Trending Tweets
Have you read the preview version of Deep Learning with Python, 2nd edition? Would you like to provide a quote we'll display on the back cover? Email me your quotes! (my email is on my website). Provide your name, your title & company/affiliation. Quotes should be 1-2 sentences.
#IgluContest competition at #NeurIPS2021 approaches how to build interactive agents that learn to solve a task while provided with grounded natural language instructions in a collaborative environment. There is still plenty of time to submit your solution to contribute to "cake":
“Biological cell-cell communication circuits … use an architecture that is the opposite of what we synthetic biologists might have designed.” — Michael Elowitz, systems biologist at Caltech https://t.co/tXH5q0kDF3
FAIR MPK Reinforcement Learning (RL) team has full-time research scientist opening. We study fundamental research problem in ML/RL-guided optimization and its real-world applications. Look for candidates in RL, optimization, and representation learning. POC: yuandong@fb.com
Great talk by @Terrybroad, summarising techniques for "Active Divergence with Generative Deep Learning", e.g. for artistic purposes - plenty of examples straight from the art world. Joint work with @sebastianberns @SimonGColton @iccc_conf #iccc21
Definitely agree. I've met interviewees who over-prepped, and it felt a bit inauthetic? I mean if our work is adjacent, some knowledge may be warranted. But if you work in architecture and know details of my proofs in Appendix C, it feels like you're just gaming the interview
We investigated whether NLI model can help us verify QA model predictions. It was cool to make use out of decontextualization system, and to identify many cases where QA models output correct answers for the wrong reasons! led by @Jifan_chen & @gregd_nlp
Check out Jifan's latest work! Continuing on his NAACL 21 work (https://t.co/iwmiavobSY), we've been thinking about how to improve robustness of QA predictions. Eunsol's decontextualization model (https://t.co/dObC0MovQj) ended up being a really useful piece to reframe QA as NLI!
congratulations to @bricksdont for successfully defending his thesis on Robust Neural Machine Translation today! with many thanks to the external examiner @mjpost!
Random Forests, Stochastic Parrots 🦜, The Pile, catastrophic forgetting: the whole field behaves like a science fiction writer who took a marketing course while dropping acid.
By inverse scaling, I mean the opposite behavior shown in OpenAI's scaling laws (https://t.co/GVzMlmP07q) - that performance gets worse as model size, data size, and compute increase
🥁 We can't wait to share our new inference product with you! 🤩 - it achieves 1ms latency on Transformer models 🏎 - you can deploy it in your own infrastructure ⚡️ - we call it: 🤗 Infinity 🚀 📅 Join us for a live event and demo on 9/28! https://t.co/fvhb86gsG7
ANSWER: A lively discussion on opportunities and challenges in human-oriented AI with Microsoft researchers, leading academics, and industry experts in reinforcement learning. What's the QUESTION?
Friday's Community Spotlight! ⚡️💜 3D object segmentation with self-supervised learning: https://t.co/wcAagyfUKe
Really happy about this work! We create counterfactual representations by taking a mirror image with respect to linear "concept subspaces", and use them to test linguistic hypotheses on the working of NNs. Check out the thread for more details! https://t.co/1Xl9XjaGDe
Multi-task RL is hard. Multi-task offline RL is also hard. Weirdly, sharing data for all tasks (and relabeling) can actually make it harder. In conservative data sharing (CDS), we use conservative Q-learning principles to address this. arXiv: https://t.co/4hhl639bgh A thread:
Varied, dynamic scene graphs of points, objects and structures are the right goal representation for #SpatialAI. We show how Gaussian BP on factor graphs can tackle the true, complex inference problem this leads to, with efficient parallel implementation on a @graphcore IPU.
Today I learned about the Superformula, which can be used for all kinds of biological shapes #JuliaLang function superformula(a,b,m,n₁,n₂,n₃;Δθ=1e-4) θ = 0:Δθ:2π r = @. (abs(cos(m*θ/4)/a)^n₂+abs(sin(m*θ/4)/b)^n₃)^(-1/n₁) return @. r * cos(θ), r * sin(θ) end
We know SOTA NLMs encode at least some useful linguistic properties. But do they *use* these properties as we might expect them to? In work accepted at CoNLL, @ravfogel @tallinzen @yoavgo and I propose a method to explore this question! Preprint: https://t.co/9Lw0UYBDQT 🧵1/N
I will be teaching a course on robot learning, starting in January 2022! This course will focus on deep reinforcement learning methods and their application to robotics. You can find more info on the course here: https://t.co/Sr1ic5KpRG
Come work with me and my amazing colleagues @j_dunkley, Romain Teyssier at @PU_Astro. We're looking for postdocs with new computational, statistical, or deep learning approaches to cosmology. Deadline Nov 1 https://t.co/ZqXaY2BRHx
Tevatron is a new toolkit for training and running dense retrieval models by @luyu_gao and @xueguang_ma - check it out! https://t.co/M2HA8Y09ya
We are pleased to share the materials of our ECML-PKDD 2021 tutorial on Machine Learning and Reasoning for Drug Discovery.: * Tutorial description: * Slides: https://t.co/BEJ4UMj3ub #drugdiscovery #artificialintelligence #machine…https://t.co/8AqNNsCKPt https://t.co/seLd8d2qGw
Somin Wadhwa with @_hamedzamani presenting “Towards System-Initiative Conversational Information Seeking” at #DESIRES2021
You can do exactly the same with models weights by the way!
Introducing #TensorFlow Similarity - a newly-released library from Google that facilitates the training, indexing and querying of similarity model https://t.co/sIU3kCsxNv
✨✨✨Quantum teleportation transfers a quantum state between two parties without sending any of its parts. The catch? A classical channel is still required to transfer information, making instantaneous communication for now only possible in science fiction.#QuTechAcademy
"How Does Counterfactually Augmented Data Impact Models for Social Computing Constructs?" analyses reasons CAD is beneficial for social NLP tasks (sentiment, sexism, hate speech) @indiiigosky @hide_yourself @ffloeck @clauwa @IAugenstein https://t.co/LKnRrYI43e #EMNLP2021 #NLProc
Successful Conversational AI: Blending Machine Learning & Human Intelligence https://t.co/Yyh4SLfi72 via @AITimeJournal
Trending Now
Microsoft Research
Founded in 1991, Microsoft Research is dedicated to conducting both basic and applied research in computer science and software engineering.
61.2
François Chollet
Deep learning @google. Creator of Keras. Author of 'Deep Learning with Python'. Opinions are my own.
58.3
Max Tegmark
Known as Mad Max for my unorthodox ideas and passion for adventure, my scientific interests range from artificial intelligence to the ultimate nature of reality
49.5
47.5
47.5
Joscha Bach
FOLLOWS YOU. Artificial Intelligence, Cognitive Architectures, Computation. The goal is integrity, not conformity.
45.3
Thomas Wolf
Co-founder & Chief Scientist at @HuggingFace – I lead the Open-Source & Science teams – 🤗Transformers & 🤗Datasets libraries – @BigScienceW research workshop
44.4
Jimmy Lin
I profess to know very little at the University of Waterloo. I used to write code for Twitter and slides for Cloudera.
43.7
43.3
Matthew Mayo
Comma separated valedictorian; flat file fanatic
#MachineLearning #DataScience #NLProc @kdnuggets
42.7
Ted Underwood
Using machine learning to study literary imagination, and vice-versa. Distant Horizons (UChicago, 2019). Information Sciences (@iSchoolUI) / English.
40.3
PyTorch Lightning
The lightweight PyTorch AI research framework. Scale your models, not the boilerplate! Use our platform @gridai_ to scale models from your laptop to the cloud.
37.0
Creative Virtual
Award-winning chatbot, virtual agent, live chat & conversational AI solutions for better customer, employee & agent engagement. #CX #EX #AI #ConversationalAI
36.3
QuTech
We keep you posted on our road to develop and foster Quantum Computing and Quantum Internet. The future is #quantum
33.2
Gautam Kamath
Professor of Computer Science @UWaterloo, affiliate @VectorInst. I lead @TheSalonML. ML, Stats, Robustness, Privacy. Prev: @SimonsInstitute @MIT @Cornell. 🇨🇦
32.8
Andrew Davison
From SLAM to Spatial AI; Professor of Robot Vision, Imperial College London; Director of the Dyson Robotics Lab; Co-Founder of SLAMcore.
32.6
CopeNLU
University of Copenhagen Natural Language Understanding research group, led by @IAugenstein #NLProc #ML #dlearn
27.7
27.0
27.0
Eunsol Choi
work on natural language processing / machine learning. assistant professor @UTCompSci, previously at @googleai, @uwcse, @Cornell. all opinions are of my own.
26.6
Emanuele Rossi
Researcher at Twitter and PhD student at @imperialcollege. Working on graph deep learning. Previously at Fabula AI (acquired by Twitter). @Cambridge_Uni alumnus
25.2
24.5
23.3
Christian Guckelsberger
Computer Scientist/Art Historian/Postdoc in Interactive Artificial Intelligence. Researching AI acting creatively in its own right, via intrinsic motivation.
22.6
21.4
Gianmaria Silvello
Information Retrieval, Data Citation and Provenance, Digital Libraries and Archives researcher @ University of Padua, Italy
20.4
Ethan Perez
PhD student in NLP at @nyuniversity advised by @kchonyc and @douwekiela. Current intern at Google DeepMind. All views my own.
20.4
Truyen Tran
Associate Professor of Artificial Intelligence at Deakin Uni, Australia. Super curious about intelligence, science and living things.
16.6
Grusha Prasad
A CogSci PhD student at JHU interested in using psycholinguistic and computational approaches to think precisely about language. Also a lover of puns. She/her
13.5
Peter Melchior
Asst. Prof. of Statistical Astronomy at @Princeton, @PU_Astro & @PrincetonSML. I weigh cosmic structures, value every pixel, express my own views.
13.5
Glen Berseth
Postdoc at Berkeley Artificial Intelligence Research Lab @berkeley_ai performing research in deepRL and robotics. Received my Ph.D. from UBC. he/him
11.1
11.1
Michiel ☁️
Machine learning and computational biology researcher. Develops models to understand ecology and synthetic biology. Likes 📚, ☕, 👨🍳, 👨💻 and ☁️s.