News

A Google Brain engineer’s guide to entering AI

Note that this guide was written in November 2018 to complement an in-depth conversation on the 80,000 Hours Podcast with Catherine Olsson and Daniel Ziegler on how to transition from computer science and software engineering in general into ML engineering, with a focus on alignment and safety. If you like this guide, we’d strongly encourage you to check out the podcast episode where we discuss some of the instructions here, and other relevant advice. Technical AI safety is a multifaceted area of research, with many sub-questions in areas such as reward learning, robustness, and interpretability.

Introspected REST: An Alternative to REST and GraphQL

In this manifesto, we will give a specific definition of what REST is, according to Roy, and see the majority of APIs and API specs (JSONAPI, HAL etc) fail to follow this model. We will see what problems a RESTful API brings and why API designers have been constantly avoiding using it but instead come up with half-way solutions or retreat to alternative models like RPC-over-HTTP or, lately, GraphQL. Then, we will propose a new model, Introspected REST, that solves the issues that REST creates and allows the design of progressively evolvable APIs, in a much simpler way than conventional REST.

How to Develop Convolutional Neural Network Models for Time Series Forecasting

Orange Skin, Thick Bones, and All the Other Ways the Human Body Could Change on Mars

Optimal Shard Placement in a Petabyte Scale Elasticsearch Cluster

The number of shards on each node, and tries to balance the number of shards per node evenly across the clusterThe high and low disk watermarks. Elasticsearch considers the available disk space on a node before deciding whether to allocate new shards to that node or to actively relocate shards away from that node. A nodes that has reached the low watermark (i.e 80% disk used) is not allowed receive any more shards.

Embracing failures and cutting infrastructure costs: Spot instances in Kubernetes

GraphQL: A success story for PayPal Checkout

At PayPal, we recently introduced GraphQL to our technology stack. At PayPal, GraphQL has been a complete game changer to the way we think about data, fetch data and build applications. This blog post takes a close look at PayPal Checkout and explains our journey from REST to Batch REST to GraphQL and lessons learned along the way.

Cross shard transactions at 10 million requests per second

Dropbox stores petabytes of metadata to support user-facing features and to power our production infrastructure. The primary system we use to store this metadata is named Edgestore and is described in a previous blog post, (Re)Introducing Edgestore. In simple terms, Edgestore is a service and abstraction over thousands of MySQL nodes that provides users with strongly consistent, transactional reads and writes at low latency.

Learning Concepts with Energy Functions

We’ve developed an energy-based model that can quickly learn to identify and generate instances of concepts, such as near, above, between, closest, and furthest, expressed as sets of 2d points. Our model learns these concepts after only five demonstrations.

Mystery particle spotted? Discovery would require physics so weird that nobody has even thought of it