What is meme-based supervision?

Post Reply
User avatar
Site Admin
Posts: 30
Joined: Sat Mar 10, 2018 3:25 pm
Location: Fort Gatlin, Francis Langhorne Dade Territory

What is meme-based supervision?

Post by colSteve » Sun May 27, 2018 8:57 am

Meme-based supervision is neither autonomy nor remote control. It is something in between. It saves robots from having to be autonomous, and saves people from having to see everything remotely and control robots every second.

Memes are small behaviors which a robot can combine into larger activities. This is similar to how letters can be combined into words and sentences. The idea that a robot should be supervised based on memes comes from several ideas:

1) People don't mind supervising robots. In fact they enjoy it like a video game.
2) People do mind standing up and moving, so the primary role of a robot is to do physical work.
3) Live control like with a joystick, is stressful and difficult.
4) Robots don't need complex thought, or to make extended plans of behavior. They just need to execute each baby step in the complex behaviors.

Many people are obsessed with the romantic notion of self-driving cars. But most cars already contain a person, and people don't mind driving. People actually enjoy driving. The reason people need a robot lawnmower, is not to steer itself around the lawn. It is to save a person having to stand up and go outside. People don't mind telling a robot lawnmower what to do, from the comfort of their computer screens.

Memes can also be shared by robots of different types. So a behavior developed for one robot, and be downloaded from the web and used by another robot. Memes like walking can be robot-independent. A meme that is robot specific, is called a "native meme." A robot translates a common meme into its own native memes.

For example, a robot could have a move-left-wheel meme, and a move-right-wheel meme. These are native memes, because not all robots have wheels. Or even a turn-on-motor meme. Not all robots will have motors with switches. A standard go-forward meme will be translated into a native turn-on-motor meme. Any robot can understand the go-forward meme. But each robot translates the meme into different personal sub-tasks. Depending on whether it has wheels, or legs, or whatever.

Sequences of memes can be combined into activities. A meme sends an output, such as going forward. Another meme collects an input, such as detect wall. When it detects the wall, it executes the stop meme. Different robots can use different hardware to execute the same meme, and do the same activity.

This saves a person from having to control what the robot is doing, every single second. And it saves the robot from having to understand the more complex task, the person is aiming toward.

Post Reply