6th Youth in High-Dimensions: Recent Progress in Machine Learning, High-Dimensional Statistics and Inference & How creative is Generative AI? Perspectives from Science and Philosophy

Description

An ICTP Hybrid Meeting

How creative is generative AI: 3 - 5 July 2025

Youth in high dimensions: 7 - 9 July 2025

The ICTP-IBM Brahmagupta AI Prize for Young Scientists: 10 July 2025

In the first part of this activity, we will discuss to what extent generative artificial intelligence systems that create text or images like ChatGPT or Dall-E are "creative’. To this end, we will bring together a diverse group of experts from the sciences and humanities. Here is the website set up for the AI & Creativity event? http://aicreativity25.github.io/

Speakers

Alice Barale, Università di Milano, Italy
Larissa Berger, Uni Siegen, Germany
Sergio Canazza, Centro di Sonologia Computazionale, Italy
Alexandre Darmon, Art in Research, France

Alessandro De Cesaris, Università di Torino, Italy
Alessandro Fiordelmondo, Università di Padova, Italy

Alexandra Gilliams, Université Paris 1 Panthéon-Sorbonne, France
Mason Kamb, Stanford, USA

Roberto Trotta, SISSA, Italy

Ardavan Vossoughi, Università di Padova, Italy

The second part of this activity will see the sixth edition of “Youth in high dimensions”, where we give young researchers from academia and industry the opportunity to gather and present recent results on high-dimensional statistical problems, in areas ranging from machine learning and inference to neuroscience or statistical physics. Typical problems in high-dimensions include estimating noisy signals, analysing (deep) neural-networks, detecting communities in large networks, or modelling disordered spin systems.

This conference aims at reinforcing the links among the interdisciplinary community of scientists working on machine learning, and to connect them with philosophers, artists and sociologists to reflect about the impact and role of this technology, and to bring forward the latest development happening in the high-dimensional world.

Youth is a satellite meeting of Statphys2025 https://statphys29.org/

Topics

High-dimensional statistics

Statistical physics

Machine learning

Computational neuroscience

Complex systems

Machine learning for science

Statistical inference

Speakers
Lorenzo Bardone, SISSA, Italy

Freya Behrens, EPFL, Switzerland

Simone Bombari, ISTA, Austria

Elizabeth Collins-Woodfin, University of Oregon, USA

Alex Damian, Princeton, USA

Hadi Daneshmand, MIT, USA

Daniil Dmitriev, ETH Zurich, Switzerland

Margalit Glasgow, Stanford, USA

Erin Grant, Gatsby, UK

Brice Huang, MIT, USA

Bingbin Liu, CMU, USA

Vidya Muthukumar, Georgia Tech, USA

Mauro Pastore, ICTP, Italy

Reese Pathak, UC Berkeley, USA

Stefano Sarao Mannelli, Chalmers, Sweden

Abishek Shetty, MIT, USA

Jingfeng Wu, Berkeley, USA

The following activity to be held from 10 to 18 July 2025 may also be of interest for you: Advanced School on Foundation Models for Scientific Discovery

Research abstracts: Applicants are encouraged to submit an abstract for a poster presentation at the “Youth in high dimensions” activity. During the application, please make sure to use our templates to format your abstract. Templates are available below for download.

Contributed talk: The possibility of giving a contributed talk will be offered to a selection of people who will present a poster.

Participation: If you don't need any financial support and/or visa, three activities can be followed separately.

Grants: A limited number of grants are available to support the attendance of selected participants, with priority given to participants from developing countries. There is no registration fee.

The ICTP-IBM Brahmagupta AI Prize for Young Scientists Ceremony will be held on 10 July 2025.

For more information on the ceremony and winner, please visit: https://indico.ictp.it/event/11057/

Material

Abstract Templates BOOKLET Post.Session CT Gen AI Booklet CT Youth Booklet Final list of Participants Group photo Poster

6th Youth in High-Dimensions: Recent Progress in Machine Learning, High-Dimensional Statistics and Inference & How creative is Generative AI? Perspectives from Science and Philosophy | (smr 4084)

Go to day

Thursday, 3 July 2025

08:30 - 09:00 Registration & administrative formalities

09:00 - 11:00

Chair: Fabiola Ricci (SISSA)

09:00 Opening remarks 40'

Speaker:	Sebastian Goldt (SISSA), Elisa Gremmo (Unipd)
Material:	Video

09:40 An experiment on beauty and rationality 5'

Speaker:	Matteo Marsili (ICTP)
Material:	Video

09:45 Artificial Optics: AI Art as Phantasmagoric Spectacle 45'

Speaker:	Alexandra Gilliams (Université Paris 1 Panthéon-Sorbonne)
Material:	Video

10:30 Coffee break 30'

11:00 - 14:00

Chair: Lorenzo Bardone (SISSA)

11:00 Towards a theory of Generative AI: A model-based approach 45'

Speaker:	Francesco Cagnetta (SISSA)
Material:	Video

11:45 An Example from the Artists? A Philosophical Journey through Creative Experiments with AI 45'

Speaker:	Alice Barale (Università di Milano)
Material:	Video

12:30 Lunch 1h30'

14:00 - 16:30 Contributed talks

Chair: Sebastian Goldt (SISSA)

14:00 Reimagining Authorship and Ownership of AI-Generated Images under EU Law 20'

Speaker:	Giovanni Chieco (Università di Trieste)
Material:	Abstract Video

14:20 Color Distributions in Nature: A Data-Driven Exploration of Color Harmony and Aesthetic Preferences 20'

Speaker:	Ortensia Forni (EconophysiX - Ecole Polytechnique)
Material:	Abstract Video

14:40 Creativity as a Human Virtue 20'

Speaker:	Patrick Hayes (University of Glasgow)
Material:	Abstract Video

15:00 Panel discussion 1h0'

Moderator: Lorenzo Carta (Università di Trieste)

16:00 Coffee Break 30'

16:30 - 17:30 Concert
- 16:30 COMPOSIZIONE AI & DISKLAVIER 1h0'
  
  Speaker: Ardavan Vossoughi (Università di Padova)

Friday, 4 July 2025

09:00 - 11:00

Chair: Elisa Gremmo (Unipd)

09:00 Five Questions on Intelligence and Creativity 45'

Speaker:	Alessandro De Cesaris (Università di Torino)
Material:	Video

09:45 The creative process in mathematics, mathematicians, and large language models 45'

Speaker:	Antonio Lerario (SISSA)
Material:	Video

10:30 Group photo & Coffee break 30'

11:00 - 14:00 Contributed talks

Chair: Fabiola Ricci (SISSA)

11:00 Generative AI: collaborator or tool? The experience of interaction in the process of artistic creation 20'

Speaker:	Puscasu Iulia-Dana (Babes-Bolyai University)
Material:	Abstract Video

11:20 Synthesizing Naturalistic Visual Textures Using Deep Neural Samplers 20'

Speaker:	Ludovica De Paolis (SISSA)
Material:	Abstract Video

11:40 AI and the problem of creative spontaneity 20'

Speaker:	Dorde Lazarevic (University of Niš)
Material:	Abstract Video

12:00 Process Versus Output: Various Illusions of Originality in Humans and AI 20'

Speaker:	Baldwin Daniel (University of Liverpool)
Material:	Abstract Video

12:30 Lunch 1h30'

14:00 - 18:30 Contributed talks

Chair: Lorenzo Bardone (SISSA)

14:00 Can Machines Philosophize? 20'

Speaker:	Michele Pizzochero (University of Bath)
Material:	Abstract

14:20 Metaphorical Masks: Capturing and Concealing LLM Creativity in Art 20'

Speaker:	Mácha Jakub (Masaryk University)
Material:	Abstract Video

14:40 Can GenAI be creative? AI-Human Co-Creativity:Enriching the Concept of Creativity in Light of Emerging GenAITowards Collective Co-Creativity of AI and Humans 20'

Remote talk

Speaker:	Kunicki Martha (Princeton University)
Material:	Abstract

15:00 Panel discussion 1h0'

Moderator: Pierluigi Masai (Università di Trieste)

16:00 Coffee break 30'
16:30 Poster session 2h0'
```
for participants 'in person'
```
Material: LINK to BOOKLET

Saturday, 5 July 2025

09:00 - 11:45

Chair: Sebastian Goldt (SISSA)

09:00 Could AI be a genius? - Kant's theory of art, artistic creativity and AI 45'

Speaker:	Larissa Berger (Uni Siegen)
Material:	Abstract

09:45 An Analytic Theory of Creativity in Convolutional Diffusion Models 45'

Speaker: Mason Kamb (Stanford)
10:30 Coffee break 30'

11:00 Beauty at the Edge of Complexity: A Quantitative Exploration of Aesthetic Preference 45'

Speaker:	Alexandre Darmon (Art in Research, France)
Material:	Abstract Video

11:45 - 12:30 Contributed talks

Chair: Lorenzo Bardone (SISSA)

11:45 From the Human/Machine Dichotomy to Distributed Creativity - The Case of AI Integration in Filmmaking 20'

Speaker:	Pierluigi Masai (Università di Trieste)
Material:	Abstract Slides

12:05 The Myth of the Creativity Dial: Temperature in LLMs 20'

Speaker:	Federico Milanesio (Università di Torino)
Material:	Abstract Slides

12:25 Conclusion 5'

Speaker: Sebastian Goldt (SISSA), Elisa Gremmo (Università di Padova)

Monday, 7 July 2025

08:30 - 09:00 Registration & administrative formalities

09:00 - 11:00

Chair: Sebastian Goldt (SISSA)

09:00 Opening remarks 5'

Speaker:	Jean Barbier (ICTP)
Material:	Video

09:05 Learning from Gaussian Data 45'

Single and Multi-Index Models

Speaker:	Alex Damian (Princeton)
Material:	Video

09:50 Propogation of Chaos in 2-Layer NNs 45'

Speaker:	Margalit Glasgow (Stanford)
Material:	Video

10:35 Coffee break 25'

11:00 - 14:00

Chair: Marco Mondelli (ISTA)

11:00 Capacity threshold for the Ising perceptron 45'

Speaker:	Brice Huang (MIT)
Material:	Abstract Video

11:45 The dynamics of bias in SGD training 45'

Speaker:	Stefano Sarao Manelli (Chalmers)
Material:	Video

12:30 Lunch 1h30'

14:00 - 16:15

Chair: Jean Barbier (ICTP)

14:00 Statistical mechanics of extensive-width shallow neural networks near interpolation 45'

Speaker:	Mauro Pastore (ICTP)
Material:	Video

14:45 Revisiting empirical risk minimization: new risk characterizations and suboptimality results 45'

Speaker:	Reese Pathak (UC Berkeley)
Material:	Video

15:30 Improving training via progressive distillation 45'

Remote talk

Speaker:	Bingbin Liu (CMU)
Material:	Video

Tuesday, 8 July 2025

09:00 - 11:00

Chair: Jean Barbier (ICTP)

09:00 High-dimensional dynamics of SGD for GLMs and Gaussian mixture models 45'

Speaker:	Elizabeth Collins-Woodfin (University of Oregon)
Material:	Video

09:45 The Lovasz number of random circulant graphs 45'

Speaker:	Daniil Dmitriev (ETH Zurich)
Material:	Video

10:30 Group photo & Coffee break 30'

11:00 - 13:30 Contributed talks

Chair: Sebastian Goldt (SISSA)

11:00 Memorization and Generalization in Generative Diffusion 15'

Speaker:	Raphaël Urfin (École Normale Supérieure PSL)
Material:	Abstract Video

11:20 An Analytic Theory of Creativity in Convolutional Diffusion Models 15'

Speaker:	Mason Kamb (Stanford)
Material:	Abstract Video

11:40 The Rich and the Simple: On the Implicit Bias of Adam and SGD 15'

Remote talk

Speaker:	Bhavya Vasudeva (University of Southern California)
Material:	Abstract Video

12:00 Beyond Scaling Curves: Internal Dynamics of Neural Networks Through the NTK Lens 15'

Speaker:	Konstantin Nikolaou (University of Stuttgart)
Material:	Abstract Video

12:30 Lunch 1h0'

13:30 - 21:15

Chair: Marco Mondelli (ISTA)

13:30 Learning beyond the Gaussian approximation: Algorithmic Hardness and Neural Network Dynamics 45'

Speaker:	Lorenzo Bardone (SISSA)
Material:	Video

14:15 A statistical approach to implicit regularization 45'

Remote talk

Speaker:	Jingfeng Wu (Berkeley)
Material:	Video

15:00 Does the [training + test] loss function matter in the ultra-high-dimentional regime? 45'

Speaker:	Vidya Muthukumar (Georgea Tech)
Material:	Video

15:45 Poster session 1h0'
```
for participants 'in person'
```
Material: LINK to BOOKLET
16:45 Hike + Aperitivo 3h0'

Wednesday, 9 July 2025

09:00 - 11:00

Chair: Marco Mondelli (ISTA)

09:00 Model systems for internal structure in Transformers 45'

Speaker:	Freya Behrens (EPFL)
Material:	Video

09:45 Internal validity of representational comparisons 45'

Speaker:	Erin Grant (Gatsby)
Material:	Video

10:30 Coffee break 30'

11:00 - 14:00 Contributed talks

Chair: Jean Barbier (ICTP)

11:00 Local minima of the empirical risk in high-dimensions: general theorems and convex examples 15'

Speaker:	Basil Saeed (Stanford)
Material:	Abstract Video

11:15 Spectral Estimators for Multi-Index Models: Precise Asymptotics and Optimal Weak Recovery 15'

Speaker:	Filip Kovacevic (ISTA)
Material:	Abstract Video

11:30 Training Dynamics of In-Context Learning in Linear Attention 15'

Remote talk

Speaker:	Yedi Zhang (University College London)
Material:	Abstract Video

11:45 What's in a caption? Finding semantics with the information imbalance 15'

Speaker:	Andrea Mascaretti (SISSA)
Material:	Abstract Video

12:00 Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers 15'

Speaker:	Lorenzo Tiberi (Harvard University)
Material:	Abstract Video

12:30 Lunch 1h30'

14:00 - 16:15

Chair: Sebastian Goldt (SISSA)

14:00 Privacy for Free in the Overparameterized Regime 45'

Speaker:	Simone Bombari (ISTA)
Material:	Video

14:45 Understanding Test-time Inference in LLMs 45'

Remote talk

Speaker:	Hadi Daneshmand (MIT)
Material:	Abstract Video

15:30 Efficient Learning with Smoothed Data 45'

Speaker:	Abhishek Shetty (MIT)
Material:	Video

Thursday, 10 July 2025

10:30 - 14:30 The ICTP-IBM Brahmagupta AI Prize for Young Scientists Ceremony

The Ceremony will be live-streamed, CLICK HERE to see the livestreaming

Click here to see the ICTP-IBM Brahmagupta AI Prize for Young Scientists web page

Chair of the Ceremony: Sandro Scandolo (Senior Coordinator of the ICTP Research Division, member of the CMSP Section)

10:30 Opening remarks 5'

Speaker: Atish Dabholkar (Director of ICTP)
10:35 Brahmagupta's legacy 15'
```
Opening Talk
```
Speaker: Claudio Arezzo (Head of Mathematics Section, ICTP)

10:50 Awarding of the ICTP-IBM Brahmagupta AI Prize for Young Scientists 10'

Speaker:

Atish Dabholkar (Director ICTP), Alessandro Curioni (IBM Research Zurich), Simon Olsson (Chalmers University of Technology)

11:00 Generative Models of Molecular Dynamics 45'

Talk by the Winner

Speaker:	Simon Olsson (Chalmers University of Technology)
Material:	Abstract

11:45 The Future of Scientific Discovery 45'
```
Keynote Talk
```
Speaker: Alessandro Curioni (IBM Research Zurich)
12:30 Group photo & Welcome Reception Lunch 1h30'

15:00 - 20:00 AI 4 Science and Society" Meetup - How AI shapes, and is shaped by science and society

3:00 pm – Check-in & Networking
3:30 pm – Introductions & Welcome by Agata Ferretti (The AI Alliance at IBM) and Serafina di Gioia (ICTP/SISSA)
3:45 pm - Talks:
    Francisco Villaescusa-Navarro (Flatiron/ CCA) - AI agents for scientific discovery
    Cristiano De Nobili (Critiqality) - Emerging technologies for Earth Observation (and beyond)
    Claudio Zeni - AI for science at Microsoft Research: from electronic structure to proteins (Microsoft Research, Cambridge)
    Jean Barbier (ICTP)
    Claudia Merger (SISSA-TSDS)
    Alberto Cazzaniga (AREA)
5:30 pm - Panel "AI and society"
    John Shawe-Taylor (UCL)
    Teresa Scantamburlo (UNITS)
    Kate Devlin (KCL)
    moderator: Prof. Roberto Trotta (SISSA)
6:30 pm – Open Networking w/ food and drinks
8:00 pm – Close

Location:

Leonardo building