12-in-1: Facebook AI’s New Framework Tackles Multiple Vision-and-Language Tasks
The model reduces the number of parameters from some 3 billion to 270 million while improving task performance by an average of 2.05 points.
Category: AI
Global machine intelligence updates.
2019 in Review: 10 AI Failures
Synced 10 AI Failures of 2019
Visual Dialogue State Tracking Improves Question Generation
Researchers from Beijing University of Posts and Telecommunication have introduced a novel visual dialogue state tracking (VDST) model that’s got pretty good at the similar, visual dialogue guessing game “GuessWhat?!”
PyTorch Deep Learning Framework: Speed + Usability
A new paper from original PyTorch developers explores the inspiration behind the library, and makes the case for its unique marriage of speed and usability.
NeurIPS 2019 Roundup: Outstanding Papers, Featured Talks, Facts and Figures
Synced Global AI Weekly December 15th
Attention Please! Remastering Vintage Films With Temporal Source-Reference Attention Networks
Researchers from the University of Tsukuba and Waseda University in Japan recently introduced a single framework designed to tackle an entire range of remastering tasks for films that have been converted to digital data.
Challenge On! Tech Giants Unite to Fight Deepfakes With AI
At NeurIPS 2019 this week Facebook unveiled its Deepfake Detection Challenge (DFDC), along with a new deepfake-specific dataset containing more than 100,000 videos.
KDD 2020 Opens Call for Research and Applied Data Science Paper Submissions
ACM SIGKDD Invites Industry and Academic Experts to Submit Advancements in Data Mining, Knowledge Discovery and Machine Learning for 26th Annual Conference in San Diego
AI Reimagines Ancient Chinese Poetry
Due to the nuanced character choices and other unique literal and aesthetical characteristics, automatic generation of Chinese poetry is challenging for AI, and high-quality poems can hardly be generated by end-to-end methods.
Baidu’s Pre-training Model ERNIE Achieves New NLP Benchmark Record
ERNIE has achieved new state-of-the-art performance on GLUE and become the world’s first model to score over 90 in terms of the macro-average score (90.1).
Stanford, Kyoto & Georgia Tech Model ‘Neutralizes’ Biased Language
Pryzant and other Stanford researchers partnered with researchers from Kyoto University and Georgia Institute of Technology to develop a novel natural language model that can identify and neutralize biased framings, presuppositions, attitudes, etc. in text.
Japanese Unicorn Preferred Networks Migrates Its DL Platform to PyTorch
Preferred Networks is migrating its deep learning research platform from its own open source framework Chainer to PyTorch.
Playing Space Invaders Blind | RL & Cross Modality Transfer
Using cross-modality transfer techniques and reinforcement learning (RL), the researchers produced an agent that can play video games with only the game audio to guide it.
NeurIPS 2019 Opens; Outstanding Paper Awards and 2021 Conference Announced
In a blog post, the NeurIPS organizing committee announced the conference’s outstanding paper awards and other awards.
What to Expect at NeurIPS 2019; Google Founders Leave Alphabet; ML Maestro Michael I. Jordan Wins John von Neumann Medal
Synced Global AI Weekly December 8th
Vancouver SkyTrain Shutdown? Bad Timing for NeurIPS Attendees
The City of Vancouver is expecting a traffic spike next week with finals approaching and thousands of attendees pouring into the Convention Center from December 8 to 14 for top-tier AI conference NeurIPS 2019.
How Face Recognition Tech Is Driving the Short Video Trend
Face recognition will continue to play a big role as short video platforms develop functions to appeal to new target groups and provide more innovative, personalized and engaging content to users.
Weekly Papers | Fine-Grained Neural Architecture Search; Adversarial Examples Boost Image Recognition; Judea Pearl Lauds “Comprehensive, Delightful, Inspiring” Max Planck Paper, and More!
On Fridays, Synced selects seven studies from the last seven days that present topical, innovative or otherwise interesting or important research that we believe may be of special interest to our readers.
King’s and University of Surrey | Deep Learning Improves Structural Health Monitoring of Civil Infrastructure
This research demonstrated that deep learning can contribute to the traditional discipline with much better performance than existing methods.
Personal Invisibility Cloak Stymies People Detectors
Researchers from Facebook AI and the University of Maryland have introduced a 21st century version — sweatshirts printed with adversarial examples that make the wearer undetectable to the AI-powered object detectors in today’s public surveillance systems.
Deep Learning Model Morphs VTube Talking Heads With a Few Mouse Clicks
A Google researcher has released a deep neural network model that makes animating a VTube persona a little easier.
Can Simple Neuron Interactions Capture Complex Linguistic Phenomena?
Researchers from The Chinese University of Hong Kong, Tencent AI Lab and University of Macau have proposed a new neuron interaction based representation composition for NMT.
CAH Black Friday AI Challenge; AI-Generated Thanksgiving Dinner Recipes; Go Master Lee Sedol Retires Because AI
Synced Global AI Weekly December 1st
Fighting Fire & Floods With Smart Emergency Systems
Simulation systems can play a valuable role in identifying and testing evacuation plans to enable individuals to promptly leave a dangerous area via the safest and fastest route.
Weekly Papers | Multi-Label Deep Forest (MLDF); Huawei UK Critiques DeepMind α-Rank
Synced selects seven studies from the last seven days that present topical, innovative or otherwise interesting or important research that we believe may be of special interest to our readers.
Herring, Not Herring: Deep Learning Accelerates Detection and Classification of Underwater Species
The collaboration produced a deep learning framework for the automatic detection of schools of herring from echograms that outperforms traditional hand-crafted machine learning algorithms on precision, recall, and F1-score metrics.
Google & Johns Hopkins University | Can Adversarial Examples Improve Image Recognition?
Rather than attempting to defend convolutional networks against them, the researchers introduce a novel enhanced adversarial training scheme, AdvProp, which treats adversarial examples as additional training examples to improve the accuracy of image classification models.
AI Helps Quantum Chemists Determine Molecular Wave Functions
An interdisciplinary team of chemists, physicists, and computer scientists have developed a deep learning algorithm that leverages fundamental quantum mechanics equations to accurately predict the quantum mechanical wave functions of molecules.
Do Human Gamers Stand a Chance Against Trash-Talking AI Bots?
Synced Global AI Weekly November 24th
AI in the E-Commerce Industry
E-commerce is a fast-growing industry. In 2017, global retail e-commerce sales reached US$2.3 trillion, and e-retail revenue is expected to grow to US$4.88 trillion in 2021.
Weekly Papers | Quoc V. Le and Kaiming He Look at Vision; Intelligence, Psychology and AI; Evolving the Hearthstone Meta and More!
Synced has surveyed last week’s crop of papers in the fields of machine learning, computer vision, computation and language, and beyond, and identified seven studies that we believe may be of special interest to our readers.
AI Is Tearing Up the Dancing Floor Again
For decades machines have been able to understand simple musical features like beats per minute. Now AI is boosting their abilities to the point that they can not only figure out what particular genre of music is playing, but also how to appropriately dance to it.
Voila! SOTA French Language Model ‘CamemBERT’ Debuts
Now, a team from Facebook AI Research, Inria, and Sorbonne Université have released CamemBERT, essentially a French version of Google AI’s game-changing pretrained language model BERT.
Google Brain’s Nicholas Frosst on Adversarial Examples and Emotional Responses
Although convolutional neural networks have reached Superman level on image classification tasks, adversarial examples remain the kryptonite that can mysteriously defeat even SOTA models.
Big RNNs Achieve SOTA Performance in Video Prediction
It’s not as easy as one might imagine to train an AI model to accurately predict what a human will do next, even when they are interacting with a relatively simple object like a ball.
Can Bots Surpass the ‘Realism’ of Human Dialogue?
Microsoft’s new tunable gigaword-scale neural network DialogGPT is a virtual master of conversation that outperforms strong baseline systems in generating relevant and context-consistent responses and attains near human level performance in conversational response generation tasks.
Google Brain’s Hugo Larochelle on Few-Shot Learning
In fact, the accuracy of results in few-shot learning, both with and without labeled data, is very high.
Huawei Tops ETH Zurich 2019 Smartphone Deep Learning Rankings
In an effort to benchmark deep learning on smartphones in the Android ecosystem, researchers from the ETH Zurich Computer Vision Lab last year developed an AI Benchmark application to measure the AI readiness of more than 200 Android devices and 100 mobile SoCs collected in the wild.
Google Health Data Backlash; Microsoft Head of AI and Research Splits; Bengio vs Marcus in Christmas Debate
Synced Global AI Weekly November 17th
AI Struts Down the Japanese Fashion Runway
Although Japan’s declining and aging population has done little to stimulate the domestic fashion industry, AI is getting very positive reviews from the country’s still massive JP¥9 trillion (US$83 billion) apparel market.