Robot Ecosystems

Robotic Mortality

ASADA--To change the topic a bit, I think that the ultimate issue is whether the robot recognizes its own mortality. A sense of one's own death is the real condition for the emergence of self. Unfortunately, the brains we're working with aren't up to the task. The autonomic nerve isn't developed yet, so it can't even tell where it hurts, or feels out of shape. Only when the robot has an awareness of its creation leading to its death, in short, when a sense of time is possible, will we finally be able to return to the analogy of the biological animal. To test this we will need a body that biodegrades, and this will bring us to wetware. Right now we have motors and gears, a body made of steel and plastic. If a gear is missing the robot doesn't feel "under the weather." To create a robot capable of replacing its own parts is completely within the realm of possibility. They could have robot hospitals where they go and request "a new arm, please, this one's no longer performing to spec." They need an autonomic nerve for even this, though.

Actually, this topic is related to the idea of robots eventually playing against human beings, because the robots will have to be able to feel pain for it to be a fair match. A human being can have a concussion heading a ball. We can't have the robots indifferent to their battle scars. (laughs) Yes, the robots have to have bodies that feel the bruises just like the humans. Precisely as in ASIMOV's Three Laws of Robotics,[*9] the robots need to learn to protect their own bodies.

SAKURA--In order to be able to monitor themselves, right? And this is an essential function?

ASADA--That's something that we haven't worked out yet.

SAKURA--Just monitoring them with language, it's hard to imagine how you would arrive at "death." I believe that death and procreation are two sides of one coin. Death is the reset button after reproduction. By the previous generation passing on, the next generation is provided for. Life is because death is.

ASADA--With robots, when their bodies become aged, it seems like you should be able to just pull out their brains, and put them in a new body, but in fact the whole point of what we've been saying is that it is through the body that the brain evolves, so, in fact, just replacing the head would be a moot exercise.

SAKURA--It would be a problem of aesthetics? (Laughs)

ASADA--When the media covers the RoboCup competitions, the single most common phrase that they use to describe it is, "robots equipped with artificial intelligence." I really wish that they'd knock it off. No matter how carefully you explain to them that it is not a matter of AI being here and bodies there, that in fact the intelligence emerges precisely because of those bodies, they run back to their office and print it as "robots equipped with artificial intelligence." If we tried taking their brains out and replacing their bodies, the value of who they are would be changed. It would be just like Jerry Lewis, in The Nutty Professor (1963). Change the bodies and everything else changes with it. The body changes the sense of values, and what is imaginable. Just changing the heads won't work, either. What we need are biodegradable bodies. Then we'll have to deal with reproduction and everything will get strange. That's where the real questions begin.

SAKURA--To put it another way, if you don't manage that aspect properly you'll never arrive at incalculable intelligence.

ASADA--We'll never get to wonder "what they're up to today?" Of course, there are a lot of limitations. Should we give them a default one-day time cycle, or let them infer their own time frame from their environment? A 12-hour period might be one day to a robot. Whatever the conclusion, if we don't at least see a concept of time emerge in them they'll never get a recognition of their own selves, far less that of others. Awareness of the "other" comes with senses of past-present-future. Living on an axis of continuum is what allows for an understanding of others' actions.

In soccer terms, we need to create robots capable of feinting actions. Feinting means a psychological grasp of the others' actions, that they can read beyond the surface of what is presented. You need to have a sense of self to fool others. You also have to be able to stand in their shoes, see what they're seeing, know who they are. Whether this is something which is genetically or experientially stipulated is an essential point.

SAKURA--That's interesting. It does implicate both nature and nurture. At Kyoto University Primate Research Institute they are testing the chimpanzee "Ai" for species differentiation competency. She understands the difference between photographs of humans and chimpanzees. But when you show her a photograph of herself, she always responds that it's a "human"! (laughs)

ASADA--According to recent developmental cognitive psychology both elements do fundamentally exist. There are ways to enable a connection to inherent knowledge. Experience, environment and other factors work together in certain patterns. It is something that I understand theoretically, but my job is to build robots. So until we really know the specifics of how to plant the seeds of this recognition, it's tough for me to just "believe."

Backleft rightNext