# Problem solving as a translation task

- Institution: CERN
- Department: EP-IT Data Science Seminars
- Speaker: Francois Charton
- Year: 2024
- Language: English
- Subjects: Computer science
- Source: [CERN](https://videos.cern.ch/record/3002434)
- Licence: [CERN](http://copyright.web.cern.ch)
- Who can watch: Tweens profiles and older (sources give no ratings; placed by Invisible Network from the source's own words)

Neural architectures designed for machine translation can be used to solve problems of mathematics, by considering that solving amounts to translating the problem, a sentence in some mathematical language, into its solution, another sentence in mathematical language. Presenting examples from symbolic and numerical mathematics, and theoretical physics, I show how such techniques can be applied to develop AI for Science, and help understand the inner workings of language models.

Francois Charton is a research engineer in Meta AI, working on applications of language models to mathematics and science. He is a former student of Ecole Polytechnique, and ENSAE (Ecole Nationale de la Statistique et de l'Administration Economique). He has been in Meta since 2019, and had a career in industry prior to that.

Coffee will be served at 10:30. .

0:00:00 Slide 1

0:04:54 Slide 2

0:06:14 Slide 3

0:07:57 Slide 4

0:11:06 Slide 5

0:12:20 Slide 6

0:14:39 Slide 7

0:15:54 Slide 8

0:16:56 Slide 9

0:18:16 Slide 10

0:19:35 Slide 11

0:21:19 Slide 12

0:22:30 Slide 13

0:25:32 Slide 14

0:26:40 Slide 15

0:27:43 Slide 16

0:29:12 Slide 17

0:30:39 Slide 18

0:31:58 Slide 19

0:34:40 Slide 20

0:36:35 Slide 21

0:39:11 Slide 22

0:40:34 Slide 23

0:42:39 Slide 24

0:44:03 Slide 25

0:48:07 Slide 26

0:49:09 Slide 27

## Lectures

1. [Problem solving as a translation task](https://invisiblenetwork.ai/learn/cern/3002434/77f7112f75f0), 1 h 6 min (CERN's player)