top of page
What can we learn about robots from goldfish?


One of the most exciting design challenges of the near future will be shaping and redefining our relationship with AI and robotics, and it’s something I want to be a part of. It’s also something I think needs a lot of work, particularly in regard to robotics, because there is a large gap between what we perceive robots can do, and what they can actually do. When we inevitably discover these limitations it's a pretty frustrating experience.
To design better robots, I wanted to understand more about them, so I decided to try and build one. This project is a work in progress, but it’s my exploration into how we can create far more satisfying interactions.


“Robots that look like people, but are no smarter than simple animals, can be viewed as deceptive and unethical. We should build human-like robots only when their intelligence matches their appearance.”

This quote from Professor Alan Winfield was the driver behind a lot of the decisions in this project, but another inspiration were goldfish - there's no personal interaction with a goldfish, it just gets on with things and swims around, but for some reason people get a lot of pleasure from them. How does this compare to the all-singing, all-dancing home robots on the market?
I think it says something on how doing very little could be much more satisfying than promising to do everything, so in this respect the goal of the robot is simple - to autonomously navigate spaces without bumping into things and without falling off things.
Robot 5_2x.png


How should a robot with this level of intelligence behave, and can something this simple still be engaging?
For the most part, the code uses randomisation to automate the 'decisions’ the robot makes, for example which direction to head in and for how long. It also has a number of built in 'instincts’ that kick in when it's about to fall off a table or crash into a wall, but other than that it just wonders about as it chooses and should be smart enough to get out of most situations (though it does occasionally needs help)
Robot 4_2x.png


There’s a universal design language for robots - white plastic, lights and a face. However, the problem with faces is that whilst they greatly increase empathy, they also lead us to believe the robot is far more intelligent than it actually is. My aim was to find a new design language that doesn’t rely on zoomorphism.
I’m currently working on producing the body for the robot, and have been experimenting with CNC’d moulds, paper pulp, and expanding foam.
Robot 1_2x.png
You can find some of my earlier experiments below...
bottom of page