Faculty Recruiting Support CICS

Assistant In a Box: Leveraging The Power of Large Language Models

13 Oct
Thursday, 10/13/2022 12:00pm to 1:00pm
Computer Science Building, Room 150/151, Zoom
Machine Learning and Friends Lunch

Title: Assistant In a Box: Leveraging The Power of Large Language Models

Abstract: Large Language Models (LLMs) are larger and more accessible than ever with more capabilities being discovered on a daily basis. Applications that were previously infeasible are now becoming possible due to the development of and access to LLMs. Join me in this demo rich presentation on the paradigm shift involved with going from writing complicated code to engineered prompts.

We will be discussing how to save time through rapid prototyping of features by leveraging the zero and few shot capabilities of LLMs. We will be discussing the different ways to use LLMs for time consuming tasks that previously would have required manual input.

The talk will also go over practical aspects of working with LLMs such as prompt engineering and combining LLMs with other NLP techniques, Common pitfalls and limitations such as bias, hallucinations, and cost. We’ll also be talking about the responsible use of this technology.

Bio: Daniel is currently working at Microsoft’s Artificial Intelligence Development Accelerator Program in the New England Research and Development center otherwise known as the NERD center. His work revolves around leveraging AI to develop AI powered applications from zero to MVP across all kinds of verticals inside of Microsoft. In concrete terms his work touches on all the components of creating AI enabled applications from model development, serving and integration. Daniel has a background in AI through his previous work at AI startups as well as his time at the University of Rochester where he graduated with a BS in Computer Science.

To find out more information about this event or to obtain the Zoom link, please see the event announcements from MLFL on the college email lists or contact wenlongzhao [at] cs.umass.edu (subject: MLFL%20Zoom%20Link) (Wenlong Zhao).