Bits & Bytes
Bionic AI.
- As a side-note to the unveiling of its latest iPhone, Apple last week introduced the world to the Neural Engine, an AI module inside of its new A11 Bionic chip. The chip will perform up to 600 billion operations per second to support functions like FaceID, Siri and image recognition on the new iPhone X. Related: In response to Apple’s FaceID announcement, Senator Al Franken asked Apple for privacy guarantees around Face ID data. It’s not clear to me that this data will be collected centrally, but if so, I’d certainly be interested in how the company responds.
More robots, more jobs.
- An interesting piece in the New York Times looks at the impact on Amazon’s warehouse labor force as they’ve introduced more robots. Long story short: they’ve reallocated workers to other tasks. The story is a bit fluffy but the pictures and videos of the warehouse are interesting.
Reinforce the cloud.
- I’m bullish on the use of machine learning for datacenter optimization, especially at cloud scale. If you share my interest in this use case, you should check out this paper on the application of reinforcement learning for optimizing cloud provisioning on AWS.
New concepts.
- Big news from Bonsai this week, as the company publishes results demonstrating a 45x improvement in deep learning training times using their “concept network” approach. They also published a paper on concept network reinforcement learning and got a nice write up in Forbes.
Only one shot.
- A couple of interesting research papers this week: Chelsea Finn’s One-Shot Visual Imitation Learning via Meta-Learning presents a new technique that allows robots to acquire new skills from just a single demonstration. 3D Face Reconstruction from a Single Image, does what it says, with pretty impressive results. The demo site lets you upload and run the algorithm on your own photo, and the project site includes links to the paper as well as the code.
