Digital computers have transformed every aspect of our world and are enabling new machines that possess intelligence like and beyond human. From automatically classifying objects in images to enabling self-driving cars to predicting human diseases, digital intelligence is bound to deliver far-reaching influence in every aspect of our lives. The evolution from fewer than a million isolated personal computers to today’s billions of Internet-connected smart devices puts forth a massive amount of data and computing power. With proper techniques to harness the data and power that have never been available before, digital intelligence is not only possible but also practical and cost-effective.
Digital intelligence consists of the use of digital computers to perform intelligent tasks, from automatically recognizing objects in pictures to manuevering an autonomous vehicle. Enabling digital intelligence involves the tasks of creating and implementing intelligence in machines. These tasks naturally translate into a set of problems on how to perceive information, how to process and represent information, how to intelligently learn from the information, and how to model and apply what has been learned. Answers to these questions incorporate fundamental theories in math and statistics as well as real-life applications on sensors and computers, all of which must engage cohesively within a system to leapfrog the next innovations in fields such as healthcare, automotive, finance, and personal entertainment.
Sitting at the interface between hardware and software, the field of computer engineering is in a unique position to tackle the interdisciplinary nature of the emerging areas of digital intelligence. Computer engineers are equally capable of building hardware devices and programming software algorithms, which constitute the two most important skills in creating new machines with increasingly intelligent capabilities. They are not only inventing and implementing intelligent machines that are fast and smart, but also witnessing first-hand how their products transform our society.
Scholars in the 2018 CATALYST Academy will explore digital intelligence through the perspective of computer engineering by exploring the process of creating and implementing intelligence in machines. In particular, we will study how to combine state-of-the-art hardware and software to perceive, process, and learn from real-world information and assemble a practical hardware-software system with demonstratable intelligence.
This week-long design experience includes three guided laboratory sessions to familiarize us with concepts in hardware, software, and machine intelligence. Based on knowledge acquired from the laboratory session, we will work in groups to build an end-to-end prototype of an intelligent machine. Here is an overview of the design activities:
We will start with learning the concepts of hardware because hardware is the foundation required to build and run our intelligent machines. We will experiment with basic analog and digital circuits and have the opportunity to build digital elements from a simple logic gate to a more complex arithmetic unit.
We will then learn to write software to program the behavior of hardware. Using an Arduino-based robotics platform, you will write software programs to control various exploratory behaviors of a robot through actuators and enable the robot to perceive its surrounding environment through sensors and react appropriately.
We will learn how to bring intelligence into machines by teaching a computer to recognize handwritten numerical digits. We will learn and implement the popular K-Nearest-Neighbor machine learning algorithm to decipher different handwritten digits using a distance-based measure.
We will combine hardware, software, and machine intelligence to build a cohesive Internet-connected intelligent system. Depending on the project selected, we will leverage our digit recognition system to wirelessly control the movement of the robot and take advantage of various sensors to enable autonomous behaviors such as rapid target finding and adaptive cruise control.