The AI ethics and safety communities are at a crossroads that will determine whether they work together in the production of safer intelligent systems, or continue down separate conflicted paths.

I opportunistically recorded a video of a delivery robot that almost immediately got into trouble. What follows is an introduction to the design elements of delivery robots, followed by how these elements produced and recovered from a series of unfortunate events.

As someone that turned down an offer to work on cashierless checkout four years ago, I was keen to learn how far the technology has come. What I found was a fascinating case study in designing the human environment for computer vision systems, and I also went home with several items that I didn't pay for...

In "2001: a Space Odyssey," the astronaut falls into an monolith floating in space. Just before the radio cuts transmission to earth, the astronaut gasps, "my god, it's full of stars." This sense of wonder and impending revelation is the view I brought to my PhD studies in artifical intelligence. Surely, I thought, we were working together to unwind our own little piece of the great mysteries of the universe. I was quickly disabused of my higher notions. In falling into the mysteries of intelligence, I came to understand that the intelligence we were collectively working to build resembeled the mysterious power of the Space Odyssey monolith in more than one way, only as I fell into my research I muttered, "my god, it's full of bugs."

Video has long been a trusted source of ground truth for world events. People know images can be manipulated, but video manipulation is rare. This has changed with the advent of manipulating videos with neural networks. Automatic video manipulation tools, like those that swapped Nicolas Cage's face onto Amy Adams, are now fast, realistic, and cheap.
Defending the public record requires methods for detecting these fakes. But is the deepfake detection game winnable?

Jesse Hostetler, Dan Garmat, and I recently hacked on DeepMind's implementation of a neural network for the Arcade Learning Environment. The above video shows the agent learning to play a game called Breakout, in which the objective is to break bricks at the top of the screen using a ball …

I am just wrapping up at the IEEE Symposium on Visual Languages and Human-Centric Computing. My paper appears in the proceedings and I presented on Monday. This has been my first Human Computer Interaction (HCI) conference, so I have been exploring the differences and similarities from my AI and FOSS …

For our forthcoming paper at the IEEE Symposium on Visual Languages and Human-Centric Computing I built a visualization for testing and debugging optimization simulators and algorithms.

The visualization is built on d3js and runs in the web browser. If you visit mdpvis.github.io you can interact with a wildfire …

I attended a talk at Open Source Bridge titled "Open Source is Not Enough: The Importance of Algorithm Transparency". The TLDR of the presentation is that algorithms shape your view of the world which can lead to socially bad outcomes. Think filter bubbles, search manipulation, etc.

The presenter's prescription was …