Chapter 15 3.5 Using real-world feedback to communicate

An overly concentrated communication load is not the only trouble with the central brain.The maintenance of central memory is also a headache.Shared memory must be strictly updated in real time and accurately - and many companies feel deeply about this.For a robot, the daunting task of the control center is to compile or update a "model of the outside world," a theory, or a representation based on its perception—where are the walls, how far are the doors, and, what else? Forget it, watch out for the stairs there. What should the brain center do if the information fed back by different sensors conflicts with each other?The eyes say that something is coming, but the ears say that the thing is going away.Who should the brain trust?The logical thing to do is to try to find out the truth.Then, the control center adjusts the dispute and re-corrects the signal to make it consistent.In a robot with a non-contained structure, the computing resources of the central brain are mostly consumed in drawing a coordinated image of the external world based on feedback signals from different perspectives.Each part of the system has a different interpretation of the massive data returned by the camera and infrared sensors, so each has a very different perception of the outside world.In this case, the brain can never coordinate everything, so nothing is always achieved.

It's so hard to coordinate a central view of the world, and Brooks finds it much easier to use the real world as a model of itself: "It's a great idea, because the world is really a pretty good model of itself." Without a model of central coercion, there is no one to undertake the work of mediating disputes, which themselves do not need mediation.Instead, different signals produce different behaviors.At the network level of containment control, behavior is selected through inhibition, delay, activation, etc. Essentially, for the robot (or for the insect, as Brooks prefers to put it), there is no reflection of the outside world.There is no central memory, no central command, no central being.Everything is distributed. "Communication through the external world avoids the problem of tuning the visual system based on data from the tentacles," Brooks wrote.The external world itself becomes the "central" controller; the environment without reflection becomes the reflection itself.This saves a lot of computing work. "Within such an organization," Brooks said, "intelligent behavior can be generated with only a small amount of computation."

Without a central agency, individuals of all stripes rise to the top or fall silent.The mechanism Brooks proposes can be understood in this way—in his words, "individuals in the brain communicate with the outside world to compete for the robot's physical resources." Notice. Those with quick brains found that Brooks's scheme was an excellent portrait of a market economy: the individuals participating in the market did not communicate with each other, but observed the effect of others' actions on the common market (not the actions themselves).From hundreds of vendors I've never met, I get information on the price of fresh eggs.The information told me (included in many other information): "A dozen eggs is cheaper than a pair of leather shoes, but more expensive than a two-minute domestic long-distance call." This information, together with many other price information, has guided thousands of chicken farms The business behavior of owners, shoemakers, and investment bankers tells them where to invest their money and energy.

Brooks' model, not only has revolutionized the field of artificial intelligence, it is also a true model for how complex organisms of any kind can function.We see containment structures and network hierarchies in all types of living systems.Brooks summed up five experiences in designing mobile robots, which are expressed as follows: Incremental builds - allowing complexity to develop itself, rather than being rigidly implanted Tight coupling of sensors and actuators - low-level reflection, not high-level thinking Module-agnostic hierarchy - splitting the system into self-evolving subunits

Decentralized control - no centralized planning Sparse communication - results of observing the outside world instead of relying on wires to transmit messages When Brooks compressed the hulking and headstrong monster of a robot into a humble, feather-light reptile, he gained new insight from that attempt at miniaturization.In the past, to make a robot "smarter," it required more computer components, which made it bulkier.The heavier it is, the bigger the drive motor needs to be.The larger the motor, the larger the battery pack required to power it.The larger the battery pack, the larger the structure of the mobile battery pack, thus falling into a vicious circle.This vicious cycle makes the proportion of the robot's brain and body tend to be smaller and smaller.

But if this cycle is reversed, it becomes a virtuous circle.The smaller the computer components, the smaller the motor can be, the smaller the battery, the smaller the frame, and the greater the structural strength corresponding to its size.This also makes the brain of the small mobile robot proportionally larger in proportion to the body, although the absolute size of the brain is still small.Most of Brooks' mobile bugs weigh less than 10 pounds.Genghis, assembled from model car components, weighs only 3.6 pounds.Brooks wants to have robotic bugs 1mm long (the size of a pencil tip) within three years.He simply called it "robot flea".

Brooks advocates not only sending such a robot to Mars, but also allowing it to infiltrate every corner of human society.Brooks says he wants to bring as much artificial life as possible into real life, not as much organic life as possible in artificial life.He wants to populate the world with cheap, tiny, ubiquitous semi-thinking machine creatures.He gave an example of smart doors.In your house, for an additional cost of $10, you can put a computer chip on a door, and it will know you are going out, or hear a message from another door that you are coming, and it will tell you that you are coming. Will notify lights when you're gone, blah blah blah.If every door in a building could talk to each other, it could help with climate control, and it could help control traffic flow.If we popularized these little invaders, injecting quick, cheap, out-of-control little wits into all the other establishments that now seem cold and boring to us, we could have millions of sensitive little ones.They serve us and are constantly learning how to serve us better.

Brooks, who was touched, predicted such a beautiful picture of the future: Our society is full of artificial creatures, living in harmony with us and interdependent, forming a new type of symbiotic relationship.Much of it goes unnoticed by us and is taken for granted.The way they solve problems is designed to be the way of insects - many firewood is powerful, many people are powerful, and individual units are insignificant.They will outnumber us as much as insects do in nature.In fact, the robots in Brooks’ eyes don’t have to serve us tea and water like R2D2 in “Star Wars”, but only need to be self-contained and assimilate with everything in places where we can’t see.

A student in the Mobile Robotics Lab built an inexpensive robot the size of a rabbit.It watches where we are in the room and constantly adjusts your stereo for optimal sound as you move around.Brooks also has an idea for a tiny robot that lives somewhere in our living room or under the sofa.It hangs around like a collector's hobby machine, waiting to vacuum up when you're not home.You don't realize the existence of the snail girl until you come home and find the floors are shiny and new.There is also a robot crawler that crawls out from the corner when the TV is turned off and secretly sucks the dust on the TV.

Everyone fantasizes about having programmable pets. "The biggest difference between a car and a horse is that you don't have to take care of the car every day, but you have to take care of the horse every day," said Keith Hansen, a popular technology evangelist. function." "We are passionate about making artificial beings," Brooks wrote in a 1985 article.He defines an artificial being as a creation that can survive in real-world environments for weeks or months without human assistance, and perform some useful work. "Our mobile robot bugs are such creations. Switch on the power, they will blend into the outside world, interact with it, and seek to achieve various goals. Other robots are very different. Program or plan to accomplish a particular task.” Brooks insists that he does not create toy environments (that is, simple, easy environments) for his being, as most robot designers do."We insist on building complete systems that can exist in the real world, so that we don't fool ourselves and avoid hard problems," he said.

Up to now, natural science has not been able to solve a difficult problem, that is, how to establish a pure consciousness.If Brooks is right, that goal may never be achieved.Instead, consciousness will grow out of the stupid body.Almost all the lessons learned from the mobile robotics lab tell us that in a real world that does not tolerate mistakes, there is no way to achieve consciousness without the body. “Thinking is action, and action is thinking,” said Heinz von Forster, an initiator of the cybernetic movement in the 1950s. “Without motion there is no life.”