Machine learning for Sciences and Humanities

Información del proyecto

SMASH

Identificador del acuerdo de subvención: 101081355

DOI

10.3030/101081355

Fecha de la firma de la CE 10 Octubre 2022

Fecha de inicio 1 Julio 2023

Fecha de finalización 30 Junio 2028

Financiado con arreglo a

• Marie Skłodowska-Curie Actions (MSCA)

Coste total

Sin datos

Aportación de la UE

€ 4 776 000,00

Coordinado por

UNIVERZA V NOVI GORICI
Slovenia

Periodic Reporting for period 1 - SMASH (Machine learning for Sciences and Humanities)

Período documentado: 2023-07-01 hasta 2025-06-30

Resumen del contexto y de los objetivos generales del proyecto

SMASH is a collaborative and multidisciplinary programme co-funded by the EU and Slovenian government. It is centred on developing cutting-edge Machine Learning (ML)/AI applications for science and humanities. SMASH aims to attract 50 outstanding postdoctoral researchers from across the world, to create a multi-disciplinary environment in which they mutually benefit from knowledge exchange, both between different fields and between academia and industry. It connects scholars from five top-level institutions in Slovenia with current 38 associated partners among Slovenian businesses and academic institutions globally.

ML/AI is changing how science is being performed today. A number of modern scientific
experiments are currently providing large amounts of data and ML/AI is becoming a necessary tool in different research fields. Through active exchange of knowledge, commitment to strict Open access policies and high standards of Ethics evaluation, SMASH aims to set example of good practices in applications ML/AI to science in Slovenia.

Main Objectives of SMASH
• Enhancing research excellence and career prospects of fellows
SMASH provides outstanding working conditions, personalized mentoring from leading European experts across diverse fields, and a comprehensive program of soft-skills training. This combination equips fellows with strong transferable skills tailored to their individual career paths, supporting their progression toward fulfilling careers in a rapidly evolving world
• Fostering idea exchange and building a Slovenian ML/AI research community
By leveraging the common language of machine learning and AI, SMASH enables efficient knowledge exchange. The program connects fellows, supervisors, and scientists working across five domains: ML/AI development from a data science perspective, climate change impact predictions, next-generation precision medicine, fundamental questions about the Universe, and the study of language and communication. While Slovenian institutions are already strong in these fields individually, SMASH aims to elevate collaboration both within Slovenia and internationally. Regular monthly seminars and annual network meetings cultivate a close-knit, creative community. Events are open to the public and hosted beyond Ljubljana, ensuring broader dissemination, knowledge transfer, and engagement with local communities throughout Slovenia.
• Setting an example of the region for high standards of meritocratic recruitment
SMASH operates a fully merit-based recruitment process, employing anonymized reviews to minimize unconscious bias. This approach is expected to improve the participation of underrepresented groups (across gender, race, ethnicity, and more), thereby strengthening diversity among future AI/ML researchers and developers
• Placing ethics at the forefront
A comprehensive ethics framework underpins the network, supported by ongoing training, assessments, workshops, Q&A sessions, and forums. Beyond technical excellence, fellows receive in-depth training on the ethical challenges and opportunities posed by AI, theories of democracy and administrative power, and issues of inequality and justice. They learn to critically assess how AI may create disparities in resources, opportunities, or power, and how it may reinforce historical injustices. This emphasis ensures that fellows develop a deep understanding of global and corporate social responsibility, equipping them to engage thoughtfully with the societal impact of AI.

Trabajo realizado desde el comienzo del proyecto hasta el final del período abarcado por el informe y los principales resultados hasta la fecha

Recruitment and Fellow Integration
During this reporting period, SMASH successfully achieved all recruitment objectives: all four calls were completed, and all 50 fellows have been recruited. Of these, 25 began their employment within this period.
Upon arrival, each fellow, as foreseen in the project design, prepared a Personal Career Development Plan (PCDP) detailing their project objectives, secondments, dissemination activities, and training pathways, tailored to their career aspirations. All PCDPs were reviewed and approved by the SMASH Supervisory Board. In addition, fellows submitted Proportionate Ethics Forms, as part of their training in ethical awareness, which were reviewed by our dedicated EEDI Advisor. Ethics Q&A sessions and individual consultations were held to support fellows and supervisors.
Currently, all SMASH projects are progressing according to these plans, as evidenced by 20 approved six-monthly PCDP progress reports.
________________________________________
Outputs and Achievements
• Publications & Open Science: 28 peer-reviewed papers or conference proceedings published; 13 open-source codes or datasets released
• Conferences & Networking: Fellows participated in ~60 international conferences and meetings.
• Secondments: 11 international secondments (in Canada, US, and EU), 5 within Slovenian academic institutions, and 2 in a Slovenian company as cross-sectorial exposure. Average duration: 2 months; longest: 6 months.
________________________________________
Training and Community Events
The SMASH network organized numerous training and outreach activities, all open to external participants, Some examples include:
• A SMASH international workshop with ~80 researchers from 11 countries, featuring a public round table on Impact of AI on Science and Society and cross-sectoral “Meet the SMEs” training.
• 14 SMASH seminars with speakers coming from US and EU organized by our fellows
• Soft-skills training, including workshops on open access (UNG Library), FAIR principles (jointly with CERN), workshop on Video making and social media posting for scientists and sessions on Communicating Ideas Clearly and Leading Small Teams Effectively.
________________________________________
Scientific Highlights
• High-impact publications: including papers in Environment International (Impact Factor 13.35)
• Major conference contributions:
o Our fellows delivered plenary talks at a number of prestigious conferences, for example EUCAIFCon 2025 (a yearly meeting of European Coalition of AI in Fundamental physics) and at the 4th EuCAPT Symposium at CERN.
o Two fellows presented at the European Geosciences Union General Assembly, one of the world’s largest geoscience conferences, attended by >15,000 scientists.
• Recognition: one of our fellows was named an Outstanding Reviewer for CVPR 2025, an award given to the top 6% of reviewers. CVPR is considered the premier global computer vision conference and the second most cited scientific publication venue worldwide (after Nature)

Avances que van más allá del estado de la técnica e impacto potencial esperado (incluida la repercusión socioeconómica y las implicaciones sociales más amplias del proyecto hasta la fecha)

In relation to our objective of enhancing the research excellence and career prospects of our fellows, we are pleased to report that four fellows nearing the completion of their SMASH contracts have already secured future positions, including one permanent appointment.
With respect to building a close-knit research community, we are encouraged to see that several fellows have initiated collaborations between them, extending beyond their original project plans. These collaborations emerged through interactions at SMASH seminars and during secondments. Furthermore, the SMASH host institutions, leveraging the strengthened ties within the network, jointly applied for a major national grant to establish a National Competence Center on AI—demonstrating the added value of collaborations fostered through SMASH. We are also in the midst of organizing the yearly SMASH conference jointly with other AI initiatives in Slovenia, resulting in a big AI4Science meeting that will take place next months in Ljubljana.
Concerning our goal of serving as a regional example of merit-based recruitment, our applicant diversity survey indicates that we successfully attracted a broad pool of candidates. Notably, the share of female fellows recruited exceeded the proportion of female applicants, suggesting that our anonymized recruitment process effectively reduced unconscious bias.
We also actively engaged in communication with the public, by authoring a number of popular articles in high-profile local magazines, by organizing and participating in panel discussions on the impact of AI on science and by participating in community events. Some of our fellows also made applications for public where results of their research can be visualized and tracked.

First SMASHing Workshop