TLDR Daily Update 2019-11-11

Facebook erases whistleblower πŸ™ˆ, AI predicts lightning ⚑

Big Tech & Startups

Facebook says it will remove posts purporting to name the Ukraine whistleblower (1 minute read)

Facebook will remove any content that claims to name the US official whose whistleblower complaint started the impeachment inquiry against President Donald Trump. This content violates Facebook's policies around coordinating harm. The new policy affects all posts on Facebook, not just ads. An ad that purported to name the whistleblower was viewed several hundred thousand times before it was taken down. Trump has called for the whistleblower to be unmasked. The impeachment inquiry is looking into whether Trump solicited help from Ukraine President Volodymyr Zelensky to investigate former Vice President Joe Biden and his son Hunter.

Here’s how Amazon employees get health care through a new app β€” a glimpse of the future of medicine (4 minute read)

Amazon's new Amazon Care app is designed to help its Seattle-area employees get more convenient and affordable health care. Some companies are moving into primary care in order to tackle rising health care costs, hoping that early detection can prevent costly emergency visits. Amazon has partnered with a medical group called Oasis Medical for the project. Health information gained from the program won't be sold. Employees who are eligible for Amazon Care are able to consult with physicians directly in-app. If an employee requires in-person care, a practitioner will be dispatched, and an in-app map will show their location and estimated arrival time. The program has been received well by employees.
Science & Futuristic Technology

AI tech predicts time and place of lightning-strikes (1 minute read)

An AI system was able to predict the time and place of lightning strikes using standard weather station data. It was trained using data from four basic weather parameters: atmospheric pressure, air temperature, relative humidity, and wind speed. The data was gathered over a 10-year period in urban and mountainous regions and cross-referenced with recordings from lightning detection and location systems. Researchers were able to predict the location and time in which lightning would strike at around an 80 percent accuracy rate. Using AI, more accurate prediction systems that require fewer inputs can be developed. This system will be able to cover more areas as it doesn't require external data from radars or satellites.

Liquid-in-liquid printing method could put 3D-printed organs in reach (1 minute read)

Replicating organs such as gastric tracts, windpipes, and blood vessels is a major challenge as they can be difficult to construct without supporting scaffolding, which can be difficult or impossible to remove after construction. Previous attempts to create support structures with liquids have failed. Researchers from China have been able to create a stable membrane using hydrophilic polymers. These liquid structures are able to hold their shape for up to 10 days.
Programming, Design & Data Science

Solved by Flexbox (GitHub Website)

Flexbox gives CSS proper layout mechanisms in order to solve problems that were previously hard or impossible to solve with CSS alone. A showcase of examples is available. Each example details a problem and then shows how Flexbox fixes the issue. Flexbox is supported in every modern browser.

PyTorch at Tesla (11 minute video)

Tesla's Autopilot can control and navigate a car through a highway, and its Summon feature can move a vehicle from a parking spot directly to the driver. It doesn't use lidar or high-definition maps. Everything is based on computer vision and machine learning on the raw footage from the cameras on the vehicle. Most of Tesla's neural networks have a shared backbone as having an individual network for each task would be too resource-heavy. Each of these shared networks is called a Hydra-net. Multiple Hydra-nets can work together to make collaborative predictions. Training all these neural networks is an expensive task. It takes about 70,000 GPU hours to compile the neural network stack for Autopilot. Most of the training process can be automated, as long as the data sets are continually improved. Tesla's custom hardware allows them to train neural networks at a higher magnitude for a low cost.

Amazon’s roadmap for Alexa is scarier than anything Facebook or Twitter is doing (2 minute read)

Amazon has plans to move Alexa from passive to proactive interactions, which means that rather than waiting and responding to requests, it will start anticipating what the user might want. In order to do this, it will need to access and process a lot of personal user data. Amazon, Microsoft, and Google are trillion-dollar companies due to their ability to collect and use data. Eventually, Amazon will start focusing on building training databases with data from individuals or very small groups. By getting Alexa to start the conversation, Amazon may be able to get even more personalized data from its customers.

Byte sized news for busy techies

Byte sized news for busy techies

TLDR is a daily newsletter with links and TLDRs of the most interesting stories in tech πŸ“±, science πŸš€, and coding πŸ’»!

or subscribe with

Join 175,000 readers for one daily email