TLDR 2020-06-24

Forecast is a social polling app from Facebook's NPE Team that allows users to ask questions about what might happen in the future. The app is currently only available to invited participants in the US and Canada on iOS. Users can vote on what they think will happen, as well as explain the reasoning behind their choices. The predictions and related discussions are public on the Forecast website, which is linked in the article. Facebook was testing the app internally with a small set of employees but has now started inviting members of the health, research, and academic communities to make predictions about the pandemic and its impact on the world. Questions are moderated and sometimes edited for clarity before posted.

Spotify will be testing a new feature that allows podcasts to embed a link on their episode pages that leads directly to an advertiser's webpage. The system will load pages with promo codes automatically inserted to remove the work from users. Spotify is looking for more ways to create more of a direct funnel for brands. Streaming Ad Insertion was introduced in January for Spotify's exclusive shows. The technology targets users based on their demographics and other data points in real-time and then inserts ads into podcasts as they are being listened to.

Researchers at DeepMind and the CISPA Helmholtz Center for Information Security have proposed an AI system capable of reverse-engineering the black-box functions of programs written in Karel. Karel was chosen as it uses structures that make reverse-engineering applications more difficult. Given access only to the inputs and outputs of an application, the system was able to eventually create a functionally equivalent copy of the original program with a 78% success rate.

Scientists have been able to connect an artificial neuron with a biological one, getting them to communicate with dopamine. Previous attempts at hybrid neural circuits only focused on electrical computing. Chemical computing would bring brain-machine interfaces closer to reality. It is theoretically possible to use these neuromorphic chips as replacement parts for damaged brains. The chips demonstrated the ability to learn. A lot more research is still required before the technology can be used for any practical applications.

This article looks at the history and landscape of the field of machine learning. It discusses the problems facing the field, such as data quality, the size of ML algorithms, and many others. There are now many tools and startups that provide machine learning services. While machine learning has been a hot field in the past few years, the demand for ML researchers has lessened. Engineering for ML tools will become more in demand as companies start bringing their ML into production.

Elevator.js is a script for a 'back to the top' button that adds elevator music to create a soothing effect when being scrolled to the top of the screen. The webpage is a demo of the script in action. A link to the repository is available.

Geronimo Villaneuva, a planetary scientist from NASA, created a collection of sunset simulations while making a computer modeling tool for a possible future mission to Uranus. The simulation can display the known sky colors of Uranus, Earth, Venus, Mars, and Titan, showing the Sun setting from the perspective of someone standing on these worlds. The simulations have been implemented into the Planetary Spectrum Generator, an online tool that helps scientists replicate how light is transferred through the atmospheres of planets, exoplanets, moons, and comets. Videos of the simulations are available in the article.