A3D3 Seminar: Miles Cranmer

Monday Nov 13, 2023, 12:00 PM → 1:00 PM US/Pacific

Title: Polymathic AI: Foundation Models for Science
 

 

Abstract:

In the last few years, natural language processing and computer vision have experienced a fundamental shift in the way these fields use machine learning. Rather than training neural networks from a randomly initialized set of parameters, researchers have often found superior performance can be achieved by fine-tuning a general pre-trained “foundation model” trained on vast amounts of diverse data – perhaps because this model comes with better “priors” than an untrained network. Polymathic AI[1] is a new research collaboration that aims to usher in the same shift in machine learning for scientific datasets. In this talk I will present the motivations behind the collaboration and describe the findings of our three new papers in this space, which examine: better numerical encodings for large language models[2], contrastive embeddings for multi-modal scientific data[3], and building machine learning models that learn from multiple types of physics[4].

1 https://polymathic-ai.org/
2 https://arxiv.org/abs/2310.02989
3 https://arxiv.org/abs/2310.03024
4 https://arxiv.org/abs/2310.02994

 

Bio:  Miles Cranmer is Assistant Professor in Data Intensive Science, DAMTP & Institute of Astronomy, University of Cambridge.  GitHub <https://github.com/MilesCranmer> | Website <https://astroautomata.com/> | Twitter <https://twitter.com/MilesCranmer> 

 

 

 

 

The A3D3 Seminar is a monthly lecture series that hosts scholars working across applied areas of artificial intelligence, such as hardware algorithm co-development, high energy physics, multi-messenger astrophysics,  and neuroscience. Our presenters come from all four domain fields and include occasional external speakers beyond the A3D3 science areas, governmental agencies and industry. The seminar will be recorded and published in YouTube. To receive future event updates, subscribe here.