Micro-Seminar: From Arithmetic to Large Language Models PART TWO

Micro-Seminars have two parts. Attendance to both parts is required. Registering for the PART ONE session will automatically enroll you in the PART TWO session on Friday.

Part 1: Thursday, August 22, 2024 from 3:00 – 4:30 pm (PST)
Part 2: Friday, August 23, 2024 from 10:00 – 11:30 am (PST)

Large Language Models seem to be everywhere today, responsible for innovations in the biological and pharmaceutical sciences, industry, communications, computer science, education, and many aspects of culture and society. And it all started with systems for translating between languages. But how do these systems actually work? Many now believe that users of such a transformative technology ought to have a basic understanding of its inner workings.

Students will be introduced to a simplified, but accurate, overview of the computations that run inside LLMs, requiring only a rudimentary understanding of arithmetic.

Lead By: Professor Khalil Iskarous

Professor Iskarous is interested in human cognition, especially linguistic cognition, and uses LLMs as models of cognition. His National Science Foundation research probes what automatic speech recognition LLMs know about human speech, and applies the results to advance progress in documenting and revitalizing endangered languages and cultures, as well as probing speech disorders such as Parkinsonian rigidity. Through the Dornsife summer program "Problems without Passports", he has taken USC undergraduates to Taiwan and the Alps to help communities document and revitalize their languages.

Academic Honors and Fellowships | Website | View More Events

Contact the organizers