Wednesday, June 30, 2010

Discussion with Robert

THIS IS AN EMAIL THREAD-- PLEASE READ FROM BOTTOM TO TOP
--------------------------------------------------------

On Fri, May 28, 2010 at 1:16 PM, asuarez510@gmail.com wrote:

Quick response cause I'm in a meeting...

Suicidal robot: asimov's three laws almost require it. A robot must obey instructions from a human above its own existence. This would never allow a bot to behave humanly, since self preservation is the strongest of human instincts.

Sent from my Verizon Wireless Phone


----- Reply message -----

From: "Robert Sobrado"

Date: Fri, May 28, 2010 12:44 pm
Subject: Chat with andrea

To: "Andrea Suarez"


I'd love to write a book with you ;). Your job is amazing what you do and the complexity of the joints you are recreating are no simple task. The wrist is an injury that never fully heals. If you guys succeed that will really make a difference in the quality of life of many people.

Back to AI. First, I never give much credit to claims of any research institute. Second, ten years ago computers got faster at an alarming exponential rate... today all we can do is add more processors a practice that has been around for almost 20 years. Third "Programming a formula that changes based on certain outcomes." The outcome may never be perfect. If for example we built a Mountain Adventure Robot of Fail (MARF.) MARF is built to run up and down a mountain as fast as possible looking for an optimal path with ground strong enough to support it moving at some maximum speed. MARF starts in a rainy Spring, his path from a wet environment will change in a dry summer, and in Winter every thing it knows will reverse and his top speed will drastically change when running down hill. There will never be one all encompassing correct program for him to run with. Only situational ones. How well he detects his environment may improve his chances at deciding what to do but he will be at a constant state of failure and improvement. The goal of an optimum path is impossible. MARF was designed to fail. A modern program can repeatedly fail and never know it, same can be said for most people. If computers ever realize all they ever can hope to do is Fail. We will create an army of suicidal robots. Humans and animals can cope with failure a machine will only follow it's programming.

This is fun! I love playing Devils Advocate.

On Fri, May 28, 2010 at 11:11 AM, Andrea Suarez wrote:

You said there is no such thing as a code that can learn- But what about a program that adapts according to the results of past attempts? That could certainly exist now. Isn't that learning? Programming a formula that changes based on certain outcomes. Doesn't repeat bad results, looks to repeat good ones (the of course, people do repeat bad results!). Kind of like an optimization. But then, what does it consider to be "good results," because that certainly varies treemendously from human to human. You would start with the same exact program, and achieve different results and different decisions from each machine, based on what decisions it is faced with, and the outcomes of each decision.
And your assumptions are based on existing technology- what about 50 years from now? There are already research institutes that claim they are getting closer and closer to accomplishing what they call "mind uploading." The brain is also a series of electrical impulses (mind you, a much more complex one). Take your smartphone example, multiplied by a huge factor. How come the brain is able to adapt and learn, but artificial intelligence is not? Is it a matter of further complexity of existing circuits, rather than a lack of appropriate technology.

Very interesting stuff!! Want to write a book with me? lol

On Fri, May 28, 2010 at 11:58 AM, Robert Sobrado wrote:


A lot can be attributed to, as you said, "our decisions are based on a subconscious formula involving a certain initial moral code and the result of past decisions." But that is incomplete. You don't give enough credit to personality and stimulus provided by immediate surroundings. A Pig headed person will attempt the same experiment multiple times and expect different results. Like extended CPR on a loved one. A puppy can fall in a pool and for the rest of its life fear pools where a human can overcome a similar experience. A guy can hold a door open for a women to pass 99% of the time but one time he can be lost in thought an just not notice the old lady struggling to open a door with a coffee pot in her hand. Two identical girls can grow up in the exact same environment doing the exact same things together their whole lives and not like the same qualities in a man. I have programmed autonomy there are no real mysteries left unsolved. It is a computer it does what you tell it to do without fear or hesitation. There is no such thing as code that can learn. There only exists code that can copy and repeat. If you teach a program how to walk on a flat surface and how to climb a vertical one it has no chance of figuring out how to climb down from a cliff. These are things babys do before they learn to speak. I am a big Nerd. If something is simple there is less room for failure. All modern technology can be broken down into very simple processes. A smart phone is just a fancy radio with a few micro processors that transfer and collect data by V=IR. With enough of them, signals, working simultaneously you can do some amazing things.


On Fri, May 28, 2010 at 10:24 AM, Andrea Suarez wrote:

haha! We're not making people into frakenstein over here! we try to rebuild your own broken bones (in tiny little parts after, say, an ATV accident) with Ti implants and screws to avoid a full prosthesis and additional surgery every 10 years. We most recently finished the elbow, and are moving onto the wrist now. Right out of surgery you can't even feel the implants- its your own joints you're using.

Funny you mention the robocop thing- I actually did a series of blogs for a class last year on the integration of bionic bots into society as workers, etc that I may be turning into a book. The professor showed interest in publishing it if I continue developing it. Figured I'd give it a shot. Right now I'm looking into the possiblities and consequences of programming free will and psychopathic robots (after all, our decisions are based on a subconscious formula involving a certain initial moral code and the result of past decisions, right?)

Monday, May 31, 2010

Robot Evolution

“It’s a Bot-Eat-Bot World”

Insect expert Laurent Keller has become the first researcher to be able to experimentally address questions of evolution. In his lab at the University of Lausanne in Switzerland, his team is using robots to condense thousands of years of evolution into days of experiment. Particularly, they are using robots to study the evolution of communication in species. Their results have been more enlightening than expected, showing how species evolve for survival from one generation to the next.
They call these small robots the “S-Bots,” which have been built so that each generation lasts two minutes. They stand only 15 cm tall and are equipped with lights, which can be turned on or off, and what Keller calls a virtual “genome.” This programming dictates their response to the environment that surrounds them.
The S-Bots were placed in an “environment” which consisted of a “food source” and a “poison source.” The methodology that follows is simple: robots that found the food source reproduced and passed their programming on to the next generation.
After 500 generations, equivalent to thousands of years of evolution, the robots had begun to communicate using their lights. Although some used them to signal food and others to warn of poison, the robots ultimately became much more efficient within their environment. Keller had expected the robots to act independently, largely unaffected by the existence of the other robots. However, not only did the robots develop a system of communication, they also became deceptive. Although the robots had managed to survive, they did so in different ways and with different programming. Robots with similar programming clearly favored each other, and even signaled “strange” robots incorrectly to decrease their chance for survival. They began to cooperate in communities, an evolution much more sophisticated than what was anticipated by researchers.
The results of this experiment were astounding. They have greatly amplified our understanding of social communication. Until now, we were limited to the evaluation of the results of the evolution of communication, but this research has allowed the evaluation of the process of this evolution. Understanding of this process is much more insightful and the information is more versatile to similar fields of research.
This experimental research that uses robots to observe social behavior, particularly the evolution of communication, is expanding our understanding of communication and widely expanding the possibilities for future research of social behavior and evolution. These robots make it possible for researchers to study thousands of years of evolution within the limits of a laboratory. Lee Dugatkin, an evolutionary biologist at the University of Louisville finds the new research invigorating: “using robots to understand the evolution of communication opens the door. It has tremendous potential to address all sorts of questions that haven't been answered yet.”

Monday, May 3, 2010

Psychopathic Robots?

"If you build artificial intelligence but don’t think about its moral sense or create a conscious sense that feels regret for doing something wrong, then technically it is a psychopath." Such is the opinion of Josh Hall, who wrote the book "Beyond AI: Creating the Conscience of a Machine." Throughout this discussion of bionics, we have only briefly mentioned the ethical issues that surround the topic and the research that could potentially make advanced bionics possible. As the last blog of the semester, it seems appropriate to dedicate some time to "robo-ethics."

Although the aid of robots have been unparalleled in many circumstances where the risk to a human would be too great, it has resulted in the death of humans. Two years ago, for example, a military robot in the South African army killed nine soldiers. This was, of course, a malfunction and not a result of the robot taking control of free will, but nevertheless, it forces us to question how to what extent are willing to trust such robots with our lives or the lives of others. In a Swedish factory, a robot machine injured one of the workers. Although this was attributed partly to "operator error," the factory was still held responsible and forced to pay a fine for the injury of its worker.

Asmiov's Three Laws of Robotics:
1. A robot may not injure a human being or allow one to come to harm
2. A robot must obey orders given by human beings
3. A robot must protect its own existence

Each of these laws take precedence over the following one. For example, a robot may not injure a human being, even if it is ordered to do so. It must obey law #1 over law #2. What becomes interesting then is law # 3: "a robot must protect its own existence." This means that if ordered to obey law #2, it must unconditionally follow any ordered given by any human being, even if it results in its own destruction. This brings up something that I had not considered in the first bog, when we were discussing what it means to be "human." Perhaps we should add a sense of self-preservation to the list we had originally compiled. In this case, if we strictly follow these 3 laws for the safety of humans in a world where robots are becoming more and more advanced, it seems that these robots could never truly behave in a human fashion as long as they are willing to follow orders over the preservation of their own "lives."

Chien Hsun Chen and Yueh-Hsuan Weng recently coauthored a very interesting paper, which was published in the International Journal of Social Robotics. It can be found here: http://works.bepress.com/cgi/viewcontent.cgi?article=1000&context=weng_yueh_hsuan. It attempts to provide solutions and guidelines to all of these ethical robotic dilemmas. For example, develop a set of guidelines for the punish of a robot and the creation of what they call a ”legal machine language” to help police future bionic bots. They explain that it is important to distinguish who takes the credit or the blame for the performance of a robot. In the example of the Swedish factory mention earlier, the factory was held responsible for the actions of the machine, even though the malfunction was attributed in part to faulty operation by the user. This reflects a general notion today: if you build a robot, you are responsible for the actions (good or bad) of that robot. But what if the robot was complex enough to make its own decision. In that case, the creator of the robot would be very much like a mother or father: you raise your kids as well as you can, and then hope that they make the right decisions. A parent does not go to jail if their son or daughter commit murder. Should this be the case with the creator of a robot that is able to make its own independent decisions?

In the paper, Chen states that a "human-robot co-existence society" could be possible as early as 2030. If he is correct, this co-existence would happen during our lifetime. As far away as a completely automated human robot may seem, there is more evidence of our quick advances than we may recognize. Looking though UM's own Miami Magazine, the topic of bionics and research toward that goal can be easily found. An article title "Differently Enabled" tells readers of the No Barriers Festival that took place at UM this past June. This event was part of the Clinton Global Initiative at UM. The BrainPort, which turns a video image into electrical impulses that are sent to the brain via the tongue, among many other new technologies were showcased at this event.

Wednesday, April 21, 2010

Rebuilding Brain Control

We have discussed using amputated nerves to work with a machine to give the user control over that machine. These nerves can even be trained to perform different tasks. But what if we were to reattach a donors limb instead of a machine? The complexity of reattaching a human body part makes the possibility of more more advance prostheses a much closer reality.

French researchers were successful in transplanting 2 hands onto each of 2 amputee patients last year. Their goal was to investigate whether the patients could relearn to use their hands in a natural way. This research had great implications. If a patient can relearn to use an implanted hand, they may be able to learn how to use a prosthetic that has as many electrical connections as a human hand. This prosthetic could then be as functionally complex and as realistic as a human hand. This research could even have implications as great as eventually being able to use a human brain to control a fully robotic machine, eliminating the need for advance artificial intelligence to make a fully bionic being possible. Perhaps Ishiguro (from previous post), could build an android with real human body parts that is control with an advanced future version of his existing software for remote control. Or perhaps he could control such an android with his own brain!

Transplanting full human body parts has only been possible in the past 10 years. The difference between limbs and organs is that limbs are much more complex in terms of their connection to the nervous system is much more complex and that they are made up of multiple tissue tissue types, as opposed to an organ such as the liver. Although nerves are the most important connections in such a transplant, reconnecting the other physical tissue types, such as bone and muscle, is also essential.

When the transplantation of the donor hands has been completed, the brain actually starts to regenerate those nerves by 1mm each day! This means that in as little as a few months, the patient is able to feel and move their new hand. The ability of the brain to regenerate the nerves after they have been connected also shows a great level of promise for the future of prostheses. This research leads to the possibility of connecting artificial nerves for a few months in the hopes of regenerating a person's natural nerves in order to be able to use them to control an artificial limb or other body part. They explain, "Our findings show that newly transplanted muscles can be recognized and integrated into the patient’s motor cortex."

One interesting this to note is that although both patients were right handed, they both agree that after the transplant they had significantly better control of their new left hand. Although the researchers dismisses this as insignificant, due to lack of data points (only two patients), it still remains an interesting phenomenon to me. According to Kass, "It could depend on just technical things. It could depend on how well the hand is connected or how much damage there was done to the nerves and muscles." Maybe the donors were actually left handed, and this has a greater effect that the fact that the patients were right handed?

I always like to read the comments at the end of such an article. I think that the average person's reaction to advances in these fields may be a good indicator of the general acceptance of these ideas, as well as the proximity of their realization. One person brings up the possibility of using a person's own stem cells (although they do not specify whether these would be adult stem cells or embryonic stem cells, and this technology is still in the distant future) to grow a donor hand that would have no chance or rejection and would eliminate the need for the patient to take immunosuppressant drugs for the rest of their lives, while preventing other complications. Another person writes: "Nah. Someone will just patent the process and then the BPAA (like MPAA, but stands for Body Parts Administration of America) will try to make it illegal to make backup copies of ourselves. Of course, we’ll just get black market versions via B-Bay (Body Bay) and BodTorrent. We’ll upload genetic blueprints that others can copy into their human DNA coding/decoding machine, downloading only compatible parts, of course. We will translate this blueprint into gene maps via the coding machine and use its store of our pre-created stem cells to “print” our organs and body parts."

A third person writes, "How much do you want to bet that by 2209 people will consider adding an extra limb or finger just like getting a tattoo. It will bring body modification to a whole new level." Although right now tissue rejection would not allow this to be possible, they did say 2209!



http://www.wired.com/wiredscience/2009/04/handtransplant/

Thursday, April 15, 2010

Building androids to understand humans?

Hiroshi Ishiguro is a roboticist at Osaka University in Japan, and he has created an android version of himself.

Although the robot does not have independent intelligence and it does not have the full body motion of its creator, Ishiguro can remotely control the robot to mimic his every move.

Ishiguro studies human-robot interaction. In our investigation of bionics and artificial intelligence throughout these blog posts, we have functioned under the general assumption that through the study of humanity, we will gain a better understanding of artificial intelligence and the creation of a fully robotic "human." However, in Ishiguro's studies he seeks to understand human behavior and intelligence through the study of androids. The reversal of this concept brings a new perspective into our discussion. ”My research question is to know what is a human,” he explains, which was the topic of my first blog post. He goes on to explain, "I use very humanlike robots as test beds for my hypotheses." Many of the questions we have discussed throughout these blog posts are being tested by Ishiguro and his robots.

Ishiguro argues that robots can already have the ability to do many of the daily tasks that humans perform. He can see robots doing household chores, caring for the elderly, etc. However, he argues that in order for the average person to accept a machine in these roles, the machine must behave as if it were a human. All of us have experienced many attempts at replacing human tasks with a machine. Sometimes, this is successful, as with ATM machines. Other times, it is frustrating and impersonal, such as with automated phone machines. In Ishiguro's mind, there is no reason that we need to be able to tell the difference. He explains that we are wired for human-human interaction, and this is essential if robots are to become an integral part of our daily lives.

The Japanese have a term called sonzaikan to describe the feeling of being in the presence of another human being. But as we have already discussed, where does the sense of humanness come from? and more importantly, how can you translate this to a mechanical robot? The world population of industrial and service robots is about 8.5 million! Still, like Ishiguro explains, these robots are further from "human" and closer to "appliances." IF you look at youtube videos of Ishiguro's android twin, the comments reflect many mixed feelings. A lot of people find the android's likeness to its creator scary and unnatural. In my opinion, an android that attempts to mimic a human likeness must be extremely successful in achieving this, otherwise I think it would be even less relateable than one that looks like a garbage can with arms and legs. When a robot looks like a robot, people know what to expect, and they are familiar with them. However, when a robot looks so much like a human, but is missing all the nuances and and fine motions and reactions of a human, I think the most common reaction would be to reject the robot, or even be afraid of it. It defeats the natural interface that Ishiguro is trying to achieve.

In Ishiguro's latest robots, he even wrote software to mimic involuntary human movements, such as blinking and breathing. While this may aid in the "human-human" interaction that he seeks to achieve, it also brings up different questions. To what extent does a robot need to appear human for humanity to accept it. Perhaps a robot with fluid motions, but without a human face, is enough. In the movie iRobot, many of the robots have a human-like appearance, but are still machines made out of metal and clearly distinguishable from a human. Still, they are able to integrate into humanity and become a part of the daily life. I think that many people find comfort in being able to tell a robot apart from a human. I think instead, it is more important for the robot to be able to react accordingly when it is spoken to. If a robot behaves in a human way, I think that people would treat it like a human regardless of how it looks.

"Humankind is always trying to replace human abilities with machines. That’s our history. I’m doing the same thing. Nothing special."

Monday, April 5, 2010

Bi+onics

Bionics is formally defined as the study of mechanical systems that function like living organisms or parts of living organisms. The word itself comes from a combination of life (bi) and electronics (onics). This concept was first widely popularized in the 1970s series The Six Million Dollar Man. After Steve Austin, a pilot, is injured in a crash, he becomes superhuman by rebuilding his body with bionic arms, eyes, and legs. Of course, modern bionics is focused on restoring human ability, rather than enhancing it to a superhuman level. Joseph Pancrazio from the National Institute of Neurological Disease and Stroke explains that to him, bionics is about "restoration."

Richard Lipschutz works at the Rehabilitation Institute of Chicago. He explains that the basic technology of human prosthetic arms hasn't changed dramatically in the last 100 years. Although materials are different, the basic idea of "hooks and hinges moved by cables or motors, controlled by levers" has not changed much. The main issue with these prosthesis is that they have to be controlled by other parts of the body, which is very unnatural. For example, a prosthetic arm may be controlled by pressing a small lever with your chin. This can be very awkward, and lead to other pain. Many times, amputees give up on these devices.

Todd Kuiken is a biomedical engineer at the Rehabilitation Institute of Chicago, and he is responsible for the "bionic arm" that Amanda Kitts has been testing (previous post). He tackled a major problem in neural prostheses: being able to connect the electric impulse from the nerve endings to the prostheses. He can't simply connect a computer cable to the nerve to pick up the electrical signal. This is dangerous to the nerve, and introduces and extremely high risk of infection. Kuiken found his missing connection in muscles. The contraction of a muscle provides the necessary amplification of the electrical signal from the nerve necessary for the bionic arm to be able to recognize it. To use this, he had to develop a technique to "reroute" the damaged nerves to other muscles. In Amanda's case, these were the upper-arm muscles.

If you search "bionics" on google, you over 527,000 websites. If you then go to google images, you will find over 50,000 images, ranging from cartoons to human prosthetics. I think we have reached a point where the average person is not only familiar with the concept of bionics, but has also formulates at least a basic opinion on the topic. I asked about 10 people among my family and friends, and the general opinion seemed to be that it is a concept in the future. When I asked them if they ever worry about bionics and its potential, everyone had the same answer: not in my lifetime. As we have discussed in these blogs, the greatest advances in bionics have been in the medical field. Restoring human motion to people who have lost it due to disease or war has become the primary focus of bionics. However, as this technology comes closer and closer to a perfect imitation of human functionality, how long will it be before all these parts are assembled into a "human" robot?

The future possibilities of bionics do not all have to be scary and reminiscent of Terminator. There may be very useful applications to these human androids. What if we were able to create these robots that instead of being artificially intelligent, could behave like an avatar and mimic our motions. They could then potentially replace our soldiers at war and turn a battle of bloodshed into a battle of technology (and money). This could revolutionize our world and save many lives in the process.

Sunday, April 4, 2010

Merging Man and Machine in 2010

Although we are stay a long way from a fully functioning artificial human body, we are taking steps toward restoring different aspects of human functionality faster than ever before. The focus of this research, however, remains focused in restoring lost human ability rather than enhancing it or creating an artificial being. In fact, most of these technologies still rely on the human brain and focus on restoring motion, sight, etc to interact with the brain as similarly to normal as possible. Without an artificial brain, these technologies would be useless on a fully bionic man. I think that this will put the concept of a fully bionic man far in the distant future. While most of these procedures and technologies remain "experimental," many patients are greatly benefiting from these advances:

Aiden Kenny, 10 months old

Aiden Kenny was born deaf beyond the help of any existing hearing aids. His cochlear implants bypass the parts of his ears that don't work by carrying an electrical signal directly to his auditory nerves. Within months of his surgery, in which he had 22 electrodes implanted under his skin, Aiden was learning his first words like any 1 year old boy. The implant is composed of several parts. an antenna receives the auditory signal from a microphone, which is contained in an external speech processor. A magnet holds the external transmitter which sends the signal to the antenna. The receiver directs the signal from the antenna to the electrodes, which are placed in the cochlea. Finally, the electrode array stimulates Aiden's auditory nerve to transmit the signal to his brain. John Niparko, a surgeon from Johns Hopkins University, calls him "a real bionic kid."

Amanda Kitts, 40 years old

Tomorrow's People- a group whose missing body parts have been replaced by machine that can respond to their own nervous system. This means that they can control the machine with their brain, similarly to how they control any other natural part of their body. These "machines" are called neural prostheses and most recently, bionics. Amanda Kitts lost her arm in a car accident in 2006. However, her brain did not lose the sense that her arm was still there. Often referred to as a "phantom limb," recent amputees can often continue to feel their missing limb and even forget that it has been amputated. This feeling is due to impulses that the brain continues to send and receive to the missing limb, even after it has been amputated. Amanda has volunteered to test one of these neural prostheses. The technique is called targeted muscle re-innervation, first used in 2002. Although this technique does require surgery to reroute the damaged nerves to major muscles nearby (see next post), the results are astounding. Amanda is now able to use this bionic arm and control it in the same way as she controlled her natural arm. She can turn her wrist and extend her elbow, and she can even feel how hard she is grabbing something, or if it is smooth or rough (thanks to pressure pads at the finger tips of the prosthetic arm!

Jo Ann Lewis, 79 years old

Jo Ann lost her sight to retinitis pigmentosa several years ago. This is a degenerative disease in which the rods and cones, the light detecting cells in your eyes, are destroyed over a period of time. The retina is normally filled with ganglion and bipolar cells, which gather signals from the rods and cones and pass them on to the optic nerve through fibers. Many patients with retinitis pigmentosa still have a surviving layer of retina, but until Mark Humayun, from Second Sight and the University of Southern California, no one knew how to feed it images in a way that the retina could understand. Jo Ann was among the first patients to use the Argus, which is a pair of dark glasses that have a very small video camera. A radio transmitter send signals from the video camera to a small computer that can be worn on a belt. This computer is then able to translate the video signal into a series of electrical impulses that the retina is able to understand. The patient also has a square array of 16 electrodes attached to the retinal surface, which are triggered that the computer. This has allowed Jo Ann to see edges and rough shapes. She says, "Now I'm able to see silhouettes of trees again. That's one of the last things I remember seeing naturally. Today I can see limbs sticking out this way and that."