Replying to @serena31649 @BeALonghorn
UT Office of admissions needs to consult its Pyschology department and understand the delirious effects of Anticipatory Anxiety on mental health of applicants (and some of them will be your future students). Stop using 'no later than' and use 'will be announced on.' #UT27
Replying to @BeALonghorn
Wave today. Its official. Check MyStatus at 7 PM Central. Ask someone to record your reaction when you see status change at 7 PM. #Longhorns
2. If your results are worse than SOTA, say that SOTA results are suspicious if the code and settings are not available.
3. Learn the use of functions like min and max. Use min of all errors for your method and max of all errors for competitors.
#NeurIPS2022 #ICLR2023
ICLR papers are due next week. If your NeurIPS paper was rejected and you got at least one 7, you just need to hope and pray for the right reviewers. Some simple suggestions to improve your paper:
1. Keep using confusing words that sound important but don't mean anything
1
I attended @WiMLworkshop three times. It is a waste of time. It is about women at Stanford, Caltech, etc. bragging about how successful they are. If you want real diversity, mentor that introvert PhD student and help her succeed rather than feeding her falsehoods. #NeurIPS2022
Replying to @NeurIPSConf
@NeurIPSConf requires privileged citizenship, sorry. It is a very selective conference. Even a workshop that promotes diversity such as @WiMLworkshop requires employment at Stanford, CMU, Amazon, Google, MSR, etc. Even diversity is privileged. #NeurIPS2022
Why do we hold #NeurIPS2022 in a closed-bordered country such as the USA? This greatly impacts many researchers with less privileged citizenships.
@NeurIPSConf
What do existing causal models have in common with Neural Networks. Both suck at causal inference. See new papers on causal models accepted by friends and family @NeurIPSConf. #NeurIPS2022
MNIST was banned this year, see #NeurIPS2022 code of conduct. Have you heard about ImageNet?
Replying to @thegautamkamath
Abolish the ACs, replace them with causal or casual Neural networks. #NeurIPS2022
Replying to @Montreal_AI
Next time, answer the question in the title. Why should I read an entire paper to get to know a simple answer.
Replying to @harsha_kokel
In future, use the Lovasz (don't pronounce it) extension of SMI to make this paper more general. #NeurIPS2022
To all my fellow @NeurIPSConf attendees, Mardi Gras is not until Feb 21, 23. Therefore NeurIPS this year is beadsless. The only reason to need and want beads if you are playing against Alpha Go or @MagnusCarlsen #NeurIPS2022
2
Don't malign overfitting. Overfitting is guaranteed to yield SOTA results and get your paper accepted. ICLR is right around the corner. If your NeurIPS 2022 paper #NeurIPS2022 didn't get in, give overfitting on the test data a try, you will love the results. #ICLR2023
1
Replying to @NeurIPSConf
@NeurIPSConf has a code of conduct. Please abide by it. Don't harass people who overfit on test data and tell them to do research on causal inference. This will not be tolerated. #NeurIPS2022
1
This is a paper about causality, 70% acceptance rate via friends and family discount. Thus, the acceptance of each paper is non-IID but acceptance rate is often treated as IID for making various decisions such as graduation, tenure, etc. That is why we need causal inference.
Replying to @etzioni
What makes you think Elon Musk understands what Artificial General Intelligence means. We cannot define AI and you won't be able to define AGI either. If you ask NLP PhD students, not researchers, they will say GPT-3 is AGI.
Replying to @dchiji_en
Interesting paper but too bad it cannot do one-shot learning without having a few shots. #NeurIPS2022
1
Replying to @DeepLearnScam @qiqihelloworld
On that note may be have two independent ACs and take a max. The ACs write 10 lines per paper and never read any papers.
1


