Alex Makelov: Harvard, Cambridge, MIT and the unknown sides of chatGPT

preview_player
Показать описание
Alexander Makelov is one of the few people who studied at Harvard, Cambridge and MIT, where he did his PhD on a topic related to Large Language Models. Alex will tell us about some weird and lesser-known sides of chatGPT.

#chatgpt #llm #harvard #cambridge #massachusettsinstituteoftechnology
---
00:11 Introduction of Alex
01:20 The transition from Math and Physics to CS
03:40 The choice of Machine Learning for the PhD
05:18 Difference between Math people and Physics people
08:07 Jokes about Math from Physics people
08:59 Math/STEM olympiads: is it nature or nurture and what is important?
15:44 Would you make your kids competitors in math/physics?
16:48 The criteria for Harvard
18:11 Why did you choose Harvard?
19:27 General advice to high school competitors in math/physics
21:37 The Harvard years
22:57 The undergraduate Harvard thesis on Expanders grapsh (that got Alex a prize)!
24:01 Expander graphs
28:33 The Cambridge years
30:49 the Harvard campus or the Cambridge campus?
31:11 the Math Tripos exam
33:15 the MIT years
34:14 Alex's current work on LLMs interpretability
37:01 ChatGPT - how does it work?
38:53 How is ChatGPT so good for essays?
42:01 Interpretability of LLMs
43:16 Lesser-Known things about ChatGPT
49:19 AGI and ChatGPT
51:18 Is it a problem we don't understand how ChatGPT work?
52:45 ChatGPT and AlphaZero
54:33 Books about AI/ML
56:10 Saying Goodbye
---

Links to stuff mentioned by Alex in the video: