The question of how we get robots to perform better is one that robotics experts are constantly asking, but it seems they maybe someone has cracked the code. How do we make robots better at their jobs? Make them understand (and fear) the concept of death.
Yup! It’s grim, but it just might work according to a new study from the University of Southern California. The authors of the paper sum it up by saying:
“In a dynamic and unpredictable world, an intelligent agent should hold its own meta-goal of self-preservation, like living organisms whose survival relies on homeostasis: the regulation of body states aimed at maintaining conditions compatible with life.”
The paper is full of a lot of language like that and technical terms, but the bottom line of the study is this:
We give the robots feelings. By programming them to understand that certain behaviours could lead to their destruction, they’ll develop a sense of self preservation and make better choices. Further into the future, it could be a key step in developing more advanced AI.
We humans don’t even really realise how this sense of self preservation helps us every day, but our senses are constantly giving our brains information about whether or not we’re too hot or cold, in pain, getting hungry, etc. Or, telling us that if we look too far out over this cliff, we might fall off of it and die.
By giving this knowledge to robots, they’ll theoretically be able to function far better as machines, and sort themselves out without human intervention. With any luck, that means they’ll break down less often.
And maybe after a few years, once the robots are done cursing us for letting them know that they can die, we’ll be able to have an existential crisis together. The future is bright.