Q: What is the most important thing that's happened in computing in the past 10 years?
A: For me, it’s the success of deep learning. I did some work on hybrid systems trying to integrate symbol-oriented approaches (expert systems, logic programming) and sub-symbolic methods (mainly neural networks, with low-level distributed knowledge representation schemes). Conceptually, this was a nice combination, taking advantage of the best of the two approaches: the knowledge bases of expert systems could be augmented by information extracted from actual data collections, and the inner workings of neural networks could be made more transparent. In practice, it didn’t work very well, however. But over time methods improved, datasets became much bigger, and processing power also increased substantially. This combination led to a breakthrough, where the neural networks used in deep learning are much larger now, and have an internal structure that corresponds to meaningful features of the respective knowledge-based models of a domain.
Q: By the end of your career, where do you think computer science will have taken us? What are you working on that might contribute to that?
A: I’m fascinated by the interaction between humans and computers, and right now, I’m looking into what I call “interaction spaces.” This spans the physical space in which we interact with computational devices, and how that has evolved from sitting in front of a desktop computer, to touch-based interaction with mobile devices, and gesture-based interaction with devices like the Microsoft Kinect and Leap Motion. For me, however, there is also a conceptual component to these interaction spaces, ranging from an interaction language (such as a set of symbols, expressed as written letters, spoken words, or gestures arranged into meaningful sequences) to interaction as the exchange of information between the participants. Until now, this interaction has been enabled by the capability of humans to work around the limitations of computers: We type commands, select menu items, click on buttons, and so on. Computers are getting better, and we can interact with them through natural language, as with Siri and Cortana, although it’s still limited. It won’t be too long, in my view, until we can engage with computational devices in more meaningful interaction sequences, be it in the form of a conversational exchange of information about a limited topic (not necessarily deep philosophical discussions), or collaborating with a robot in a physical activity. This requires the ability to construct models that specify the behavior of the participants, and the anticipation of the likely next actions by others. For a conversation in natural language, the interaction space encompasses the vocabulary, syntax, and usage of the language, together with the domain knowledge and context of the conversation topic. In real-world interaction with a robot, the interaction space model captures the capabilities of the participants, the task within the domain and the context, and how the task can be accomplished by breaking it down into interleaving sequences of actions according to those capabilities.
Q: Who is your favorite historical figure? Why?
A: Albert Einstein. We were both born in the same city, and he managed to achieve a lot under challenging circumstances.
Q: If you weren't working in the computer science field, what would you be doing instead?
A: If I had followed tradition, as the first-born son I’d have taken over my parents’ farm. So I’d be spending more time on tractors and combines, but I’m quite sure that I still would have gotten involved with computers in some way.
Q: What is your favorite type of music?
A: Having just returned from a trip to Jamaica, right now it’s reggae.