If ChatGPT were playing Jeopardy grand finale in November 2022, it would destroy the opponents.
58 correct answers out of 61.
Not only that, there are some emerging behaviors while it plays. 🧵⤵️
Before we jump in, here is a quote from the first human battle against IBM Watson: "People don't realize how tough it is to write a program that can read in a natural language and understand all the double meanings, the puns, the red herrings, unpack the meaning of the clue."
Despite these challenges, chatGPT understands natural language almost perfectly. Here are a few examples:
Here are some stats:
5513 submitted (4990 in 2020)
1184 accepted: 166/1018 long/short talks (1088 in 202)
21.5% acceptance rate (21.8% in 2020)
Thread on the top authors, countries, organizations, etc. 🧵
Top authors:
Masashi Sugiyama is #1 for the second year (11 in 2020 -> 14 in 2021).
Sergey Levine is #2 (5 -> 13)
Z. Yang, Z. Wang, G. Niu are # 3 (all with 11 accepted)
Top organizations:
Google, MIT, Stanford, Berkeley, Microsoft are top-5.
DeepMind (#9): 51 in 2020 -> 38 in 2021.
Compared to virtual #ICLR2020, I found virtual #ICML2020 lacks a few features.
1. Papers are scheduled for 2 days. instead of 1. I prefer to collect all papers of interest for the day and then attend only those. Now I need to keep in mind if I attended a poster previously.
1/n
2. Videos are 10-15 minutes long. This forces me to *really* want to attend the poster. Having tl;dr version (<5 min) or short/long videos as in ICLR is preferred.
3. 1 hour of the poster session is often not enough for the amount of content proposed. In session 1, I have 5+ posters, 15 minutes each. It physically doesn't fit into the slot.