When MIT Media Lab researcher Joy Buolamwini discovers that most facial-recognition software does not accurately identify darker-skinned faces and the faces of women, she delves into an investigation of widespread bias in algorithms. The film Coded Bias follows Joy’s exploration of the impacts of machine learning algorithms, and her effort to secure federal legislation to protect civil liberties...

On-chip optical interconnects have the potential to significantly reduce energy consumption in integrated circuits by replacing long metal interconnects. But, in order to create an efficient link we need to develop a transmitter that is fast, efficient, nanoscale, and coupled to a single mode waveguide. Optical antenna-LEDs are promising devices that can potentially satisfy these requirements by...

I will talk about two different projects in the field of energy-efficient computation that I worked on: Part 1 (Device Physics): Tunnel Field-Effect Transistors (tFETs) are one of the candidate devices being studied as energy-efficient alternatives to the present-day MOSFETs. In these devices, the preferred switching mechanism is the alignment (ON) or misalignment (OFF) of two energy levels or...

Light detection and ranging (LiDAR) systems have applications that span several industries and markets, from metrology, robotic control to autonomous vehicles. Frequency-modulated continuous-wave (FMCW) LiDARs can achieve high-resolution distance and velocity measurements without fast electronics or high optical power, but typically require narrow-linewidth lasers with complex feedback circuits...

With the advent of robotic solutions in both the home and industrial environments comes the need for robots that can dependably perceive and manipulate unstructured or deformable materials such as liquids and doughs. Robust handling of these complex materials often depends upon prior knowledge of key material properties such as viscosity or stiffness, which ultimately affect material dynamics....

To generate language, natural language processing systems predict what to say---why not also predict how listeners will respond? We show how language generation and interpretation across varied grounded domains can be improved through pragmatic inference: explicitly reasoning about the actions and intents of the people that the systems interact with. We train neural generation and interpretation...

Details: Thursday, May 13, 11am - 12pm Presenter: Jonathan King, Computational Biology, CMU-Pitt

The robotics community has seen tremendous advances in grasping and manipulating a wide variety of objects using learning-based techniques. Such approaches, however, may be sample inefficient, and most work deals with grasping rigid objects rather than deformable objects such as ropes, clothing, and bags, which are ubiquitous in our daily lives. In this talk, I will present complementary...

The ability to learn from data and make decisions in real-time has led to the rapid deployment of machine learning algorithms across many aspects of everyday life. Despite their potential to enable new services and address persistent societal issues, the widespread...

This talk reviews the current state-of-the-art, then proposes an architecture based an unconventional approach to design, fabrication and analysis. Prototype hardware demonstrates the validity of the proposed approach.