After scanning through the news coverage of the German federal election in all the major online newspapers I must say I liked @zeitonline the best. Great online video analysis as well. Not that data coverage in other sources was particularly bad… https://t.co/0b6GoJbSkh
— Simon Kuestenmacher (@simongerman600) September 27, 2021
Transformers Generalize Linearly
— AK (@ak92501) September 27, 2021
abs: https://t.co/ud0iUEYDyx
Transformers fail to generalize hierarchically across a wide variety of grammatical mapping tasks, but they exhibit an even stronger preference for linear generalization than comparable recurrent networks pic.twitter.com/VzbM2SQTZl
Very interesting thread on plagiarism in machine learning papers at top conferences in this post on r/ML
— hardmaru (@hardmaru) September 26, 2021
The comments are especially quite revealing, with links to detailed discussions on Chinese discussion forum Zhuanlan Zhihu, in case you read Chinese.https://t.co/RD5VkUfrnx
I don't know who needs to hear this, but proofreading something twice with two different fonts and large/smaller font sizes helps you catch more typos. The effect of font size is especially large.
— François Chollet (@fchollet) September 25, 2021
Muzic: Music Understanding and Generation with Artificial Intelligence
— AK (@ak92501) September 25, 2021
github: https://t.co/2XtiUh3h8e pic.twitter.com/4PTPX4qMia
Amusing! Object detection cast naively into language modeling framework + borrowing many of the tips&tricks.
— Andrej Karpathy (@karpathy) September 24, 2021
- random object ordering seems fine ✅
- coords, class labels flattened into a single softmax 😂
- sequence augmentation is the most gnarly part, almost as yucky as nms 😬 https://t.co/FxSz5UbpxY
AI is a marketing term meant to intimidate and engender blind trust. https://t.co/EvSP02sy2I
— Cathy O'Neil (@mathbabedotorg) September 23, 2021
Have you wondered why object detection, unlike classification, has so many sophisticated algorithms?
— Ting Chen (@tingchenai) September 23, 2021
With Pix2Seq (https://t.co/ygsG3aAIbG), we simply cast object detection as a language modeling task conditioned on pixels!
(with @srbhsxn, Lala Li, @fleet_dj, @geoffreyhinton) pic.twitter.com/aTYZ5IvJc9
Elegant way of showing global fertility data for every country in 2019. It's actually a rather elegant visual. Spend a few minutes studying it and might become as big as a fan of the piece as I am. Great work by @SeppalaVilleEN. Source: https://t.co/UKJK0IWBn5 pic.twitter.com/gASjveZvkq
— Simon Kuestenmacher (@simongerman600) September 22, 2021
Neural Distance Embeddings for Biological Sequences
— AK (@ak92501) September 22, 2021
abs: https://t.co/fvlN4fcGpB
github: https://t.co/Z0Ok68cWhH pic.twitter.com/nmNGqO8IRy
If you’re a data scientist working for terrible people, know that your work is actively pointless and being fully ignored.
— Angela Bassa (@AngeBassa) September 20, 2021
It’s way past time for an ethical framework for our profession. https://t.co/l4DFnV3w2W pic.twitter.com/CqCxFSsdBM
Primer: Searching for Efficient Transformers for Language Modeling
— AK (@ak92501) September 20, 2021
abs: https://t.co/JM9v7pNoSI
github: https://t.co/xhA7uGyC7H
Experiments show Primer’s gains over Transformer increase as compute scale grows and follow a power law with respect to quality at optimal model sizes pic.twitter.com/CXq1yYMfUA
