Large Language Model at AWS - Generative AI and BloomBergGPT sharing

Generative AI is a type of AI that can create new content and ideas, including conversations, stories, images, videos, and music. It is powered by large models that are pre-trained on vast amounts of data and commonly referred to as foundation models (FMs). With generative AI on AWS, you can reinvent your applications, create entirely new customer experiences, drive unprecedented levels of productivity, and transform your business. You can choose from a range of popular FMs, or use AWS services that have generative AI built in, all running on the most cost-effective cloud infrastructure for generative AI.

Speaker: Yanwei Cui, Machine Learning Specialist Solutions Architect, AWS

Measuring the Impact of Remote Work Across the United States

We develop a framework to measure remote work at the firm-level using novel data of the daily internet activity for over hundreds of thousands of firms in the United States from 2019 to 2021. For each firm, we measure the fraction of employee internet activity arising from residential, business, VPN and mobile networks. Validating this classification, we document a 30 percent increase in remote IP traffic in March 2020 at the onset of the crisis and a negative covariance of -0.756 between the share of remote IP traffic in a county and mobile phone data on workplace visits. Using cross-sectional variation in remote work decisions, we characterize the determinants and consequences of remote work.

Speaker: Dr. Alan Kwan, tenure-track faculty member, University of Hong Kong

Can a pretrained neural language model still benefit from linguistic symbol structure? Some upper and some lower bounds

 In this presentation I introduce one way in which deep-learning-based language models (LMs) and symbolic linguistics can potentially be reconciled. The contributions I talk about include: an almost stupidly simple vector encoding of labeled and unlabeled linguistic structure, which performs much faster than and equally as effective as established methods; a comparison of different linguistic representations on the task of next-word prediction; as well as an analysis of robustness against noise. I conclude that if we had human-like linguistic knowledge resources for large amounts of data, we could indeed achieve drastic improvements to LM perplexity, which are even robust to certain types of "well-behaved" errors. However, it remains unclear if automatic parsers can be good enough to only produce well-behaved errors and avoid bad ones, and if yes, if the effort is worth it. This is joint work with Emmanuele Chersoni, Nathan Schneider, and Lingpeng Kong.

Speaker: Jakob Prange, postdoctoral fellow, the Department of Chinese and Bilingual Studies, Hong Kong PolyU

Comparing and Predicting Eye-tracking Data of Cantonese and Mandarin

This introduces the first deeply-annotated joint Mandarin-Cantonese eye-tracking dataset, from which we achieve a unified eye-tracking prediction system for both language varieties. In addition to the commonly studied first fixation duration and the total fixation duration, this dataset also includes the second fixation duration which is thought to reflect later and structural processing. A basic comparison of the features and measurements in our dataset revealed variation between Mandarin and Cantonese on fixation patterns related to word class and word position. The test of feature usefulness suggested that traditional features are less powerful in predicting structural processing, to which the linear distance to root makes a leading contribution in Mandarin. In contrast, Cantonese eye-movement behavior relies more on word position and part of speech.

Speaker: Junlin Li, postgraduate student, the department of Chinese and Bilingual Studies, Hong Kong PolyU