Chapter 7
Defense Advanced Research Projects Agency Development Center, Nevada
June 6, 2005, 0855 Pacific Standard Time
“So Charles, you managed to persuade DARPA that Porcupine plus would work.”
“It wasn't that difficult. DARPA was already receptive to the concepts behind our proposal, and they have funded several similar projects. By the way, Porcupine has a new name. Its now called the Mobile Autonomous Area Denial System.
“MAADS, an unfortunate acronym. I am sure we can think of something more warm and cuddly. What persuaded DARPA to fund MAADS?”
”I made the presentation and covered all of our main points, but what swayed them was the fully autonomous aspect, plus, a lot of what we need, we have already developed for Porcupine.
“Richard was there, and made a compelling argument that robots need to be autonomous. Richard, why don't you summarize your arguments for Alan? I'm sure he would find them interesting.”
The badly dressed, overweight man, who had been leaning against the wall, gave a slight smile that was hard to see through his thick beard.
“I am at my best when faced with a room full of expectant faces, but I think I can recall the argument I made. It was along the lines of - a robot is an entity capable of initiating complex behavior in order to achieve a goal. If a person controls the behavior, then you do not have a robot. What you have is a slave automaton. You must be able to give the robot an objective, and then let it determine the behavior required to achieve the objective. This is not to say you cannot intervene to set and change constraints on its behavior. What it does mean is you cannot direct its behavior. Anything claiming to be a robot that does not act autonomously is not a robot.”
Alan said, “So Richard, you told DARPA, all those robotic systems, they are funding, are not in fact robots because they are not fully autonomous.”
“Someone needed to tell them.”
“Did DARPA bring up that robots never seem to realize their promise?”
“No, they didn’t, but I can explain the reason for those failures.” Richard continued without waiting to see if either of the other two indicated they wanted to hear his reasoning. “Too much development effort in robotics is expended on solving supposed problems that are in fact existing solutions constructed to optimize human capabilities.
“If you want to teach a robot to cook, don’t start in a kitchen. A kitchen is a solution to cooking, designed to optimize human capabilities in preparing food, not the problem you have to solve.
“Getting robots to copy what people do, makes the problem a lot more difficult because people have very hard to replicate skills. What you need to do is go back to the root problem, deconstruct it, figure out the simplest way to solve it, and then design a robot around that solution. If you take humans out of the food preparation process, the solution would look nothing like a kitchen.
“We have kitchens because we cannot change people, or we are limited in the extent we can change people. We have no such restrictions with robots. We can change a robot as often and as much as we like.”
Richard obviously warming to his subject continued, “Many years of effort have gone into trying to get robots, and other artificial intelligence systems, to acquire progressively more complex behavior by learning. Those efforts have been a total failure. I can tell you, getting robots to learn, completely misses the point. What the artificial intelligence community is trying to do is replicate the solution that humans use. If you analyze the problem of 'how do I get a robot to exhibit complex behavior resulting in predictable outcomes', you would in all likelihood come up with a very different solution.”
Charles stopped Richard lecturing on one of his favorite subjects. “You need to leave, if you are going to get to Vegas by eleven a.m.”
Charles turned to Alan. “Richard is meeting this morning with one of our software vendors. A couple of their senior people are in Las Vegas for a convention. He should be back sometime this afternoon.”
When Richard had left, Charles wondered what kind of impression he had made on the Englishman.
Alan said. “Interesting guy! I think I will like him.”
Charles thought if Alan did like Richard he would be in a minority. Richard had a real talent for alienating people. Even though Charles had known him for years, he still found him hard to take at times.
“Will two weeks here in Nevada be enough time for what you need to do?”
“Probably not, but I'll come back every couple of months or so. My job is to make sure everyone settles in, and the right lines of communication are in place. After all, this a joint project and the UK are picking up part of the bill.”
Charles knew 'satisfying one master’ was hard enough, and didn’t relish the prospect of two sets of requirements, one from the US side and the other from the UK.
“Why is the UK paying part of the bill, and providing one of its best research scientists, not to mention some valuable intellectual property, for a project that is 100 percent US based?”
“Since the Second World War the UK’s overriding defense imperative has been to maintain the Trans-Atlantic Alliance. Maybe, you don’t remember Suez, but we do. In defense matters, we need you a whole lot more than you need us, and the UK government will jump at any chance to work with you on next generation weapons development. If we can bring something to the party you need, but do not have, then so much the better.”
Alan added, “We are under a lot more pressure than you are to stop using landmines, and our military insist they must have a workable replacement. Consequently, area denial is a research priority for us.”
“Alan, my main concern is we don’t get conflicting requirements from the US and UK sides. It will help me a lot if you can make sure the UK is onboard with whatever requirements get to me.”
“That's the other reason I’m here. I’ll spend most of next week at DARPA making sure the British and the Americans are on the same page. I’m sure you know, they offered me the Deputy Director position under you, but I didn’t want to pull my kids out of school, and move them to the middle of a desert in a strange country, so I declined the offer. But rest assured, I’ll do everything I can to ensure the politics don’t get in the way of you getting your job done.”
“I understand about not wanting to uproot your family or be away from them for any period. I’m married too, and intend to spend every weekend I can with my beautiful wife, but I am not looking forward to the weekend commute back to Portland. If there is anything you need me to do, then let me know.”
“There is one thing. Try to ensure Fiona doesn't overwork herself. She is smart and extremely conscientious, but she feels, because she said the birds would work, then she is responsible for every aspect of ensuring they do. Make sure she relaxes every once in a while, and doesn't take things too much to heart, especially if they go wrong.”
“I'll keep an eye on her.”
Charles glanced at his watch. “We need to get moving. Otherwise, we will be late for our first MAADS team meeting.”
The two men chatted amiably as they left the office. Each carried a laptop computer in a shoulder bag and a thick stack of stapled meeting notes. They turned left at the end of the white fluorescent-lit corridor, and entered a conference room where ten people were already seated around a large table.
Charles recognized everyone seated around the table, including the two new faces. Fiona, he had met, for the first time, the previous evening, and the soldier in uniform who had walked into his office at eight thirty that morning and introduced himself as 'Sergeant Jackson reporting for duty'. Charles at the time, was busy organizing his materials for the meeting and had been rather curt. He told the soldier he was sharing an office with Fiona, a scientist from the UK and he should attend meetings and try to learn as much as possible about the project.
Bringing a serving soldier onto the project team had not been his idea. It was one of those things that sounded good in a management meeting, but in practice could cause significant problems, by introducing someone who was unfamiliar with the culture of a technology development project. Charles resolved to make the time to talk to Jackson, and explain his role in more detail.
Charles walked to the head of the table, and stood in front of a large whiteboard. He sensed the mood of subdued excitement that always occurs with new projects. The hard reality of solving problems within the schedule would soon temper that mood.
“Ladies and gentlemen, welcome to the MAADS project. Some of you are new to the team, and I intend to describe the project's history for your benefit. Those of you, who were on the Porcupine project, will find some important changes with our new sponsor and project requirements, and we will have to rethink several important aspects of the system.
“This is now a joint USA UK project. We can do the full introductions later, but this is Alan Jones, my counterpart from the UK. He is here for two weeks, helping us setup the new project and coordinate requirements. A couple of people from the UK have joined our development team, including Dr. Fiona Lachlan. She is a biomechanical engineer specializing in bird flight. They all have full security clearances, and you should treat them no differently than the American staff.
“I need to introduce one other person. This is Sergeant Jackson, a weapons specialist in a US Army Special Forces unit. He is here to advise us on integrating a weapon into MAADS, and to give us a combat soldier's perspective.”
The medium-height man in his early thirties wearing a US Army uniform nodded to the other meeting participants. He had propped the cane he used to walk, against his chair. His air of causal indifference accurately reflected his attitude toward his own presence in the room. He didn’t want to be here, but while he was, he had decided to play the role of a by-the-book soldier.
Charles connected the overhead projector to his laptop, and started his presentation.
“Project Porcupine is now called the Mobile Autonomous Area Denial System, MAADS for short.”
A collective groan passed around the room.
“DARPA, the Defense Advanced Research Projects Agency, and the UK’s Ministry of Defense are jointly funding MAADS. Alan and Fiona both work for the Defense Research Agency. I won’t go into the details of DRA’s role, but you can think of it as a consultancy specializing in defense and weapons research.
“OK, let's start with our project objective. DARPA wants us to produce a weapon system capable of denying an area to an enemy force that operates autonomously for an extended period, and is highly selective in its targeting. The other requirements are listed in your handout.”
“How big an area?”
“As big as possible, at least a square kilometer.”
A short heavyset man in his forties reading through the handout asked. “What does 'guaranteed nonlethal self-destruction' mean?”
“It means the system, and its components, need to be capable of destroying themselves without killing bystanders.”
“What about 'conducive to rapid large-scale production'?”
“Sammy, why don't I run through my presentation? It should answer a lot of your questions.”
The man smiled and held up a hand to signify his agreement. The black lines under his fingernails were a sure sign of someone who worked with machines.
Charles showed the first of his slides. “Mobile means it can move around its area of operation, both on the ground and in the air. It doesn't mean the entire system is capable of moving. At least one major component will be static.
“Autonomous means it performs its functions without a human operator controlling it, and can continue to do so for an extended period. We and others, are continually pushing the limits of what can be performed autonomously. MAADS autonomy includes finding and targeting hostile forces, extends to refueling, and could potentially include other functions like rearming. I've had several discussions on these topics with DARPA, and frankly, I've had to rein in some of DARPA's expectations.
“Area denial means it can stop an enemy force occupying or crossing its area of operation. Traditionally, landmines are used for this function, but landmines have a very serious drawback. They are indiscriminately lethal, killing the bad guys, the good guys, and anyone else unlucky enough to be close, when the mine detonates.
“Many mines are left around, killing and injuring people for years after they are laid. This has resulted in a movement to totally ban them. All the mines used by the US and UK armed forces either are recoverable or can be destroyed in situ. However, this has done little to reduce the pressure from the people who want a complete ban.
“An important objective for MAADS is to be a selective alternative to landmines for area denial.
“And our overall objective is to produce several working MAADS systems as a proof-of-concept. By a proof-of-concept, I mean the system demonstrates the feasibility and viability of a fully autonomous weapon capable of area denial.
”Fully autonomous doesn't mean a 'deploy and forget weapon'. We need to closely monitor what MAADS does when it's deployed. The Army needs to exercise command and control over the system. This means we have to provide the capacity to override decisions arrived at autonomously.
“From our perspective, we need to see how MAADS performs, when it's deployed, and learn how we can improve it.
“Safety is another important issue. It's going to be a while before people are sufficiently comfortable with the safety record of autonomous weapons that they will allow them to operate without a human capable of overriding the robot's behavior.
“By the way, I'd prefer us to avoid using the word 'robot'. It has unfortunate connotations from science fiction books and movies. Try to talk about autonomous units and autonomous behavior.”
Charles took a drink of water before continuing. “MAADS will be composed of a number of semi-autonomous, but interdependent units, that work together to achieve the system's overall objectives. While it is not a specific requirement, multiple MAADS systems could cooperate and interoperate to achieve an objective beyond the capabilities of a single system.”
Chapter 8
June 6, 2005, 0855 Pacific Standard Time
“So Charles, you managed to persuade DARPA that Porcupine plus would work.”
“It wasn't that difficult. DARPA was already receptive to the concepts behind our proposal, and they have funded several similar projects. By the way, Porcupine has a new name. Its now called the Mobile Autonomous Area Denial System.
“MAADS, an unfortunate acronym. I am sure we can think of something more warm and cuddly. What persuaded DARPA to fund MAADS?”
”I made the presentation and covered all of our main points, but what swayed them was the fully autonomous aspect, plus, a lot of what we need, we have already developed for Porcupine.
“Richard was there, and made a compelling argument that robots need to be autonomous. Richard, why don't you summarize your arguments for Alan? I'm sure he would find them interesting.”
The badly dressed, overweight man, who had been leaning against the wall, gave a slight smile that was hard to see through his thick beard.
“I am at my best when faced with a room full of expectant faces, but I think I can recall the argument I made. It was along the lines of - a robot is an entity capable of initiating complex behavior in order to achieve a goal. If a person controls the behavior, then you do not have a robot. What you have is a slave automaton. You must be able to give the robot an objective, and then let it determine the behavior required to achieve the objective. This is not to say you cannot intervene to set and change constraints on its behavior. What it does mean is you cannot direct its behavior. Anything claiming to be a robot that does not act autonomously is not a robot.”
Alan said, “So Richard, you told DARPA, all those robotic systems, they are funding, are not in fact robots because they are not fully autonomous.”
“Someone needed to tell them.”
“Did DARPA bring up that robots never seem to realize their promise?”
“No, they didn’t, but I can explain the reason for those failures.” Richard continued without waiting to see if either of the other two indicated they wanted to hear his reasoning. “Too much development effort in robotics is expended on solving supposed problems that are in fact existing solutions constructed to optimize human capabilities.
“If you want to teach a robot to cook, don’t start in a kitchen. A kitchen is a solution to cooking, designed to optimize human capabilities in preparing food, not the problem you have to solve.
“Getting robots to copy what people do, makes the problem a lot more difficult because people have very hard to replicate skills. What you need to do is go back to the root problem, deconstruct it, figure out the simplest way to solve it, and then design a robot around that solution. If you take humans out of the food preparation process, the solution would look nothing like a kitchen.
“We have kitchens because we cannot change people, or we are limited in the extent we can change people. We have no such restrictions with robots. We can change a robot as often and as much as we like.”
Richard obviously warming to his subject continued, “Many years of effort have gone into trying to get robots, and other artificial intelligence systems, to acquire progressively more complex behavior by learning. Those efforts have been a total failure. I can tell you, getting robots to learn, completely misses the point. What the artificial intelligence community is trying to do is replicate the solution that humans use. If you analyze the problem of 'how do I get a robot to exhibit complex behavior resulting in predictable outcomes', you would in all likelihood come up with a very different solution.”
Charles stopped Richard lecturing on one of his favorite subjects. “You need to leave, if you are going to get to Vegas by eleven a.m.”
Charles turned to Alan. “Richard is meeting this morning with one of our software vendors. A couple of their senior people are in Las Vegas for a convention. He should be back sometime this afternoon.”
When Richard had left, Charles wondered what kind of impression he had made on the Englishman.
Alan said. “Interesting guy! I think I will like him.”
Charles thought if Alan did like Richard he would be in a minority. Richard had a real talent for alienating people. Even though Charles had known him for years, he still found him hard to take at times.
“Will two weeks here in Nevada be enough time for what you need to do?”
“Probably not, but I'll come back every couple of months or so. My job is to make sure everyone settles in, and the right lines of communication are in place. After all, this a joint project and the UK are picking up part of the bill.”
Charles knew 'satisfying one master’ was hard enough, and didn’t relish the prospect of two sets of requirements, one from the US side and the other from the UK.
“Why is the UK paying part of the bill, and providing one of its best research scientists, not to mention some valuable intellectual property, for a project that is 100 percent US based?”
“Since the Second World War the UK’s overriding defense imperative has been to maintain the Trans-Atlantic Alliance. Maybe, you don’t remember Suez, but we do. In defense matters, we need you a whole lot more than you need us, and the UK government will jump at any chance to work with you on next generation weapons development. If we can bring something to the party you need, but do not have, then so much the better.”
Alan added, “We are under a lot more pressure than you are to stop using landmines, and our military insist they must have a workable replacement. Consequently, area denial is a research priority for us.”
“Alan, my main concern is we don’t get conflicting requirements from the US and UK sides. It will help me a lot if you can make sure the UK is onboard with whatever requirements get to me.”
“That's the other reason I’m here. I’ll spend most of next week at DARPA making sure the British and the Americans are on the same page. I’m sure you know, they offered me the Deputy Director position under you, but I didn’t want to pull my kids out of school, and move them to the middle of a desert in a strange country, so I declined the offer. But rest assured, I’ll do everything I can to ensure the politics don’t get in the way of you getting your job done.”
“I understand about not wanting to uproot your family or be away from them for any period. I’m married too, and intend to spend every weekend I can with my beautiful wife, but I am not looking forward to the weekend commute back to Portland. If there is anything you need me to do, then let me know.”
“There is one thing. Try to ensure Fiona doesn't overwork herself. She is smart and extremely conscientious, but she feels, because she said the birds would work, then she is responsible for every aspect of ensuring they do. Make sure she relaxes every once in a while, and doesn't take things too much to heart, especially if they go wrong.”
“I'll keep an eye on her.”
Charles glanced at his watch. “We need to get moving. Otherwise, we will be late for our first MAADS team meeting.”
The two men chatted amiably as they left the office. Each carried a laptop computer in a shoulder bag and a thick stack of stapled meeting notes. They turned left at the end of the white fluorescent-lit corridor, and entered a conference room where ten people were already seated around a large table.
Charles recognized everyone seated around the table, including the two new faces. Fiona, he had met, for the first time, the previous evening, and the soldier in uniform who had walked into his office at eight thirty that morning and introduced himself as 'Sergeant Jackson reporting for duty'. Charles at the time, was busy organizing his materials for the meeting and had been rather curt. He told the soldier he was sharing an office with Fiona, a scientist from the UK and he should attend meetings and try to learn as much as possible about the project.
Bringing a serving soldier onto the project team had not been his idea. It was one of those things that sounded good in a management meeting, but in practice could cause significant problems, by introducing someone who was unfamiliar with the culture of a technology development project. Charles resolved to make the time to talk to Jackson, and explain his role in more detail.
Charles walked to the head of the table, and stood in front of a large whiteboard. He sensed the mood of subdued excitement that always occurs with new projects. The hard reality of solving problems within the schedule would soon temper that mood.
“Ladies and gentlemen, welcome to the MAADS project. Some of you are new to the team, and I intend to describe the project's history for your benefit. Those of you, who were on the Porcupine project, will find some important changes with our new sponsor and project requirements, and we will have to rethink several important aspects of the system.
“This is now a joint USA UK project. We can do the full introductions later, but this is Alan Jones, my counterpart from the UK. He is here for two weeks, helping us setup the new project and coordinate requirements. A couple of people from the UK have joined our development team, including Dr. Fiona Lachlan. She is a biomechanical engineer specializing in bird flight. They all have full security clearances, and you should treat them no differently than the American staff.
“I need to introduce one other person. This is Sergeant Jackson, a weapons specialist in a US Army Special Forces unit. He is here to advise us on integrating a weapon into MAADS, and to give us a combat soldier's perspective.”
The medium-height man in his early thirties wearing a US Army uniform nodded to the other meeting participants. He had propped the cane he used to walk, against his chair. His air of causal indifference accurately reflected his attitude toward his own presence in the room. He didn’t want to be here, but while he was, he had decided to play the role of a by-the-book soldier.
Charles connected the overhead projector to his laptop, and started his presentation.
“Project Porcupine is now called the Mobile Autonomous Area Denial System, MAADS for short.”
A collective groan passed around the room.
“DARPA, the Defense Advanced Research Projects Agency, and the UK’s Ministry of Defense are jointly funding MAADS. Alan and Fiona both work for the Defense Research Agency. I won’t go into the details of DRA’s role, but you can think of it as a consultancy specializing in defense and weapons research.
“OK, let's start with our project objective. DARPA wants us to produce a weapon system capable of denying an area to an enemy force that operates autonomously for an extended period, and is highly selective in its targeting. The other requirements are listed in your handout.”
“How big an area?”
“As big as possible, at least a square kilometer.”
A short heavyset man in his forties reading through the handout asked. “What does 'guaranteed nonlethal self-destruction' mean?”
“It means the system, and its components, need to be capable of destroying themselves without killing bystanders.”
“What about 'conducive to rapid large-scale production'?”
“Sammy, why don't I run through my presentation? It should answer a lot of your questions.”
The man smiled and held up a hand to signify his agreement. The black lines under his fingernails were a sure sign of someone who worked with machines.
Charles showed the first of his slides. “Mobile means it can move around its area of operation, both on the ground and in the air. It doesn't mean the entire system is capable of moving. At least one major component will be static.
“Autonomous means it performs its functions without a human operator controlling it, and can continue to do so for an extended period. We and others, are continually pushing the limits of what can be performed autonomously. MAADS autonomy includes finding and targeting hostile forces, extends to refueling, and could potentially include other functions like rearming. I've had several discussions on these topics with DARPA, and frankly, I've had to rein in some of DARPA's expectations.
“Area denial means it can stop an enemy force occupying or crossing its area of operation. Traditionally, landmines are used for this function, but landmines have a very serious drawback. They are indiscriminately lethal, killing the bad guys, the good guys, and anyone else unlucky enough to be close, when the mine detonates.
“Many mines are left around, killing and injuring people for years after they are laid. This has resulted in a movement to totally ban them. All the mines used by the US and UK armed forces either are recoverable or can be destroyed in situ. However, this has done little to reduce the pressure from the people who want a complete ban.
“An important objective for MAADS is to be a selective alternative to landmines for area denial.
“And our overall objective is to produce several working MAADS systems as a proof-of-concept. By a proof-of-concept, I mean the system demonstrates the feasibility and viability of a fully autonomous weapon capable of area denial.
”Fully autonomous doesn't mean a 'deploy and forget weapon'. We need to closely monitor what MAADS does when it's deployed. The Army needs to exercise command and control over the system. This means we have to provide the capacity to override decisions arrived at autonomously.
“From our perspective, we need to see how MAADS performs, when it's deployed, and learn how we can improve it.
“Safety is another important issue. It's going to be a while before people are sufficiently comfortable with the safety record of autonomous weapons that they will allow them to operate without a human capable of overriding the robot's behavior.
“By the way, I'd prefer us to avoid using the word 'robot'. It has unfortunate connotations from science fiction books and movies. Try to talk about autonomous units and autonomous behavior.”
Charles took a drink of water before continuing. “MAADS will be composed of a number of semi-autonomous, but interdependent units, that work together to achieve the system's overall objectives. While it is not a specific requirement, multiple MAADS systems could cooperate and interoperate to achieve an objective beyond the capabilities of a single system.”
Chapter 8
1 Comments:
He had propped the cane he used to walk, against his chair.
Don't get it.
Post a Comment
<< Home