California Startup Builds Worlds Largest AI Chip; Open AI Releases 774M Parameter Language Model; Waymo Shares Its Open Dataset
Synced Global AI Weekly August 25th
Tag: technology
Huawei’s First Commercial AI Chip Doubles the Training Performance of Nvidia’s Flagship GPU
Ascend 910 delivers performance of up to 256 teraFLOPS under FP16 and 512 teraOPS under IN8 with declared max power consumption of 310W.
MIT & Adobe Introduce Real Time AR Tool for Storytelling
Researchers from MIT Media Lab and Adobe Research recently introduced a real-time interactive augmented video system that enables presenters to use their bodies as storytelling tools by linking gestures to illustrative virtual graphic elements.
Shared Machine Learning: Ant Financial’s Solution For Data Privacy
One of the world’s largest financial companies, Alibaba subsidiary Ant Financial, recently introduced Shared Machine Learning (SML) as their solution for data privacy.
How to Hide Your Feelings From AI Voice Assistants
No matter whether the conversations commence with “Ok, Google”, “Hello Alexa”, or “Hey Siri”, tech giants Apple, Amazon, Facebook, Google and Microsoft have all granted third-party contractors access to their users’ many interactions with voice assistants.
Seoul National University AI Uses Text Tags to Colourize Line Drawings
Let’s say you’ve just sketched a portrait but don’t have the inclination to hand-colour your creation. Wouldn’t it be nice if you could simply tell an artificial intelligence assistant what colours you want and have it do the rest? Well, now you can!
Why Automation Isn’t the Answer to Smart Manufacturing
While management teams and consultants may spout trendy tech terms such as “process”, “automated”, or “smart”, and seem keen to hand over precisely-timed, measured tasks to machines, the truth is that AI tech is not a panacea — because manufacturing is hard.
Israeli AI Research Company Led by Stanford Professor and MobileEye CEO Introduces Model to Solve Text Ambiguity
Israeli research company AI21 Labs today published the paper SenseBERT: Driving Some Sense into BERT, which proposes a new model that significantly improves lexical disambiguation abilities and has obtained state-of-the-art results on the complex Word in Context (WiC) language task.
Does Deep Learning Still Need Backpropagation?
Now, researchers from the Victoria University of Wellington School of Engineering and Computer Science have introduced the HSIC (Hilbert-Schemidt independence criterion) bottleneck as an alternative to backpropagation for finding good representations.
New Snip Smartphone App Converts Math Screenshots Into LaTeX
Not so long ago, mathematics students and researchers had to spend a tedious amount of time writing out equations in the technical and scientific documentation typesetting system LaTeX.
PyTorch 1.2 Supports Transformer and Tensorboard; Summer Hackathon Announced
PyTorch v1.2 brings the machine learning community further improvements, including official support for Transformer, TensorBoard, and more.
Tencent-backed Video App Kuaishou Is Turning Chinese Country Folk Into Hollywood Directors
Thanks to social video app Kuaishou and its plug-and-play tools, even rural Chinese with cheap smartphones are quickly and easily producing short videos featuring Hollywood-style special effects.
AI Analyzes Chess Commentary to Learn to Play Chess
What if an AI could learn to play Chess based only on human game commentary? That interesting task has been tackled by SentiMATE, a new Chess algorithm proposed by University College London.
Robot Masters String Puppetry
A group of researchers from ETH Zurich have introduced a robotic system that can perform animation of real-world string puppets, aka marionettes.
Baidu’s ERNIE 2.0 Beats BERT and XLNet on NLP Benchmarks
Today, the research division of the Chinese search giant released their updated ERNIE 2.0, a pretrained language understanding model with significant improvements.
Julia Computing & MIT Introduce Differentiable Programming System Bridging AI and Science
A research team from Julia Computing and the Massachusetts Institute of Technology propose that extensive scientific computing and machine learning domains both require linear algebra support on their underlying structures.
Deep TabNine: A Powerful AI Code Autocompleter For Developers
The developers behind TabNine have introduced a new deep learning model, “Deep TabNine,” which significantly improves suggestion quality.
Harvard’s Tiny, Solar-Powered Flying Robot
Last month, Harvard Microrobotics Lab researchers introduced RoboBee X-Wing, a solar-powered micro-aerial vehicle that fits in the palm of your hand.
‘An Absolute Monster Bluffer’ – Facebook & CMU AI Bot Beats Poker Pros
Don’t simply “all in” if there’s a bot at your Texas hold’em poker table, because Facebook and Carnegie Mellon University’s new *Pluribus* AI system just beat five human pros at the same time — including a couple of World Series of Poker Champs.
New Repository Provides NumPy Implementation of ML Models
If you are a developer who has ever wished for a rich collection of machine learning algorithms implemented exclusively in NumPy, then you’ll love Numpy-ml — a new GitHub project that’s winning accolades from across the global machine learning community.
Alibaba, Baidu and Xiaomi Open-Source Their SOTA Research
Synced Global AI Weekly July 7th
DeepMind Transporter: Unsupervised Learning of Object Keypoints
In common reinforcement learning (RL) domains, existing keypoint discovery methods struggle to perceive keypoints in an image when there are limited objects present.
Local Temporal Bilinear Pooling for Fine-Grained Action Parsing
Parsing fine-grained temporal actions is vital in application scenarios that require an understanding of detailed and precise operations over long-term periods, such as daily activity understanding, surgical robots, human motion analysis and animal behavior analysis.
The Staggering Cost of Training SOTA AI Models
While it is exhilarating to see AI researchers pushing the performance of cutting-edge models to new heights, the costs of such processes are also rising at a dizzying rate.
AI Pioneer Li Deng Named Canadian Academy of Engineering Fellow
Citadel Chief AI Officer Li Deng has been named a Fellow of the Canadian Academy of Engineering (CAE) in recognition of his notable achievements in deep learning and speech recognition.
New Facebook PyTorch Hub Facilitates Reproducibility Testing
In a bid to provide a smoother reproduction experience, Facebook has announced the beta release of PyTorch Hub, a new pretrained model repository designed to facilitate research reproducibility testing.
Alan Turing Institute Releases ML Framework Written in Julia
UK-based national research organization The Alan Turing Institute has released a new machine learning toolbox, Machine Learning in Julia (MLJ), which provides a uniform interface enabling users to easily train, evaluate, and tune machine learning models.
OpenCV Launching Kickstarter Campaign for New AI Courses
The non-profit organization behind the popular worldwide library of computer vision programming functions, OpenCV (Open Source Computer Vision), is launching a kickstarter campaign to raise funds for a series of summer 2019 AI courses.
Robotic Catheter Autonomously Navigates in a Beating Heart
New research from Harvard Medical Group shows AI-powered autonomous device navigation is possible in minimally invasive surgical procedures, and even in heart surgery.
Close Your Eyes and I’ll Scan You: Chinese Face Payment System Vulnerabilities
Everyone, however, sleeps. And Synced has found a simple way to challenge cutting-edge facial recognition security scans: just point a camera at the unaware subject while they are sleeping.
TensorFlow, PyTorch or MXNet? A comprehensive evaluation on NLP & CV tasks with Titan RTX
Thanks to the CUDA architecture [1] developed by NVIDIA, developers can exploit GPUs’ parallel computing power to perform general computation without extra efforts. Our objective is to evaluate the performance achieved by TensorFlow, PyTorch, and MXNet on Titan RTX.
Stanford ML Releases MRNet Knee MRI Dataset
The Stanford ML Group led by Andrew Ng has released its MRNet Dataset, which contains more than 1000 annotated knee MRI Scans; and announced an associated public model competition.
Upgrading CNN With OctConv
Researchers from Facebook, the National University of Singapore, and the Qihoo 360 AI Institute have jointly proposed OctConv (Octave Convolution), a promising new alternative to traditional convolution operations. Akin to a “compressor” for Convolutional Neural Networks (CNN), the OctConv method saves computational resources while boosting effectiveness.
A Google Brain Program Is Learning How to Program
The idea of using machine learning to teach programs how to automatically write or modify code has always been tempting for computer scientists.
‘Snip’ Converts Math Screenshots Into LaTeX
A new math tool called “Snip” is creating a buzz on Twitter. Thousands of netizens are sharing the tool, which is being heralded as a “life changer” for scientific writing.
New Google Brain Optimizer Reduces BERT Pre-Training Time From Days to Minutes
Google Brain researchers have proposed LAMB (Layer-wise Adaptive Moments optimizer for Batch training), a new optimizer which reduces training time for its NLP training model BERT (Bidirectional Encoder Representations from Transformers) from three days to just 76 minutes.
AI Biweekly: ML & Ethics; Skepticism re AI-Powered Surgeries
10 AI News You Must Know from March W 3 – W 4
Let Boston Dynamics ‘Handle’ That Package
In a scene that looks like it’s from a sci-fi movie, a YouTube video posted today by robotics company Boston Dynamics shows a huge, ostrich-like robot “Handle” whirling round while deftly moving boxes in a warehouse. The video has garnered over 138,000 views in less than four hours.
AI Local Rainfall Nowcasting Using Weather Radar Maps
To provide accurate precipitation nowcasting, Beijing startup ColorfulClouds Tech is applying machine learning (ML) techniques using observed radar echo maps to generate high resolution, minute-by-minute rainfall forecasts on its ColorfulClouds Weather mobile app.
NVIDIA CEO Says No Rush on 7nm GPU; Company Clearing Its Crypto Chip Inventory
NVIDIA CEO and Co-Founder Jensen Huang says a rumored next-generation GPU architecture is not a priority for the company, and that he remains optimistic about clearing the chip inventory built up for cryptocurrency mining. Huang made the remarks in a press conference Tuesday at the GPU Technology Conference (GTC) in Santa Clara.