News

Slow Thought: a manifesto

The only thing for certain is that everything changes. The rate of change increases. If you want to hang on, you better speed up.

Seeing the brain’s electrical activity

MIT researchers have come up with a new way to measure electrical activity in the brain. Their new light-sensitive protein can be embedded into neuron membranes, where it emits a fluorescent signal that indicates how much voltage a particular cell is experiencing. This could allow scientists to study how neurons behave, millisecond by millisecond, as the brain performs a particular function.

HotSwap: Bringing hot code reloading to Buck

Every day hundreds of Facebook engineers make thousands of code changes, each of which requires at least one, and usually many, iterations of the edit-compile-run development cycle. To speed up this process, we built and open-sourced Buck, a build tool designed from the ground up for fast iteration, allowing engineers to compile and run changes quickly.

How and why our experiments with virtual reality motion made us ill

Mobile Real-time Video Segmentation

We use big convolution kernels with large strides of four and above to detect object features on the high-resolution RGB input frame. Convolutions for layers with a small number of channels (as it is the case for the RGB input) are comparably cheap, so using big kernels here has almost no effect on the computational costs.

Can AI Ever Learn To Follow Its Gut?

Academics, economists, and AI researchers often undervalue the role of intuition in science. Here’s why they’re wrong.

Google-Landmarks: A New Dataset and Challenge for Landmark Recognition

A few examples of images from the Google-Landmarks dataset, including landmarks such as Big Ben, Sacre Coeur Basilica, the rock sculpture of Decebalus and the Megyeri Bridge, among others.

Learning by playing

Our new paper proposes a new learning paradigm called ‘Scheduled Auxiliary Control (SAC-X)’ which seeks to overcome the issue of exploration in control tasks. SAC-X is based on the idea that to learn complex tasks from scratch, an agent has to learn to explore and master a set of basic skills first. Just as a baby must develop coordination and balance before she crawls or walks—providing an agent with internal (auxiliary) goals corresponding to simple skills increases the chance it can understand and perform more complicated tasks.

Feeding Frenzy for AI Engineers Gets More Intense

Demand for software engineers with AI expertise continues to increase, while supply flattens

Computing With Random Pulses Promises to Simplify Circuitry and Save Power