Skip to main content
Talk

AI for Code and Science — Universidad Peruana Cayetano Heredia (UPCH)

Dr. Omar Costilla Reyes gave the webinar “AI for Code and Science” at Universidad Peruana Cayetano Heredia (UPCH) in Lima in May 2023. He shared the recording publicly in July.

Recording of Dr. Costilla Reyes’s AI for code and science talk for the Faculty of Sciences and Engineering at Universidad Peruana Cayetano Heredia. Watch on YouTube.
UPCH announcement poster for the May 2023 webinar, which Dr. Costilla Reyes later shared as a public recording. Credit: Image courtesy of Dr. Omar Costilla Reyes.
A workshop session at MIT, February 2023. Related photograph from AI for Code and Science Workshop. Credit: Image courtesy of Dr. Omar Costilla Reyes.
Figure from the paper showing how the neurosymbolic programming (NP) lifecycle feeds into and draws on the scientific process, from choosing a research question to reporting to the community, shared with the paper announcement. Related photograph from Neurosymbolic programming for science. Credit: Image courtesy of Dr. Omar Costilla Reyes.

A webinar for Peruvian engineering students

The Ingeniería Informática (computer engineering) program in the Facultad de Ciencias e Ingeniería at Universidad Peruana Cayetano Heredia (UPCH) in Lima invited Dr. Omar Costilla Reyes to give the webinar “AI for Code and Science” in its webinar series. The announcement poster, in Spanish, presents the session as an online webinar in May 2023, with Dr. Costilla Reyes introduced as a research scientist at MIT CSAIL. Gissella Bejarano hosted him, and in July he shared the recording publicly, thanking her and the university and describing it as a pleasure to help inspire the next generation of scientists.

The subject: AI for code and science

Two research areas meet in the title. AI for code applies machine-learning models trained on very large amounts of source code, such as the transformer-based models CodeBERT, CodeT5, and Codex, to tasks like completing, explaining, and repairing programs. AI for science uses machine learning to find patterns, models, and equations in scientific data. Neural networks learn well from data but are hard to read from the inside; symbolic methods such as program synthesis, which automatically generates a program that meets a specification, produce models a researcher can inspect and check. Blending the two is an active research direction that links machine learning, programming languages, and software engineering.

The same subject was the focus of the MIT AI for Code and Science Workshop that Dr. Costilla Reyes co-organized in February 2023, a two-day hands-on tutorial session on combining neural-network methods and program synthesis for science, whose recordings and code were released through the Neurosymbolic Programming for Science project. The webinar also followed his 2019 visit to Lima, where he spoke to the first student cohort of Aporta’s Institute for Advanced Analytics and Data Science. Both are described on their own pages on this site.

Topics