Deep Learning Generation: Novel Methods for Ensuring AI Can Be Trusted

Date: June 22 2024 Saturday Noon - 14:00 EDT

Title: Deep Learning Generation: Novel Methods for Ensuring AI Can Be Trusted
Topic:
Artificial Intelligence, powered by deep learning architectures, has become ubiquitous in our daily lives, granting computers the ability to see, create art, and interpret videos and photographs. Despite the remarkable capabilities of deep learning, they are not infallible and can make errors when confronted with conditions different from their training data.

Generalization, or the ability of a model to accurately interpret or respond to new, previously unseen data, based on the knowledge it has acquired during its training phase, is a dynamic area of research. This talk will introduce the overall problem of model generalization, give an overview past attempts to solve it and recent trends in this field, elucidate relevant techniques, and introduce a novel open source library that can be used to predict generalization performance.

Speaker:
Elliott Miller boasts a decade of experience as an AI practitioner, with a specialized focus on productized machine learning. His expertise shines in the realms of geospatial computer vision and natural language processing. Constantly intrigued by emerging trends in data comprehension, Elliott remains deeply engaged in the field. Beyond his professional pursuits, he actively participates in various local DC meetups, relishing opportunities to exchange insights and knowledge with fellow enthusiasts.

Moderators: Dr. Pawel Gora, CEO of Quantum AI Foundation, Dr.Sebastian Zajac, member of QPoland