About This Document (IRD-34355).
This Microsoft Word document created by the La Trobe University Inclusive Resources Development team. This document has been created as a transcript of the supplied audio/video and contains only narrative/spoken content. No audio description has been included.
While every care has been taken to accurately transcribe the original material there may still be errors contained in this conversion.
Project Number.
34355.
Student Name.
Graduate Research School.
Subject Code.
IRDTREQ - IRD Transcription Request.
Article Title.
Murray Neuzerling: How to avoid robots.
Publication.
3MT.
Publisher.
La Trobe University.
Date of publication.
2015.
Copyright Notice.
Copyright Regulations 1969.
WARNING.
This material has been copied and communicated to you by or on behalf of La Trobe University pursuant to Part VA of the Copyright Act 1968 (the Act). The material in this communication may be subject to copyright under the Act. Any further copying or communication of this material by you may be the subject of copyright protection under the Act.
Do not remove this notice.
Start Transcript.
Murray Neuzerling
How to avoid robots. We know how to avoid robots, we just walk around them. What we don't know is what goes on in our heads when we do that sort of navigation. There's a process there involving spatial relations, and relations are what I study. Not just spatial, but temporal, even familiar. You see, we can look at relations like familial relations in a mathematical way using concepts like and, or, not, of. So the sister of my father is my aunt. My brother or sister is my sibling. Those are mathematical concepts.
And the same reasoning holds for spatial relations. One team in Germany has used relations to study the problem of a human and a robot walking past each other in a corridor. The human didn't even know the robot would be there, but that's fine, humans know how to avoid robots. But for the robot it's a very difficult encounter. One reason for this might be that machines think quantitatively, a distance of 3 metres, and a bearing of 280 degrees. Whereas humans think qualitatively. To my left. And to my left is a relation. So we can use all those mathematical concepts, and, or, not, of, to solve problems like this one. But mathematicians, we like to solve entire classes of problems at once. So I'm looking at all models involving only a handful of relations, but in my work the relations can be anything; spatial, temporal, familial, it's all the same to the mathematical model.
Now in general when you take a problem and you model it with relations, it's hard. And when mathematicians say "hard" we're very precise about what we mean by that. It's the exact same kind of hard as solving a Sudoku puzzle, it involves a bit of guess and check. So I'm going through all of these problems, and classifying them as hard or easy, so that researchers know what they're up against. But hard problems are hard. If you guess a square in a Sudoku puzzle and you later find out that you made a mistake, you have to undo all of your work and start again.
The difference between and easy problem and a hard one can be the difference between a fraction of a second and a hundred years. So my work will help a researcher choose the right kind of model for their robot, because it has to do more than just "work", it has to work in time for the robot to avoid us.
Thank you.
End Transcript.