robust language understanding is about building discourse models across multiple sentences (see eg https://t.co/C6XodPts1W & recent thread by @mpshanahan)
— Gary Marcus (@GaryMarcus) October 27, 2019
on simple word problems like these, GPT-2 did not succeed once in 30 min testing
my 6-yr old son scored 100% pic.twitter.com/FLN31mdhcU
Does #deeplearning "understand" natural language? Has common sense been solved? Try for yourself at https://t.co/8LkSUMTrDv, from @AdamDanielKing.
— Gary Marcus (@GaryMarcus) October 26, 2019
Here's a sample from this morning, my question in bold, transformer network answer at bottom, starts "It's the same as... " pic.twitter.com/SMd8VDLh0h
text-to-text-transfer-transformer - https://t.co/ts5xMqoo9q
— Python Trending (@pythontrending) October 25, 2019
Contextual Word Embedding is now part of Google Search https://t.co/l7Lp49RosE Terrific insight and context by @tsimonite
— Oren Etzioni (@etzioni) October 25, 2019
🚨🚨Big #nlproc claim from Google: "we .. [introduce] a unified framework that converts every language problem into a text-to-text format." https://t.co/znJ5DefCoG pic.twitter.com/pBZrdompF9
— Delip Rao (@deliprao) October 24, 2019
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer https://t.co/1oyuqNC8hG
— /MachineLearning (@slashML) October 24, 2019
One year ago, @seb_ruder asked the https://t.co/GEOZuodrZj community for help with multilingual language modeling.@eisenjulian @PiotrCzapla @misterkardas answered the call, and we now have MultiFiT, an EMNLP paper & code for multilingual training!https://t.co/FwY6QGV0gA pic.twitter.com/JYQRpz8wHT
— Jeremy Howard (@jeremyphoward) October 23, 2019
Most of the world’s text is not in English. We are releasing MultiFiT to train and fine-tune language models efficiently in any language.
— Sebastian Ruder (@seb_ruder) October 22, 2019
Post: https://t.co/LXMo4KGcVS
Paper: https://t.co/3kh3fpfKcj
With @eisenjulian @PiotrCzapla Marcin Kadras @GuggerSylvain @jeremyphoward pic.twitter.com/QtcWhKqxyL
there, you have them on one slide pic.twitter.com/DiEuULhtKB
— Thomas Wolf (@Thom_Wolf) October 21, 2019
The Illustrated GPT-2 (Visualizing Transformer Language Models) https://t.co/85cGvGRwb9
— Nando de Freitas (@NandoDF) October 20, 2019
💥 @LysandreJik just published a great comparison of inference speed & memory requirements for a wide range of recent NLP models:
— Thomas Wolf (@Thom_Wolf) October 18, 2019
- RoBERTa, Bert, GPT2, XLNet, DistilBert, CTRL...
- comparing PyTorch & TensorFlow 2.0
- on typical CPU and GPU setups
- summarizing good-practices👇 https://t.co/2G3cYaUozL
Machines’ Language Skills Just Surpassed Our Own. But Is This Understanding?
— Sebastian Ruder (@seb_ruder) October 17, 2019
Nuanced take by @QuantaMagazine on the latest BERT-based models and what this means for NLP, with comments from @sleepinyourhat @annargrs @omerlevy_ and others https://t.co/uVPNr4PqYo
