Dr. Robertson began his work in Artificial Intelligence in graduate school at Purdue University. For his Master’s project he developed an analog computer model of a simple biological nervous system, the cockroach stretch receptor. In his Ph.D. research, he developed techniques (including early neural networks) for recognizing materials and objects in multispectral satellite imagery.
At Bell Laboratories, he designed software agents that were embedded in electromechanical billing equipment throughout the US, to recognize fault patterns and direct repair work.
At TASC (The Analytic Sciences Corporation) he led a division that contributed to a revolution in large-scale map making, using digital image processing and pattern recognition to replace film and microscopes.
As Director of Engineering at Lockheed Martin Advanced Simulation, he led the development of distributed, virtual simulations that allowed humans to interact with AI-controlled forces on the virtual battlefield.
As Vice President and General Manager at Atlantic Aerospace, he led the development of systems based on recognizing patterns in signals and images, including equipment installed on nuclear submarines and video tracking of pitched baseballs to assess MLB umpire performance.
At BAE Systems, he led the Deep Green project, a $40M project to automate military planning using AI, sponsored by the Defense Advanced Research Projects agency.
While at Thinking Teams, his current organization, he has supplemented his Organization Development work by learning computer languages currently used in AI development (Python, Swift, TensorFlow, Keras), and becoming certified in multiple aspects of deep neural networks (dense, convolutional, and recurrent networks; hyper-parameter tuning). He has also spoken in the US and internationally on the impact of AI on the workplace and society as a whole.
Below are some examples from this work.
Organization Development in the Time of AI
A virtual presentation given June 2020 for Organization Development Network Oregon.
Emergency Benefits and Risks of Artificial Intelligence
Presentation at the TIEMS 2019 Annual Conference in Goyang, Korea:
Video Baseball Pitch Tracking
Visual pattern recognition from digital video can be used to track baseball pitches, and together with video measurement of batter stance, show whether a pitch is a ball or a strike. Our system was used to enhance television broadcasts in the US and internationally, and it was used by Major League Baseball in the US to evaluate umpires. At first the umpire union objected, and Dr. Robertson was called as an expert witness to testify before lawyers in a labor dispute. Eventually, use of the system was accepted by the umpires.
Multispectral Image Partitioning
The LANDSAT 1 satellite sent digital images from space, showing the reflectance of the earth in four spectral or ‘color’ bands. These images arrived as streams of numbers, four reflectance values for each 30 meter patch of earth. Statistical techniques had been developed to classify materials on the earth’s surface based on distinctive patterns in the four numbers, however these methods were not able to take advantage of shape and texture, hugely important to human photo interpretation. Dr. Robertson developed a technique to ‘partition’ the digital image stream into areas of uniform reflectance or texture, so that shape and texture could be used in the automatic classification process.
Modeling a Cockroach Stretch Receptor
The nervous systems of humans and cockroaches share a common architectural element – neurons. Understanding neurons is important to understanding human brains and intelligence. Cockroach neurons are easier to study, because they a big, and an individual neurons can be more easily instrumented and recorded. To better understand what goes on inside a neuron, we built a circuit on an analog computer that was able to produce a series of pulses in response to a stimulus pattern, that was the same series of pulse as produced by a cockroach neuron when its stretch receptor is stretched and released with the same stimulus pattern.