LIPS

21 Feb 2024, 09:00 → 23 Feb 2024, 18:00 Europe/Berlin

Auditorium (DESY)

 

---

Registration deadline for in-person attendance: 09.02.2024

Remote (zoom-only) registration will be possible until the end of the event. 

Abstract submission deadline: 04.02.2024

---

Recent Large Language Models like ChatGPT show impressive capabilities, e.g. in the automated generation of text and computer code. These new techniques will have long-term consequences, including for scientific research in fundamental physics. 

 

This first symposium aims to chart a path forward and bring together different applications and developments of large language models in physics research. 

 

Planned topics include:

• Direct LLM research use in fundamental physics
• Automated text and code generation 
• LLM integration in traditional workflows
• Implicit physics understanding
• Symbolic computation
• Prompt Engineering
• Alignment, Ethics, and Reliability

 

In-person attendance is encouraged. However, there will also be an option to connect remotely (please specify on the registration form, how you will attend, if you have not done so yet). No fee will be charged for remote participation. Zoom connection details will be shared with the registered participants before the meeting. 

 

Invited talks

•  Semantic taxonomies for radio astronomy - Micah Bowles (Manchester)
• AI and symbolic math - Francois Charton (Facebook AI Research)
• Language models for quantum simulation - Juan Carrasquilla (ETHZ)
• LLMs usage ar Anthropic AI - Marat Freytsis (Anthropic AI)
• cosmosage: a 7B LLM Fine-Tuned on Cosmology Papers and Textbooks - Tijmen de Haan (KEK)
• Prospects of LLMs for Fundamental Physics - Mariel Pettee / Siavash Golkar (Berkeley / Polymath AI)
• AstroLLaMA-Chat - Rui Pan (Hong Kong University of Science and Technology)
  
  and others 

 

 

Local Organisers 

• Marcus Brueggen (Uni Hamburg, mbrueggen@hs.uni-hamburg.de)
• Sarah Heim (DESY, sarah.heim@desy.de)
• Gregor Kasieczka (Uni Hamburg, gregor.kasieczka@uni-hamburg.de)

 

International Advisory Committee:

• Sven Krippendorf (LMU)
• Gilles Louppe (Liege)
• Anna Scaife (Manchester)
• David Shih (Rutgers)
• Daniel Whiteson (UC Irvine)

 

 

 

Group picture

Wednesday, 21 February

Knowledge extraction, LLMs & general physics

1 Arrival and registration

2 Welcome

Speakers: Gregor Kasieczka (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Marcus Marcus Brüggen (UHH), Sarah Heim (ATLAS (ATLAS Dark Matter with Higgs))

LIPS_intro.pdf

3 Extracting Measurements from (legacy) publications (20'+10')

Speaker: Peter Steinbach (HZDR)

20240221-lips24-PeterSteinbach.pdf

link to slides

4 Exploring the Strong Coupling and Dark Matter through Natural Language Processing (30'+10')

Speakers: Annika Eichler (MSK (Strahlkontrollen)), Antonin Sulc (MCS (Control System))

Presentation.pdf

15:20 Conference photo - either in auditorium or on the front steps

15:30 Break

5 ChATLAS: An AI Assistant for the ATLAS Collaboration (20'+10')

Speaker: Cary Randazzo (Louisiana Tech)

Chatlas Overview.pptx

6 Exploring LLM performance on Physics 101 coursework in different languages (20'+10')

Speaker: Marcel Völschow (Hamburg University of Applied Sciences)

LIPS2024.pptx

LIPS_PHYS101_VOEL.pdf

7 Educational Outreach with AI-Assisted CERN Open Data Analysis (20'+10') [zoom]

Speakers: Paul Philipp Gadow (None), Philipp Gadow (Max-Planck-Institut für Physik)

2024-02-21-Lips-ATLASOpenData.key

2024-02-21-Lips-ATLASOpenData.pdf

Link to animations

17:30 Welcome reception

Thursday, 22 February

Astronomy and assistants

8 Language Models for Multimessenger Astronomy (20'+10')

Speaker: Dmitriy Kostiunin (Z_DV (Datenverarbeitung))

kostunin-lips-2024.pdf

9 Cosmosage: a 7B LLM Fine-Tuned on Cosmology Papers and Textbooks (25'+10') [zoom]

Speaker: Tijmen de Haan (KEK)

cosmosage

10 LLMs in astronomy (25' + 10') [zoom]

Speaker: Rui Pan (The Hong Kong University of Science and Technology)

10:40 Coffee break

11 Semantic taxonomies for radio astronomy (25' + 10')

Speaker: Micah Bowles (The University of Manchester)

20240222_LIPS_SemanticTaxonomies.pdf

20240222_LIPS_SemanticTaxonomies.pptx

12 Semantic association of astronomical images with natural language (20'+10')

Speaker: Siddharth Mishra-Sharma (MIT)

13 MetaInsight: An LLM-Powered Research Assistant (20'+10') [zoom]

Speaker: Mohamed El Ghafiani (LPMR, Faculty of Science, Mohammed First University, Oujda, Morocco)

demo-1.mkv

Meta-Insight.pdf

Combining Domain Knowledge and LLMs, efficient computation

14 Language models for quantum simulation (25'+10') [zoom]

Speaker: Juan Carrasquilla (ETH Zurich)

Language models for quantum simulation Hamburg (online for me), Feb. 22nd, 2024 .pdf

15 AI Honesty and Self-Supervision (25' + 10') [zoom]

Speaker: Marat Freytsis (Anthropic)

16 AI and symbolic math (25'+10')

Speaker: Francois Charton (Meta AI)

AIFromGeneratedData.pdf

16:00 Break

17 Prospects of LLMs for Fundamental Physics (25' + 10') [zoom]

Speakers: Mariel Pettee (Lawrence Berkeley National Laboratory), Siavash Golkar

SGolkar MPettee LIPS 2024.pdf

18 Efficient Matrix Multiplication Algorithms for Quantized Language Models (20'+10') [in person, but demo on zoom]

Speaker: Johannes Gäßler (Karlsruhe Institute of Technology)

demo.mkv

Efficient_Mat_Mul.odp

Efficient_Mat_Mul.pdf

19 Helmholtz Blablador: An Inference Server for Scientific Large Language Models (20'+10')

Speaker: Alexandre Strube (Heymholtz AI - Juelich Supercomputing Centre)

Single HTML file with all resources

Friday, 23 February

Accelerators and wrap up discussion

20 PACuna: Automated Fine-Tuning of Language Models for Particle Accelerators (20'+10')

Speaker: Antonin Sulc (MCS (Control System))

LIPS_LLM-2.pdf

21 Building an Intelligent Accelerator Operations Assistant using Advanced Prompt Engineering Techniques and a High Level Control System Toolkit (20'+10')

Speaker: Dr Frank Mayet (MPY1 (MPY Fachgruppe 1))

LIPS_2024_Talk_FrankMayet.pdf

22 Large Language Models for Particle Accelerator Tuning (20'+10')

Speaker: Jan Kaiser (DESY)

20240223 LIPS LLM Accelerator Tuning.pdf

23 AccGPT: A Vision for AI Assistance at CERN's Accelerator Control and Beyond (20'+10')

Speakers: Florian Rehm (CMS (CMS Fachgruppe Searches)), Florian Rehm (CERN)

AccGPT-LIPS_v6.pdf

11:00 Break

24 Large Language Models in Fundamental Physics: An Interdisciplinary Roadmap (25'+10')

Speaker: Sascha Caron (Radboud University and Nikhef)

HamburgLargeLanguageModels_caron.pdf

25 Discussion: Taking stock and the path ahead

26 Good-Bye