TLDR 2019-02-15

Amazon cancels HQ2 🏢, Facebook's record fine 💰, AI writing ✎

Big Tech & Startups

Facebook could reportedly face multibillion-dollar FTC fine over privacy violations (1 minute read)

The FTC is in discussions with Facebook over the exact amount of a multi-billion dollar fine as a result of privacy violations by Facebook. If an agreement is not reached over the amount of the fine, the FTC is likely to bring Facebook to court. The largest fine that the FTC has previously issued was to Google in 2012, at $22.5 million, as the company tracked users of Apple’s Safari browser after explicitly saying it wouldn’t. Privacy advocates argue that a multi-million dollar fine may not be effective in persuading Facebook to change its behavior, so the FTC is pushing for a much larger deterrent in addition to the threat of a court trial.

Amazon Pulls Out of Planned New York City Headquarters (5 minute read)

Amazon has canceled its plans to build a new headquarters in New York City after heavy resistance by lawmakers, progressive activists, and union leaders. Groups opposing the deal contended that the government should not provide the tech giant with the up to $3 billion dollars in incentives that it had promised. Amazon decided that dealing with the heavy resistance would not be beneficial to the company in the long term, however, it has decided that it will still continue growth in its existing branches in the city regardless of government subsidies.
Science & Futuristic Technology

JP Morgan is rolling out the first US bank-backed cryptocurrency to transform payments business (3 minute read)

The first cryptocurrency created by a US bank is JP Morgan’s JPM Coin. The bank has created the cryptocurrency in order to create instant wire transactions, instant securities transactions, and to replace actual dollars held in subsidiaries across the world. Using blockchain technology will allow the bank to quickly transfer value anywhere in the world in a secure manner. A JPM Coin will be worth exactly a dollar, and initially will only be available for use by its larger clients.

New AI fake text generator may be too dangerous to release, say creators (3 minute read)

OpenAI, a research company backed by Elon Musk, has produced a new AI fake text generator which is so impressive that the company has declined to release a full report on its capabilities for fear of misuse. The AI, GPT2, is able to generate plausible text that matches the style and subject of its inputs, without the common quirks of other previous AI systems such as forgetting what it is writing about midway through paragraphs or creating mangled sentences. For example, it is currently able to generate an infinite number of positive (or negative) reviews for products. There are examples of its outputs in the article, and a video demonstrating a version of the AI.
Programming, Design & Data Science

Sloth (GitHub Repo)

Sloth is a MacOS application that allows users to see all open files and sockets used by all running processes on the system. It creates a graphical user interface over the command line tool, lsof, and adds functionality to the output, such as being able to filter through the results or kill specific processes. Using the tool, users can easily inspect and control what their apps are doing on their system.

Introducing draft pull requests (1 minute read)

GitHub has introduced the ability to create draft pull requests, allowing developers to create pull requests which may not be aimed for deployment, but to create discussion and continued collaboration. Users can now opt to create a draft pull request by using the dropdown menu when selecting ‘Create pull request’. The draft pull request is styled differently, and users can choose to change the state of the draft to a full pull request by selecting the ‘Ready for review’ button near the bottom of the pull request.
Miscellaneous is face-generating AI at its creepiest (2 minute read)

Nvidia has used a new Generative Adversarial Network (GAN) to create a neural network that can generate an unlimited number of high-quality, realistic faces. A GAN works by having one neural network produce an image, while another neural network tries to see whether the image is real or fake. If the image is deemed to be ‘real’, then the network passes the image, otherwise the generator restarts and tries again. The generator at is a result of both training using the GAN and modifications to the AI architecture by the research team.

After a brief rebellion, the EU link tax and upload filter will move to a final vote (2 minute read)

The final text for the EU's Copyright Directive has been completed and will be voted on sometime in March or April. There had been disputes over the wording of the directive, specifically in Articles 11 and 13, which define the controversial ‘link tax’ and ‘upload filter’. Under the directive, content distributors such as YouTube are required to proactively scan user content for copyright infringement, and content aggregators may be liable for charges and fees if they reprint content from the original articles.
TLDR is a daily newsletter with links and TLDRs of the most interesting stories in startups 🚀, tech 📱, and programming 💻!
Join 1,250,000 readers for one daily email