TLDR Daily Update 2020-07-28

Manipulating dreams 💭, Mars helicopter 🚁, how GPT-3 works 💻

Big Tech & Startups

Airbus' self-flying plane just completed successful taxi, take-off, and landing tests, opening the door for fully autonomous flight (4 minute read)

Airbus just completed flight testing for its Autonomous Taxi, Take-off, and Landing project. Over 500 flights were conducted with the new Airbus A350-1000 XWB, with the pilots only lining up the plane with the runway before letting the plane take flight on its own. The program relies heavily on the onboard cameras that are used by pilots to help guide the plane while on the ground. Airbus' success means that fully autonomous flights might soon be possible as autopilot already handles most functions while airborne.

Google will keep employees working remotely until July 2021 (1 minute read)

Google will keep its 200,000 full-time and contract employees working remotely until at least July 2021. Sundar Pichai, CEO of Alphabet, made the announcement last week following a meeting with top company executives. Google has reopened some offices for a limited number of employees, but working in the offices is optional. Other companies such as Facebook, Twitter, and Square have announced plans to keep employees working from home indefinitely.
Science & Futuristic Technology

MIT Dream Research Interacts Directly With an Individual’s Dreaming Brain and Manipulates the Content (4 minute read)

Targeted Dream Incubation is a method to record dream reports and guide dreams towards particular themes. Dormio is a sleep-tracking device that can alter dreams by tracking dream states and delivering audio cues based on incoming physiological data. Directing dreams requires precise timing. Guided dream content can be used to complete tasks such as creative story writing. Dream studies have shown that dream incubation is tied to performance benefits in tests of creativity.

NASA’s Ingenuity—the First Ever Off-World Helicopter—Is Set for a ‘Wright Brothers Moment’ on Mars (5 minute read)

NASA's Perseverance rover will be carrying a four-pound helicopter in its belly. Named Ingenuity, it will attempt up to five powered flights on Mars. The first flight will replicate test flights previously conducted on Earth. After that, Ingenuity will start testing its limits, eventually flying up to 150 feet away on its final test. Each trip will last about 90 seconds from takeoff to landing, which is the maximum time available due to Ingenuity's battery capacity. Mars' atmosphere is less than 1 percent the density of Earth's atmosphere, so Ingenuity's blades have to spin 10 times faster than helicopters on Earth to create an upward lift. It will take a whole Martian day to recharge between flights.
Programming, Design & Data Science

Realtime Face Movement Tracking (GitHub Repo)

This repository contains code for basic face movement tracking that can convert face movement into keyboard commands. Two demo games are available.

How GPT3 Works - Visualizations and Animations (5 minute read)

GPT-3 was estimated to cost 355 GPU years and around $4.6m to train. 300 billion tokens of text were used to train the model. It generates one token at a time. The model has 175 billion parameters which it uses to calculate which token to generate at each run. It converts each word into a list of numbers representing the word, computes a prediction, and then converts the resulting vector into a word. Animations that explain the structure of GPT-3 and deep learning models are available in the article.

Japanese construction giant to build massive dam almost entirely with robots (2 minute read)

The Obayashi Corporation is building a giant dam in the south-east corner of Japan's main island. Almost every stage of the construction involves some sort of automation technology. The robots, while automated, are directly overseen by human workers for safety reasons. Productivity on the site has increased by about 10 percent, with the company aiming to eventually cut building time by 30 percent.

Disney, Microsoft, Nintendo, and 50 more hit by massive source code leak (3 minute read)

Source code from more than 50 high-profile companies was leaked onto a public GitLab repository. The code was obtained by scanning third-party sources and misconfigured DevOps applications. Some of the code includes credentials. Leaked source code can have serious impacts as it gives cybercriminals all the information they need to exploit the software. A list of the affected companies is linked in the article.

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 📱, science 🚀, and coding 💻!

Join 500,000 readers for one daily email
515 W 18th St. Unit 621
New York, NY, 10011
515 W 18th St. Unit 621
New York, NY, 10011