Google Lens, TPU v2, AI fake news, Learning common sense from videos, Deep Learning at Twitter, Google SyntaxNet, Nvidia Volta and more
ARTIFICIAL INTELLIGENCE NEWS #61 May 18th 2017
In the News
Thanks to machine learning, it’s becoming easy to generate realistic video, and to impersonate someone.
Two interesting A.I.-related announcements at Google's I/O event this week:
- Google Lens — point your camera at something and Google will try to tell you what you are looking at and give you contextual info
- 2nd generation TPU chips for faster A.I. training and inference. They will be available on Google Cloud Platform soon.
Yann LeCun says the next frontier in machine vision is software that learns just by observing the world.
Will AI come to play a role in courts?
Also in the news...
- Apple acquires AI company Lattice Data for $200M. More
- Cisco acquires conversational AI company Mindmeld for $125M. More
- Melinda Gates and Fei Fei Li want to "liberate AI from guys with hoodies". Interview
- Nvidia reaches out to VCs to fund and build the AI ecosystem. More
- Taser will use police body cameras to anticipate criminal activity. More
Open AI has trained a robotics system, trained entirely in simulation and deployed on a physical robot, which can learn a new task after seeing it done once.
Twitter provides interesting details on its deep learning-powered ranking algorithm used to create users' timelines.
Facebook published research results using a novel convolutional neural network approach for language translation that achieves state-of-the-art accuracy at nine times the speed of recurrent neural systems. Source code and trained systems are available on Github.
GLG is transforming the way the world’s top professionals learn. Watch our video series on AI's capacity to disrupt business featuring innovators from companies like Brightenai, Clarifai, and x.ai.
Do you have AI expertise? Join our membership of 500,000 professionals for free and start earning for what you know.
Software tools & code
SyntaxNet is an open-source neural network framework implemented in TensorFlow that provides a foundation for Natural Language Understanding (NLU) systems.
"Deformable Convolutions are a drop-in tool for neural networks to let you sample from a large and more disparate set of points over an image, potentially helping with more complex classification tasks. "
Nvidia unveiled the Volta GPU, a massive new GPU — 5120 CUDA cores and 15 TFLOP/s of single precision performance
While recent advancements in Deep Learning have been fueled by GPUs, there is speculation that application-specific circuits (ASICs) such as Google's TPU might overshadow GPUs in the future. NVIDIA has announced they are working on their own TPU and will open-source the resulting technology.
Brains & Neurons
Perceived images are hard to decode from fMRI scans. But a new kind of neural network approach now makes it easier and more accurate.
Google generating custom smileys/illustrations in Allo
Presentation by Frank Chen (partner at a16z) on the way companies are using A.I. today and what is just a few years away from productization in research labs.
This newsletter is a collection of AI news and resources curated by @dlissmyr. If you find it worthwhile, please forward to your friends and colleagues, or share on your favorite network!
Share on Twitter · Share on Linkedin · Share on Google+
Suggestions or comments are more than welcome, just reply to this email.