Intelligence and Autonomy
On 4 July 1997 the Pathfinder probe landed on Mars and its six-wheeled, multi-jointed, robotic rover, the Sojourner, began analyzing the atomic composition of rocks. This was one of the first of a series of missions planned to be “lighter, faster, and cheaper,” according to NASA administrator Daniel Goldin. Previous rover projects called for a brawny device as large as a garden tractor. The Mars Rover was as small as a skateboard, but it was brainy.
The team that built the Mars rover programmed it to have a certain level of autonomy. They could issue a command on the level of “seek rock” and the rover would find a rock to analyze. This autonomy was necessary because Mars is far away. The distance between Earth and Mars varies from as little as 56 million kilometers (35 million miles) to as much as 378 million kilometers (235 million miles). It takes radio waves or laser signals anywhere from 3 to 21 minutes to carry a command to Mars, depending on the distance. When a robot executes the command there is an equal interval of time before the controllers on Earth can see the results. The delay between giving the command and seeing the results is unavoidable because radio waves and light travel at a high but limited speed. No signal can go faster than the speed of light. A robot on Mars needs some autonomy to detect a hazard and respond appropriately before its Earth-bound controllers can tell it what to do.
Mars missions have never occurred at the time of closest approach to Earth. Aerospace engineers must choose the probe trajectory that takes the least fuel. They launch the probe when the planets are approaching each other. The probe arrives when the planets are receding again. When the probe finally lands, the distance between Earth and Mars is much larger than the minimum. Typically the roundtrip light-speed delay is 15 minutes.
Suppose that the Earth-bound controllers see an interesting rock a hundred meters away and send their rover toward it at a slow crawl, just fast enough to arrive in 20 minutes. What happens if, 80 meters from the starting point, there is a crevice in the ground, a crevice the controllers couldn’t see when they commanded the rover to move? Suppose that the rover’s television camera first sees the crevice when it is still ten meters away and sends the picture back to Earth. As soon as the controllers see the crevice they signal the rover to hit the brakes. By the time their command arrives, the rover has lain broken at the bottom of the crevice for 13 minutes.
Obviously, the rover needs sufficient visual processing capability on board to let it recognize the hazard and apply the brakes without waiting for intervention from Earth. It must have intelligence and some autonomy.
We have sent a robotic probe to fly by and photograph Pluto and some other Kuiper-Belt objects. The roundtrip light-speed time was some 10 hours and 40 minutes. The flight controllers pre-selected the scenes to photograph at closest approach when resolution was highest. There was only one opportunity, and there wasn’t time to tell the probe to take a second close look at anything interesting that appeared in the photos taken at closest approach. If we ever send a robotic rover to Pluto it will need considerable artificial intelligence and autonomy to send interesting reports to the folks back home.
The nearest star is four light years from Earth, and extra-solar planets are farther still. To begin exploring beyond the solar system we must develop artificial intelligence that will give a robot nearly complete autonomy. However, the robot must not have enough autonomy to decide that its mission is too dangerous to fulfill!
Extra-solar exploration robots will also have to be very lightweight to arrive in a reasonable time, because there are limits on the fuel and money we can dedicate to any given project, and the cost of propulsion goes up faster than the mass of the instruments. The lightest weight of all would be a constructor robot built on a molecular scale, capable of building many larger robots from local materials found near the distant exploration site. The larger robots would be of various types and would have various programs for the purpose of exploring and reporting.
Such a system of robots imitates biology by storing compactly immense catalogs of information. The information consists of sets of instructions, some for building various kinds of structures, and other sets of instructions for fulfilling a purpose or mission. There has to be some initial structure capable of finding and using available resources to manufacture the other structures, or at least to make the first robot factory. When robots make other robots at the destination, their activity will be analogous to biological reproduction.
Designers can hardly do better than living organisms at storing structural information on the molecular level. Let’s review how living cells specify their structure and manufacture new cells, cells either like themselves or adapted for other purposes.
Biological Structure and Reproduction
A cell nucleus contains deoxyribonucleic acid (DNA). DNA is a long-chain molecule composed of certain atomic structures called nucleic acids. These are arranged like a twisted ladder, a spiral staircase, or like the mating teeth on a long, winding zipper. The sequence of nucleic acids is the genetic code.
The structure is stable most of the time. But when the cell is reproducing itself, an enzyme partly unzips the DNA in the middle of a strand. It acts like the kind of zipper that one can open in the middle by moving two zipper sliders in opposite directions. Another enzyme reproduces one of the unzipped strands of DNA. It makes a new mating half for that strand. When the mating half is finished and floating away free inside the cell, the first enzyme zips up the original DNA again. Then other enzymes make a matching half to the floating half strand, and zip the two halves up. At the end of the process there are two zipped zippers instead of one.
The new zipper or DNA strand moves away from the original DNA strand to the far end of the cell. The cell is then ready to pinch itself in the middle and divide in two, with one DNA strand going to each half. This is the normal method of cell reproduction.
The two strands of zipped-up DNA are identical, because only certain pairs of nucleic acids interlock precisely to make the teeth of the zipper. In this way a cell makes a precise copy of its own genetic information before it divides and reproduces.
Once a cell reproduces it has to grow. Other enzymes read the relevant sections of the DNA code in the nucleus and provide instructions for manufacturing all the enzymes the cell needs. These enzymes produce all the proteins of the cell and all the molecules needed to digest food.
Enzymes all by themselves are fragile. Some enzymes must be kept on ice or they will denature. How then can they operate in living organisms at body temperature? Part of the requirement for life is the ability of the structure to maintain itself. When part of the maintenance system goes down, the whole organism collapses and dies. At the moment of death it still has most of the structure necessary for life, if only the broken part could be repaired quickly. However, within minutes thermal agitation disrupts the delicate coiling or folding of proteins that makes some of them active as enzymes. The enzymes denature and cannot catalyze the necessary chemical reactions. The structures they were maintaining then break down, and death becomes irreversible. It takes large systems of enzymes to maintain life.
Engineering is still far from miniaturizing its information storage and retrieval systems to the molecular level, or manufacturing such tiny robots as the enzymes in living cells.
The team that built the Mars rover programmed it to have a certain level of autonomy. They could issue a command on the level of “seek rock” and the rover would find a rock to analyze. This autonomy was necessary because Mars is far away. The distance between Earth and Mars varies from as little as 56 million kilometers (35 million miles) to as much as 378 million kilometers (235 million miles). It takes radio waves or laser signals anywhere from 3 to 21 minutes to carry a command to Mars, depending on the distance. When a robot executes the command there is an equal interval of time before the controllers on Earth can see the results. The delay between giving the command and seeing the results is unavoidable because radio waves and light travel at a high but limited speed. No signal can go faster than the speed of light. A robot on Mars needs some autonomy to detect a hazard and respond appropriately before its Earth-bound controllers can tell it what to do.
Mars missions have never occurred at the time of closest approach to Earth. Aerospace engineers must choose the probe trajectory that takes the least fuel. They launch the probe when the planets are approaching each other. The probe arrives when the planets are receding again. When the probe finally lands, the distance between Earth and Mars is much larger than the minimum. Typically the roundtrip light-speed delay is 15 minutes.
Suppose that the Earth-bound controllers see an interesting rock a hundred meters away and send their rover toward it at a slow crawl, just fast enough to arrive in 20 minutes. What happens if, 80 meters from the starting point, there is a crevice in the ground, a crevice the controllers couldn’t see when they commanded the rover to move? Suppose that the rover’s television camera first sees the crevice when it is still ten meters away and sends the picture back to Earth. As soon as the controllers see the crevice they signal the rover to hit the brakes. By the time their command arrives, the rover has lain broken at the bottom of the crevice for 13 minutes.
Obviously, the rover needs sufficient visual processing capability on board to let it recognize the hazard and apply the brakes without waiting for intervention from Earth. It must have intelligence and some autonomy.
We have sent a robotic probe to fly by and photograph Pluto and some other Kuiper-Belt objects. The roundtrip light-speed time was some 10 hours and 40 minutes. The flight controllers pre-selected the scenes to photograph at closest approach when resolution was highest. There was only one opportunity, and there wasn’t time to tell the probe to take a second close look at anything interesting that appeared in the photos taken at closest approach. If we ever send a robotic rover to Pluto it will need considerable artificial intelligence and autonomy to send interesting reports to the folks back home.
The nearest star is four light years from Earth, and extra-solar planets are farther still. To begin exploring beyond the solar system we must develop artificial intelligence that will give a robot nearly complete autonomy. However, the robot must not have enough autonomy to decide that its mission is too dangerous to fulfill!
Extra-solar exploration robots will also have to be very lightweight to arrive in a reasonable time, because there are limits on the fuel and money we can dedicate to any given project, and the cost of propulsion goes up faster than the mass of the instruments. The lightest weight of all would be a constructor robot built on a molecular scale, capable of building many larger robots from local materials found near the distant exploration site. The larger robots would be of various types and would have various programs for the purpose of exploring and reporting.
Such a system of robots imitates biology by storing compactly immense catalogs of information. The information consists of sets of instructions, some for building various kinds of structures, and other sets of instructions for fulfilling a purpose or mission. There has to be some initial structure capable of finding and using available resources to manufacture the other structures, or at least to make the first robot factory. When robots make other robots at the destination, their activity will be analogous to biological reproduction.
Designers can hardly do better than living organisms at storing structural information on the molecular level. Let’s review how living cells specify their structure and manufacture new cells, cells either like themselves or adapted for other purposes.
Biological Structure and Reproduction
A cell nucleus contains deoxyribonucleic acid (DNA). DNA is a long-chain molecule composed of certain atomic structures called nucleic acids. These are arranged like a twisted ladder, a spiral staircase, or like the mating teeth on a long, winding zipper. The sequence of nucleic acids is the genetic code.
The structure is stable most of the time. But when the cell is reproducing itself, an enzyme partly unzips the DNA in the middle of a strand. It acts like the kind of zipper that one can open in the middle by moving two zipper sliders in opposite directions. Another enzyme reproduces one of the unzipped strands of DNA. It makes a new mating half for that strand. When the mating half is finished and floating away free inside the cell, the first enzyme zips up the original DNA again. Then other enzymes make a matching half to the floating half strand, and zip the two halves up. At the end of the process there are two zipped zippers instead of one.
The new zipper or DNA strand moves away from the original DNA strand to the far end of the cell. The cell is then ready to pinch itself in the middle and divide in two, with one DNA strand going to each half. This is the normal method of cell reproduction.
The two strands of zipped-up DNA are identical, because only certain pairs of nucleic acids interlock precisely to make the teeth of the zipper. In this way a cell makes a precise copy of its own genetic information before it divides and reproduces.
Once a cell reproduces it has to grow. Other enzymes read the relevant sections of the DNA code in the nucleus and provide instructions for manufacturing all the enzymes the cell needs. These enzymes produce all the proteins of the cell and all the molecules needed to digest food.
Enzymes all by themselves are fragile. Some enzymes must be kept on ice or they will denature. How then can they operate in living organisms at body temperature? Part of the requirement for life is the ability of the structure to maintain itself. When part of the maintenance system goes down, the whole organism collapses and dies. At the moment of death it still has most of the structure necessary for life, if only the broken part could be repaired quickly. However, within minutes thermal agitation disrupts the delicate coiling or folding of proteins that makes some of them active as enzymes. The enzymes denature and cannot catalyze the necessary chemical reactions. The structures they were maintaining then break down, and death becomes irreversible. It takes large systems of enzymes to maintain life.
Engineering is still far from miniaturizing its information storage and retrieval systems to the molecular level, or manufacturing such tiny robots as the enzymes in living cells.