AI for Ukraine / AI HOUSE

AI In Biomedical Imaging: How Computer Scientists Can Contribute To Medicine I Recording

Experimental biologists gather vast data on organisms . Key among these are images that need to be analyzed and interpreted with the help of advanced techniques. That's the focus of Dmytro's biomedical computer vision lab. In this talk, Dmytro will cover recent projects and illustrate how computer scientists can enhance healthcare.

Register here

Dmytro Fishman

Assistant Professor of Artificial Intelligence at the University of Tartu; Lecturer of ML in Ukrainian Catholic University and Chief Science Officer at Better Medicine

Arize LLM Observability, How To and Best Practices | Recording

Dive into the world of Arize LLM Observability in this technical seminar. Gain insight into LLM evaluations, RAG workflows, prompt engineering debugging, root cause analysis, and debugging using embeddings flows, traces, and spans. Learn to navigate troubleshooting flows effectively. Join us to elevate your engineering skills in LLM observability with the leading LLM Observability
Platform, Arize.

Register here

Dat Daryl Ngo

ML Solutions Architect at Arize AI

A Hands-On Introduction to Automated Machine Learning | Recording

It's often necessary to optimize a machine learning approach to get good performance, in particular through hyperparameter tuning. Join the webinar to dive into some of the background of hyperparameter tuning and automated machine learning before a hands-on session that will show you how to use AutoML tools for your ML application.

Register here

Lars Kotthoff

Associate Professor, University of Wyoming

Lessons Learned Productionising LLMs for Stripe Support | Recording

In this talk, Sophie discusses lessons learned productionising Stripe’s first application of Large Language Modeling – providing answers to user questions for Stripe Support.

Register here

Sophie Daly

Staff Data Scientist at Stripe

The Ethical Implications of AI: How to Strike a Balance between Progress and Humanity | Recording

This talk could explore the ethical considerations that come with the development and application of AI, such as privacy, bias, and job displacement. It could also discuss possible solutions for balancing technological progress with human values.

Register here

Haseeb Khan

Senior AI ML Engineer at Google

Balancing Ethics and Creativity: Leveraging LLMs for Responsible Content Generation I Recording

How can you leverage LLMs to avoid any potential pitfalls related to reputation, responsibility, and regulations? To address these challenges, organizations must establish rigorous control measures guided by industry best practices. In this workshop, Viktoriia will share some of her experience in developing an engine for tailoring reliable and trustworthy marketing content that may be useful for you as well.

Register here

Viktoriia Oliinyk

Data Scientist in University of Oxford

Representation Learning on Relational Data to Automate Data Preparation I Recording

Preparing data is a huge part of any statistical learning. In this lecture, Gael will present progress on learning representations with data tables, overcoming the lack of simple regularities. He will show how these representations decrease the need for data preparation: matching entities, aggregating the data across tables.

Register here

Gael Varoquaux

Research director at INRIA

Deep Learning Decade and GPT-4 I Recording

How did ChatGPT happen and what preceded it? Lukasz stood at the origin of this revolution as a part of the OpenAI team, so he has some stories to tell. This lecture will help you understand the evolution of Transformers and the process of creation of GPT4. As a result, you’ll get a better idea of what to expect in the years to come.

Register here

Lukasz Kaiser

Deep Learning Researcher at OpenAI

Bridging the gap between current deep learning and human higher-level cognitive abilities

This talk raises many interesting research questions ranging from Bayesian inference and identifiability to causal discovery, representation learning and out-of-distribution generalization and adaptation, which will be discussed in the presentation

Yoshua Bengio

Professor at Université de Montréal, Founder and Scientific Director of Mila

Data-driven Reinforcement Learning with Transformers

In this talk, Misha Laskin will give a tutorial on these methods that he will broadly refer to as Reinforcement Learning Transformers (RLTs). Misha will show how RL agents can be pre-trained to learn in-context like LLMs. He will then present new work showing that transformers can be pre-trained to reinforcement learn in-context.

Misha Laskin

Senior Research Scientist at DeepMind

Evaluating machine learning models and their diagnostic value

Gaël will first discuss choosing metric informative for the application, stressing the importance of the class prevalence in classification settings. Then he will then discuss procedures to estimate the generalization performance, drawing a distinction between evaluating a learning procedure or a prediction rule, and how to give confidence intervals to the performance estimates.

Gael Varoquaux

Сo-founder of scikit-learn, Director at INRIA

Model Compression for Deep Learning

In this presentation we will discuss why and when to compress ML models, survey major model compression techniques and best practices, and review state-of-the-art approaches to model compression. We will focus on pruning and quantization, but also cover other techniques, like knowledge distillation, deep mutual learning, and architecture search.

Sergiy Matusevych’s lecture is available in closed access, which can be obtained by registering at this link.

Sergiy Matusevych

Principal Data and Applied Scientist at Microsoft

Modeling Personal Experiences Shared in Online Communities

In this talk, Maria will share work that seeks to reliably represent individual experiences within their social contexts and model interpretive dimensions that illuminate both patterns and outliers, while addressing social and humanistic questions. She will share two case studies that highlight both the opportunities and the risks in reusing NLP models for context-specific research questions.

Maria Antoniak

PhD Candidate at Cornell University

Multimodal Grounded Learning with Vision and Language

Recently we increasingly see language being used to enhance visual models by enabling zero-shot capabilities, improving generalization, mitigating bias, etc. Anna is deeply interested in building models that can ingest language advice to improve their behavior.

In her talk Anna will cover work that tries to achieve the aforementioned capabilities and discuss challenges as well as exciting opportunities that lie ahead.

Anna Rohrbach

Research Scientist at UC Berkeley

One and Done - Automatic Machine Learning with AutoGluon

In the past Machine Learners had to choose between simplicity, speed and accuracy when using AutoML tools. In this talk Alex Smola will show how AutoGluon provides all three of them, for a wide range of applications. This includes tabular data, text, images, and time series among the many options.

Alex Smola

Principal Data and Applied Scientist at Microsoft

Prompting, Metadatasets, and Zero-Shot NLP

This talk focuses on T0, a large-scale language model trained on multi-task prompted data (Sanh et al 2022). Despite being an order of magnitude smaller than GPT-3 class models, T0 exhibits similar zero-shot accuracy on unseen task categories. In addition to the modeling elements, this talk highlights the community processes of collecting data, dataset, and prompts for models of this scale.

Alexander Rush’s lecture is available in closed access, which can be obtained by registering at this link.

Alexander Rush

Assoc. Professor at Cornell, Researcher at Hugging Face

Search in imperfect information games

The combination of decision-time search and value functions has been present in the remarkable milestones where computers bested their human counterparts in long-standing challenging games — DeepBlue for Chess and AlphaGo for Go. Until recently, this powerful framework of search aided with (learned) value functions has been limited to perfect information games.

We will talk about why search matters, and about generalizing search for imperfect information games.

Martin Schmid

CEO and Co-Founder of EquiLibre Technologies

Towards Causal Representation Learning

Francesco will first review fundamental concepts of causal inference and present new approaches for causal discovery using machine learning.
Then, he will broadly discuss how causality can contribute to modern machine learning research. As well, Francesco will introduce causal representation learning as an open problem for both communities: the discovery of high-level causal variables from low-level observations.

Francesco’s lecture is available in closed access, which can be obtained by registering at this link.

Francesco Locatello

Senior Applied Scientist at Amazon Web Services

Towards Practical Reinforcement Learning: Offline Data and Low-Adaptive Exploration"

In this talk, Yu-Xiang Wang will first share some recent theoretical advances on the offline RL algorithms. Then he will talk about the limitation of offline RL over its online counterpart, and describe a new model called RL with low-switching cost that could get the best of both world.

Yu-Xiang Wang

Assistant Professor of CS at UCSB, Director of Scalable Statistical ML Lab

Unveiling Transformers with LEGO

In this work we propose a synthetic task, called LEGO (Learning Equality and Group Operations), to probe the inner workings of transformers. We obtain some insights on multi-head attention, the effect of pretraining, as well as overfitting issues.

Joint work by the ML Foundations group at Microsoft Research with co-authors Yi Zhang, Arturs Backurs, Ronen Eldan, Suriya Gunasekar, and Tal Wagner.

Sebastien Bubeck

Sr Principal Research Manager at Microsoft

AI for Ukraine

Speakers

Watch the lectures recordings

Watch the lectures recordings

SEASON 2

SEASON 1 | RECORDINGS

Report

Questions & answers