Profile cover photo
Profile photo
Wayne Radinsky
18,601 followers -
Software Design Engineer
Software Design Engineer

18,601 followers
About
Posts

Post has attachment
One hour of imaginary celebrities.
Add a comment...

Post has attachment
"How China's WeChat is tackling fake news differently from Facebook." "While the Silicon Valley juggernaut has sought to distance itself from the fake news epidemic, Chinese tech firms are taking it head on. In China, there are no debates around the role that companies play in deciding what the public sees. The Great Firewall, which blocks thousands of sites like Facebook and The New York Times, might keep information from the outside world from filtering in, but domestically, the government expects tech giants to take responsibility."

"Here's how WeSeer works: articles that are shared between accounts, both on personal newsfeeds and via public accounts, create a path when they travel through the social network."

"To identify potential falsehoods, Qu's research group must delve deeper into the accounts themselves to assess how credible they are."
Add a comment...

Post has attachment
"The US Defense Advanced Research Projects Agency (DARPA) is looking to change the way AI works through a program it calls L2M, or Lifelong Learning Machines. The agency is looking for systems that learn continuously, adapt to new tasks, and know what to learn and when. 'We want the rigor of automation with the flexibility of the human,' says the program's director Hava T. Siegelmann. The US $65-million program has already chosen 16 groups for 4-year projects, but according to Siegelmann there is still opportunity to propose 12- or 18-month projects."

"If you raise the net in a basketball game by 30-centimeters, players will miss most of the time at first, but as they continue playing they'll learn to score at the new height. You don't have to pull them off the court and teach them the entire game over again."
Add a comment...

Post has attachment
The current population growth rate of 1.11% per year cannot be sustained for much longer.
Add a comment...

Post has attachment
An intuitive explanation of the Single Shot Detector MultiBox object detection technique. "A few years ago, by exploiting some of the leaps made possible in computer vision via CNNs, researchers developed R-CNNs to deal with the tasks of object detection, localization and classification. Broadly speaking, a R-CNN is a special type of CNN that is able to locate and detect objects in images: the output is generally a set of bounding boxes that closely match each of the detected objects, as well as a class output for each detected object.

"In the last few years, new architectures were created to address the bottlenecks of R-CNN and its successors, enabling real-time object detection. The most famous ones are YOLO (You Only Look Once) and SSD MultiBox (Single Shot Detector)."

"To better understand SSD, let's start by explaining where the name of this architecture comes from: Single Shot: this means that the tasks of object localization and classification are done in a single forward pass of the network, MultiBox: this is the name of a technique for bounding box regression developed by Szegedy et al, and Detector: the network is an object detector that also classifies those detected objects."
Add a comment...

Post has attachment
Ethereum now handles more transactions than all digital currencies combined. "Ethereum currently processes 52.3% of all cryptocurrency transactions, followed by bitcoin at 33%, while Bitcoin Cash, currently third in market cap, is handling only 1.8% as pictured above."

"More than half a million transactions were processed yesterday by ethereum, a level that is now becoming common for the platform, rising considerably since early last year when it stood at just 10,000."
Add a comment...

Post has attachment
The sweet spot for intermittent fasting. "Ted Naiman suggested that 'the sweet spot for intermittent fasting' occurs between 18 and 24 hours of fasting, since this is the time period that sees the greatest drop in insulin and increase in lipolysis  --  the breakdown of fat."
Add a comment...

Post has attachment
"Why Russia is sending robotic submarines to the Arctic." "The Arctic Ocean is estimated to hold billions of barrels of oil, and trillions of cubic feet of natural gas -- accounting for 16-26% of the Earth's undiscovered reserves."

"Enter Russia's Project Iceberg: an ambitious plan to use extreme technology for equally extreme conditions." "The centrepiece of Iceberg is the 182m-long (600 ft) Belgorod, the largest nuclear submarine ever built. The Belgorod will carry out underwater surveys and lay communication cables under the ice, but its main role will be to act as a mothership for a flotilla of smaller submarines."

"But perhaps the most ambitious part of Project Iceberg are the plans for the word's first underwater nuclear power plants to act as pitstops for the swarms of submarines that will be deployed."
Add a comment...

Post has attachment
An autonomous drone with AI made by NASA JPL engineers was raced against human drone pilot Ken Loo in an indoor obstacle course. The AI averaged 13.9 seconds per lap and Ken Loo averaged 11.1 seconds per lap. So the AI is not quite at human-level performance. The AI also had the benefit of having a map pre-loaded into its memory before the competition.
Add a comment...

Post has attachment
The original WaveNet model for speech generation generated one sound sample at a time and was not suitable for real-time production use for massive numbers of people. To make the production system, DeepMind came up with a system called Probability Density Distillation, where a parallel network was trained using the original WaveNet network as the 'teacher'. They say it has no significant difference in quality with the original WaveNet, can generate speech samples 20x faster than real-time, and is deployed online by Google Assistant, including serving multiple English and Japanese voices.

"This student network is a smaller dilated convolutional neural network, similar to the original WaveNet. But, crucially, generation of each sample does not depend on any of the previously generated samples, meaning we can generate the first and last word -- and everything in between -- at the same time

"During training, the student network starts off in a random state. It is fed random white noise as an input and is tasked with producing a continuous audio waveform as output. The generated waveform is then fed to the trained WaveNet model, which scores each sample, giving the student a signal to understand how far away it is from the teacher network's desired output. Over time, the student network can be tuned -- via backpropagation -- to learn what sounds it should produce. Put another way, both the teacher and the student output a probability distribution for the value of each audio sample, and the goal of the training is to minimise the KL divergence between the teacher's distribution and the student's distribution."

"The training method has parallels to the set-up for generative adversarial networks (GANs), with the student playing the role of generator and the teacher as the discriminator. However, unlike GANs, the student's aim is not to 'fool' the teacher but to cooperate and try to match the teacher's performance."
Add a comment...
Wait while more posts are being loaded