provide continuous machine support to each young mind who first encounters mathematical problems: will it change something? Should we mathematicians, who have had a traditional education, occupy part of our time in thinking how to allow the people who are currently in ages 6-23 to have access to the broader formative aspects of mathematics, that may be greatly reduced by an "intuitive" use of AI? Young students asking an LLM how to solve a mathematical problem without doing the effort, may well learn how to solve that mathematical problem, but the overfitting with this procedure is extremely higher, and the struggle gets close to zero. Isn't this a societal long term risk too?

  50. Amar Yumnam
  51. Devi Prasad verified email Indian Institute of Science
  52. Marta Sanz-Solé ORCID Professor emeritus , University of Barcelona
  53. Emanuele Prati verified email University of Padua
  54. Adrian Detavernier ORCID Faculteit Ingenieurswetenschappen en Architectuur, Universiteit Gent
  55. Chen Qian verified email Xi'an Jiaotong University
  56. Raphael Appenzeller ORCID Heidelberg University
  57. Alessandro Della Corte ORCID Professor (Associate) of Mathematical Physics, Università degli Studi di Camerino
  58. Nikolaos Panagiotopoulos ORCID On Board Computers and Data Handling Engineer, European Space Research and Technology Centre
    Comment

    My affiliation does not reflect the ESA's viewpoint on AI. Furthermore, I would like to add that the current technology which has been baptised as "AI" is simple a group of algorithms (very good ones though) and are able to automate tedious mental work with high risks and error margins. It requires diligent investigation and background work (here is where the human intelligence is required most) before it is implemented on real world applications, in order to bring the desired benefits, always in the frame of logic and ethics.

  59. David Evans Prifysgol Caerdydd - Cardiff University
  60. Elima Shehu ORCID MathCoRe Postdoc, Otto-von-Guericke University Magdeburg
  61. Annegret Burtscher ORCID Associate Professor, Radboud University Nijmegen
  62. Alessandra Cattafi verified email University of Pisa
  63. Jiangyu Su verified email Fudan University
  64. Julien BREMONT ORCID Collège de France
  65. Hyeonwoo Kim verified email Pohang University of Science and Technology
  66. Leon Kester Transdisciplinary Scientist, TNO
    Comment

    Thank you so much for this initiative. I fully support the declaration and recommendations. I hope this and other initiatives evolve towards a transdisciplinary declaration on science and philosophy.

  67. André Schlichting ORCID Full Professor, Universität Ulm
  68. Johannes Lederer ORCID Chair of Data-Driven Mathematics at the University of Hamburg
    Comment

    No one knows what AI will bring tomorrow. But as mathematics community, we can—and should—take part in shaping that future.

  69. Alberto Gervani verified email Scuola Normale Superiore di Pisa
  70. Ahmed Ratnani ORCID Director of The UM6P Vanguard Center, Université Mohammed VI Polytechnique
  71. Ingemar Kaj ORCID Uppsala University
  72. Jireh Loreaux ORCID Associate Professor, Southern Illinois University Edwardsville
    Comment

    While there any many further steps to be taken, the Leiden declaration is a good starting point.

  73. Rémy Degenne ORCID Researcher, Inria Centre at the University of Lille
  74. Giuseppe Alessio D'Inverno ORCID Postdoctoral Researcher, Laboratoire de Mathématiques d'Orsay, Université Paris-Saclay
  75. GIUSEPPE MOLTENI ORCID associate professor, University of Milan
  76. Mingcheng Wu ORCID Master Student, Shandong University
    Comment

    I strongly support the Leiden Declaration on AI and Mathematics. Preserving human authorship, transparency, and deep understanding in mathematics is crucial.

  77. Stephan Wagner ORCID TU Graz and Uppsala University
  78. Axel Flinth verified email Umeå University
  79. Yuriy Drozd ORCID Institute of Mathematics National Academy of Sciences of Ukraine
  80. Olivier Hénard ORCID Assistant Professor (MCF), Université Paris Saclay
  81. Hendrik De Bie ORCID full professor, Ghent University
  82. Georgios Skantzaris verified email Eindhoven University of Technology
  83. Jess Stubbs ORCID Data Engineer (Student)
    Comment

    What I appreciate about this declaration is that it's not a tantrum. It's not "AI bad, stop everything." It's a precise community with a precise relationship to truth, asking a precise question: what happens to knowledge when the infrastructure of verification gets undermined faster than anyone can adapt? I just hope we're not too late...

  84. Matthew Bisatt verified email Adam Mickiewicz University of Poznan
  85. Perdita Stevens ORCID University of Edinburgh, Prof. Em. Mathematics of Software Engineering
  86. Evgeny Khukhro ORCID Emeritus Professor of Pure Mathematics, University of Lincoln, UK
  87. Bart de Keijzer ORCID Department of Informatics, King's College London
  88. Shouxi Wang verified email University of Warwick
  89. Carlo Mantegazza ORCID Math. Department, University of Naples Federico II
  90. Balasubramanian Sury ORCID
  91. Anton Fehnker ORCID PhD Student, University of Copenhagen
  92. Julian Berman ORCID Columbia University
  93. Raul Rojas ORCID University of Nevada Reno
  94. Florian K. Richter ORCID École Polytechnique Fédérale de Lausanne (EPFL)
  95. Danko R. Jocic verified email University of Belgrade
    Comment

    Every result in Mathematics should be categorized as (one of these): - HUMAN verified (standard) or - AI verified. Alternatively, if not verified by humans, every AI claim in Mathematics should be classified as: AI Mathematics. D.J.

  96. Juliet Floyd ORCID Borden Parker Bowne Professor of Philosophy, Boston University; Director, Boston University Center for the Humanities
    Comment

    We need such dialogue among professionals and others about our future with AI.

  97. Jan Draisma ORCID Universitat Bern
    Comment

    I find it amazing how such relatively simple models for the brain perform so well at tasks so far thought to be doable by humans only. This includes mathematical reasoning, and I am sure that these systems are here to stay. But on how they should be used in mathematical research, I wholeheartedly endorse the values and recommendations formulated out in this declaration. (Among the very real threats, I found that cognitive offloading might have been also included more prominently. Just like it is now very tempting for students to ask an LLM to do their homework, it will also be increasingly tempting for research mathematicians, and researchers in other areas, to delegate to a machine thinking that they would previously have done themselves. I am concerned about the long-term effects that this will have on our minds, and on humanity as a whole.)

  98. Michele Caselli ORCID Postdoctoral Researcher, The University of Sydney
  99. Bian Yang verified email Norwegian University of Science and Technology
    Comment

    A good initiative! But I'd suggest we use a more open and collaborative attitude towards industry and government. "Don’t believe the hype", "Regulate the artificial intelligence industry", ... maybe we can use better phrases (less emotional, more scientifically evidenced) to encourage the stakeholders to talk and find a solution together.

  100. Nicola Ciccoli ORCID University of Perugia