Large Models for All and How to Adapt Them Efficiently
Alex Smola
CEO - Boson AI
Foundation models have revolutionized machine learning, yet using them is still a challenge.
In this talk I will give an overview over such models, and how to use them, both for mixed data in the context of AutoML and through fine-tuning and prompting, as is commonly done in Large Language Models. In particular, AutoGluon allows users to integrate tabular data with representations for different data types, including text and images.
I will discuss how this can be accomplished efficiently, taking advantage of the latest models available on public repositories. When it comes to pure language models, I will cover how to adapt them efficiently, e.g. by automatic prompt and example generation, thus alleviating the need for manual generation, and by parsimonious parameterization, thus reducing the computational cost significantly.
About the Speaker Alex Smola
Alex was VP, Distinguished Scientist with Amazon, a lecturer at Stanford for Machine Learning, and now started his owna company.
He co-authored ‘Dive into Deep Learning,’ a widely used book adopted by 400 universities internationally, and he is one of the contributors to the opensource AutoML tool AutoGluon.
