What if computers could engage with people fluently in plain English, researching and providing fresh information, teaching them about the world? New models trained on large-scale data, arranged for reusability and explainability will make it happen. That's what I've been working on lately: interactive, natural-language dialogue systems that do not only sound natural, but, most importantly, speak the truth.
The deep question I have been asking is about models that describe what makes people and computers intelligent. Why and when are humans and large language models effective communicators and decision-makers? I have adopted a computational view of the human mind, and have used big-data methods and NLP to study cognitive processes. Recently, my focus has been on AI.
I joined Google Research in New York City after a career in academia. I was as a tenured professor at Penn State, after Carnegie Mellon University and the University of Informatics (PhD). I have published 100+ papers in computer science and cognitive science, was funded by over $1.8M in grants from the National Science Foundation and others, and mentored a group of graduate students and post-docs who, without exception, do very well professionally. My computational work aims to bring real value to my organization's information mission with measurable improvements to dialog systems and conversational interfaces. In the long term, though, I also address the scientific puzzle of how the mind works. My most prominent work concerned alignment and information distribution in conversations: I examined what makes conversational successful, and a model of human memory can explain how people adapt their individual language choices to one another. To do so, my group used shallow and deep natural-language processing tools, statistical modeling, cognitive architectures and neural networks to better understand the cognitive and social processes involved in language-based interaction and decision-making. Recently, deep learning has become an essential tool in our quest to understand and augment the human mind. And yes, I'm happy to be hands-on: I started and led Aquamacs Emacs.
Before my research career, I worked briefly as a freelance radio journalist in the German ARD network. I am a licensed, instrument-rated private and commercial pilot in the USA and in the U.K., in gliders, single engine, multi-engine (land) airplanes. I have helped manage some great flying clubs in Edinburgh, Pittsburgh, and Central Pennsylvania.
I am a U.S. and German citizen, and I live in New York City.
Contact: David Reitter, Google. E-mail