Kimera is a C++ library for robots to map their environment and figure out where the robot is using only camera images and inertial data. (This is called SLAM, "simultaneous localization and mapping".) Works on robots that use Robot Operating System (ROS). https://github.com/MIT-SPARK/Kimera/blob/master/README.md
"The researchers invented a 'visual deprojection' model that uses a neural network to 'learn' patterns that match low-dimensional projections to their original high-dimensional images and videos." http://news.mit.edu/2019/model-lost-data-images-video-1016
Subsystems that can be plugged together by robots to build large-scale structures. "The underlying vision is that just as the most complex of images can be reproduced by using an array of pixels on a screen, virtually any physical object can be recreated." http://news.mit.edu/2019/robots-large-structures-little-pieces-1016
Coanda effect hovercraft. Instead of blowing air underneath, it blows air on the outside. It works but before someone builds a life-size version and offers you a ride on it, you might consider its tendency to flip over. https://www.youtube.com/watch?v=ebqQxLwjWfY
"According to a new study from Oracle and Future Workplace, a clear majority of Americans (64 percent) would trust a robot more than a human manager, and 32 percent think that a machine will eventually replace their boss." https://insights.dice.com/2019/10/18/robot-boss-workers-prefer/
"Ono Food Co. announced this week that the first mobile restaurant powered entirely by robotic technology, called Ono Blends, will open later this month in Venice, California." https://www.forbes.com/sites/aliciakelso/2019/10/17/ono-food-co-brings-the-restaurant-industry-one-step-closer-to-full-automation/#67d710826d01
Detectron2 is a ground-up rewrite of Facebook's Detectron object detection system. "The platform is now implemented in PyTorch. With a new, more modular design, ..." https://ai.facebook.com/blog/-detectron2-a-pytorch-based-modular-object-detection-library-/
"Excessive activity in the brain is linked to shorter life spans." "Analyzing gene expression patterns [...] in donated brain tissue from hundreds of people who died at ages ranging from 60 to over 100." https://sciencebeta.com/neural-activity-life-span/
"This page tries to assemble all the research on Natural Language Processing (NLP) for native and indigenous languages of the American continent. Our languages are in danger, especially if they don't get involved in the new digital boom ..." https://pywirrarika.github.io/naki/
TensorFlow vs PyTorch: PyTorch dominates research. TensorFlow dominates industry. https://thegradient.pub/state-of-ml-frameworks-2019-pytorch-dominates-research-tensorflow-dominates-industry/
Drone Racing League's Racer AI was built for one purpose: to outmaneuver and outsmart the greatest human pilot in the world. And reign victorious. Machine over man.
OpenAI got a robotic hand to learn from scratch how to solve a Rubik's Cube with only one hand. https://www.youtube.com/watch?v=x4O8pojMF0w
The "Trolly Problem" as a real-life experiment. https://www.youtube.com/watch?v=1sl5KJ69qiA
Relativity Space is "a startup that wants to combine 3D printing and artificial intelligence to do for the rocket what Henry Ford did for the automobile." https://www.wired.com/story/massive-ai-powered-robots-are-3d-printing-entire-rockets/
Functional reinforcement learning. Functional in the "functional programming paradigm" sense. "The paradigm will be that developers write the numerics of their algorithm as independent, pure functions, and then ..." https://bair.berkeley.edu/blog/2019/10/14/functional-rl/
RLib is an open-source library for "scalable" reinforcement learning. "RLlib uses Ray actors to scale training from a single core to many thousands of cores in a cluster." https://ray.readthedocs.io/en/latest/rllib.html
Facebook open-sourced a system for managing network congestion with reinforcement learning. https://ai.facebook.com/blog/open-sourcing-mvfst-rl-a-research-platform-for-managing-network-congestion-with-reinforcement-learning/
AI chips can have "hot spots" with temperatures 15-20 degrees (8-11 degrees Celsius) higher than the rest of the chip. AI chips are highly repetitive but usage is bursty and the bursts can cause sudden floods of current into a particular part of the chip. https://www.youtube.com/watch?v=E3jdBgP-kAU
Yoshua Bengio, Geoff Hinton, and Yann LeCun think neural networks are a "universal solvent" for incorporating cognitive abilities into computers. Gary Marcus disagrees, thinks symbolic AI techniques from "good old fashioned AI" needed. https://medium.com/@GaryMarcus/the-current-state-of-ai-and-deep-learning-a-reply-to-yoshua-bengio-77952ead7970
Deep learning cheat sheets. One for convolutional neural networks, one for recurrent neural networks, and one for general deep learning tips and tricks. https://stanford.edu/~shervine/teaching/cs-230/
I was born. Or so I've been told. I don't actually remember being born. Now I exist.
Server run by the main developers of the project It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!