Philosophy 101 Unit 1

Next Sunday, the $100,000 Loebner Prize group will run their annual Turing test competition to determine the most linguistically competent computer. You can interact with most of these computers online, I’ll dig around for links. In the mean time:

Intelligent Computers put to the test (Guardian Online)

No machine has yet passed the test devised by Turing, who helped to crack German military codes during the Second World War. But at 9am next Sunday, six computer programs - ‘artificial conversational entities’ - will answer questions posed by human volunteers at the University of Reading in a bid to become the first recognised ‘thinking’ machine. If any program succeeds, it is likely to be hailed as the most significant breakthrough in artificial intelligence since the IBM supercomputer Deep Blue beat world chess champion Garry Kasparov in 1997. It could also raise profound questions about whether a computer has the potential to be ‘conscious’ - and if humans should have the ‘right’ to switch it off.

Professor Kevin Warwick, a cyberneticist at the university, said: ‘I would say now that machines are conscious, but in a machine-like way, just as you see a bat or a rat is conscious like a bat or rat, which is different from a human. I think the reason Alan Turing set this game up was that maybe to him consciousness was not that important; it’s more the appearance of it, and this test is an important aspect of appearance.’

The six computer programs taking part in the test are called Alice, Brother Jerome, Elbot, Eugene Goostman, Jabberwacky and Ultra Hal. Their designers will be competing for an 18-carat gold medal and $100,000 offered by the Loebner Prize in Artificial Intelligence.

For this post I would like to present a hypothetical situation, and then hear your comments and thoughts on it!

With current trends, it wouldn’t be surprising if one day we have physical robots that are indistinguishable from humans. Perhaps it won’t be anytime soon, but programming has come a long way, and robotics is also moving along. So, in our hypothetical situation, taking place sometime in the future, we are still in the United States and our governmental system is still in place. In our situation we have this robot who looks exactly like a human, and has incredibly complex programming, that allows him (the sex of our robot friend in this story is a male, thus “him”) to behave indistinguishably from a human. Let’s also say he capable of learning, and he can do all the complex things humans are capable of doing. 

To listen to my hypothetical situation, you have to be able to accept all of the above paragraph.

Let’s name him… Aidan (or if you prefer, A.I. Dan).

Now, lets say Aidan is an experiment, of a robot that has been let out into society to see how he’ll interact with people, the environment, and what he’ll learn. It doesn’t really matter why he is out on his own, the point of our hypothetical situation, is that Aiden is out on his own.

One day Aidan is out wandering about, and runs into an angry looking man, vehemently hitting the bottom of a small defenseless child (we would recognize this as a father spanking his child, and maybe look the other way, but Aidan does not). The child is profusely weeping, and the father continues to look upset. Aidan runs over and commands the man to stop hitting the poor child. The father says something along the lines of “mind your own business” but with more profanity. As the father continues to spank the child, Aidan reaches out his hand to stop the man from hitting the child any longer. The man is now even more enraged (our hypothetical man could use some anger management), so he gets up to confront Aidan. As the man rises, he accidentally trips over something and falls. The man now has a broken arm. Aidan tries to assist the stubborn man, but he refuses help. Because the man looks seriously hurt, Aidan calls for an ambulance. When it arrives Aidan comes with the man and his son to the hospital, to make sure he is okay.

Later on the man is informed that the person who helped him get to the hospital, Aidan, is a robot (he couldn’t tell before because robots behave indistinguishably, remember?). 

This really upsets the man, and so now he is convinced he needs to take the robot to trial and see justice. The man accuses Aidan of battery (because he grabbed the man’s arm against his will).

Let’s say it goes to court. Aidan pleads guilty to grabbing the man’s arm.

I’m going to stop my situation here, and ask some questions that I would like discussion about:

-Does Aidan deserve to be on trial, should a robot go to trial? If not, why doesn’t he deserve the justice that any human would get?

-If he deserves to be on trial (innocent until proven guilty, that’s our system of justice) can there be any argument that A.I. Dan doesn’t deserve to be treated with the exact same respect we would give any human?

-What kind of punishment should Aidan get? No punishment, a fine, community service, jail time, be shut off?

-If you answered that Aidan should be shut off, why should he be? That seems to me like killing a person, and what right do we have to do that to Aidan? (Because Aidan is indistinguishable from a person, Aidan confessed that he doesn’t want to die, or be shut off).

-If you answered that Aidan should receive any other punishment: what good is it to punish a robot?

Please add any other thoughts or questions to our fun hypothetical future situation! 

I’ve been wondering why a majority of people are very eager to doubt the possibility of artificial intelligence.  Is it arrogance?  Is it a belief that “God” made us “special”?  Is it a belief in a vital force or soul?  Are religious teachings preventing “the faithful” from accepting the possibility of artificial intelligence?

I have realized that most people who deny A.I. can’t really support why they don’t believe its possible.  It is a common thought in our culture that humans and other living things have a soul, but a computer is not living and therefore does not have a soul.  It seems as though people make these claims so they will remain to feel comfortable and special.  Nobody can deny that machines are extremely advanced, but they will deny that A.I. is impossible because humans contain some sort of “vital force” that machines will never have.  Machines are made of circuit boards, wires, and other electronic items.  It is hard for humans to hear ourselves referred to as “biochemical machines” because we’d like to think that we are something more than just a “biochemical thinking machine”. 

Religion also plays a significant role in the reason why artificial intelligence can be easily doubted without a concrete logical argument.  In particular, the Christian faith teaches that humans are created in the image of God, but are less perfect.  If religious followers accept the idea of A.I., then the whole basis of their religious beliefs may start to crumble.  God’s divinity might be challenged if a human has the capacity to create an intelligent machine because if we were not created as perfect or intelligent as our creator, then how is it possible for a human to create a being that is equally or more intelligent than a human.  This is what might frighten the faithful from believing in A.I. 

In addition, the Christian faith teaches its followers that someone must live a “good” life on earth in order to achieve eternal life.  We define our conscience as an internal compass that tells us the difference between “right” and “wrong”.  Christians could argue that it is impossible for a bundle of wires to comprehend these “spiritual” things.

In class on Tuesday we watched a documentary on DARPA’s Grand Challenge event back in 2005.  DARPA stands for the Defense Advanced Research Projects Agency.  DARPA is an agency of the United States Department of Defense whose job is for the development of new technology for use by the military.  In order to research new technologies DARPA created an event where teams were challenged to have a driver-less vehicle, one that is not controlled by a remote or any type of human control.  It must be able to navigate unknown terrain on its own based solely off the construction of input devices and difficult programming.  The wining team would receive $2 million dollars.  The first grand challenge took place in 2004 and was set on a 150 mile route through the Mojave Desert.  There was no winner in this contest with the best vehicle traveling not even seven and half miles.  DARPA held the contest again in 2005 with much better results.  All but one of the 23 finalists in the 2005 race surpassed the 7.36 mile distance completed by the best vehicle in the 2004 race.  Five vehicles successfully completed the race.  First place taken by “Stanley”, built by Stanford University.  Second and third places were filled by Carnegie Mellon University with “Sandstorm” and “Highlander” respectively.

The vehicles in DARPA’s grand challenge where designed to be driver-less vehicles.  The main way that the driver-less vehicles were designed were with the following criteria in mind:

• sensors: the car knows where an obstacle and other things that are around it
• navigation: how to get to the target location (i.e. the finish like) from the starting location;
• motion planning: getting through the next few meters, steering, and avoiding obstacles while also abiding by rules of the road and avoiding harm to the vehicle and others;
• control of the vehicle itself: actuating the system’s decisions.

In fact some people have argued that the problem is strong-AI, that a safe, reliable driver-less car would need to use all the skills of a human being, including commonsense and affective computing.  The concern is that driver-less cars will not be able to preform as well as human beings in emergency situations that require judgement and the ability to communicate with other drivers and police.  For example, how should a driver-less car react to a person waving a flare in the middle of the road?  Should it just avoid the person and treat it as an obstacle??  Will it just stop as the roadway is blocked???  These questions are the ones that require the strong-AI systems.  Yet, strong-AI has not been created yet.  Machines cannot act as humans and will have difficulty in these situations.  It is basically an impossible task to create a computer system that can account for all the unknowns while driving a vehicle.  People and bicycles can dart onto the street, animals run across roadways, and being a native of Chicagoland, fellow motorists will always do something completely unexpected, potentially dangerous, and usually inherently stupid.  Can a computer with AI be expected to account for these random, unpredictable events?  I believe the answer to be no.  Technology does not allow computers to behave in this fashion.  So while autonomous vehicles are fun for the races, they do not seem to be at the point where they are practicial.  Currently the best option for a driver-less vehicle would be controlling one by remote control, but that technology has limitations as well.

"Boss" Carnegie Mellon's 2007 Grand Challenge Winner

The 2007 Grand Challenge was another step forward in creating an urban race style versus the desert track DARPA used back in the 2005 race.  The course involved a 60-mile urban area course, to be completed in less than 6 hours.  The major rules including to obey all traffic regulations while negotiating with other traffic and obstacles and merging into traffic.  The 2007 Grand Challenge winner was “Boss”, a collaborative effort by Carnegie Mellon University and General Motors Corporation.  “Boss” won the 60 mile race in just over 4 hours with an average speed of 14 miles an hour.  Coming in second this year was Stanford University’s vehicle “Junior”.  A strong rivalry is present beween both Carnegie Mellon and Stanford with the two teams switching places.  Anyway one is to look at it though, it just shows how advanced technology is becoming and that the future holds amazing machines that will be able to preform tasks that will continue to make them ever closer to humans.

Hello ladies and gentlemen, welcome to the strongest, least debatable, most over used argument in philosophy!

            Alright so we were talking about computers being able to “think”. We argued that nothing they do is by their own right. Yes, they can have sensory input, yes they can store that information, but no, right now computers cannot think. They can give direct output based on information presented to them. They can be programmed very well. Stanley was an amazing machine. It was impressive what it people were able to program it to do. But no matter what people say that machine wasn’t thinking. It did a hell of a job at color coordination by being able to keep on the road, but it didn’t think at all. Today in class we had a debate where I was on the side of machines can’t think. The opposing side’s strongest arguments were:

1: In the example of the Chinese Room, the ledger of Chinese is in our heads, so in fact either, we haven’t really learned English, we just have the ledger in our heads, or computers that have the ability to translate language can think.

2: Everything we do, everything we experience, think, feel, is completely based solely on sensory input.

            What you guys were saying for the first argument I wasn’t able to grasp or something. To me it was sounding like you guys were saying “If you know Chinese then you know Chinese.” Which I figured was one of those given things. But with a computer, it may be able to take a character from the Chinese alphabet and pair it up with letters from the English alphabet and translate, but that doesn’t mean it is able to comprehend the words that it’s showing on its screen. Just having the words doesn’t mean it captures the essence of the object. With humans when we say tree, we think tree, we picture a tree in our minds. Now the tree we picture can be completely original from person to person. If a computer was to produce an image it would be something that a human has put into it or something that it has taken in from some type of optical sensory device. But no matter the compute it will not be able to come up with an original tree. It will come up with an image that has already been perceived by someone, or something else.

Next…

            Yes MOST things that we feel are based on sensory input. But the decisions we make based on those inputs are completely voluntary. Say you take two computers give them the same exact sensory input. Say for argument, you put one robot in a room with a book on the floor. You tell this robot to go pick up the book… what is this robot going to do? IF this robot is programmed to do this, it’s going to go pick up the book, if it’s not programmed to it will not.  That simple it will or it won’t. Humans on the other hand, give two humans the same sensory input and you could get two very different reactions. Whether it’s from different experiences or whatever they have the ability to give you millions up on millions of reactions. Computers will not because they can’t feel. Well, they can’t have feelings anyway. Say you tell one kid to go in and pick up the book that is sitting in the middle of the floor. The kids an obedient kid so he/she goes over and picks it up and brings it back to you. Awesome, for all we know this kid could be a robot. Now are you ready for this? What if… just maybe you had the same kid, cloned, same DNA and same life experiences. Now you tell this kid to go pick up that book in the middle of the room. Now this kid has the ability to turn and tell you to quit being a lazy ass and go pick up the book yourself. AMAZING! Ah the possibility of free will. You could give the same person the same sensory input and get a zillion different reactions from him/her.

            So I’ll take a definite stand. Computers cannot think they are not intelligent. They can be hell of tools, but they aren’t able to by themselves do squat. Someone has to program them, allow them to have input, start them, and then take care of them after they have completed their job. If someone doesn’t will the machine/computer care? Hell no. The computer can’t feel sad, happy, mad, silly, or abandoned. The machine would sit there for all eternity until someone gave it a new command, or set of commands.  

Steven Spielberg created the movie Artificial Intelligence, about a little robot boy adopted by two parents whose own child was frozen due to illness. Due to algorithms, the boy had the ability to love, be happy, and fear. Yet is it possible for this to occur in real life? An algorithm is “a rule or procedure that is guaranteed to give a result meeting certain conditions (you just turn the crank and out it pops)” (Haugeland, 41). Thus using an algorithm to produce the emotion of love, fear, and happiness implies that those are set states of being, with certain definable characteristics for each one consistent each time that state is achieved and the same for every human being. To simulate these emotions, though, it’s impossible to “just turn the crank” and produce them. Firstly, love takes many, many forms: sexual love, familial love, material love, on and on. Secondly, within those categories, there are again many variations; within sexual love, for example, people feel passionate love, intimate love, any mix of both that changes over time, and within familial love exists similar patterns, etc. So for humans to believe that we can reproduce real versions of these emotions is absurd, and to use algorithms to do so is just impossible. Haugeland points out that for chess, there are more possible moves than the total number of seconds since the Earth began, and for any formal system to carry out an algorithm to find the right move would take centuries. Given that love, fear, happiness, and emotions in general are much more complicated and intricate than chess games and the possibilities and variations within each emotion are endless, a formal system cannot possibly reproduce, to the same level a human feels, these or any emotions.

Love, happiness, and fear are all merely primary emotions, however; if humanity wants to produce more emotions than that, say the amount that an average human feels on any given day, we have to account for secondary emotions as well. Secondary emotions are much more complicated than primary emotions, especially if attempting to reproduce them. Primary emotions are innate; they are “biological,” so to speak, existing cross-culturally around the world, possible for every human to feel without outside influence. Words for primary emotions appear in every language, which is a good indicator of their universality. Secondary emotions, however, are not all found around the world. They consist of blends of primary emotions: for instance, contempt is equivalent to anger and disgust. Secondary emotions far outnumber primary emotions, and tend to differ for each person each time they experience it (the same is also true for primary emotions, although the former tend to vary quite a lot more than the latter). If an algorithm can’t even reproduce one primary emotion in a reasonable amount of time, there is no way it will be able to produce secondary emotions.

Primary emotions and secondary emotions are all complex and varied, but almost none of them can be felt on command; humans cannot just tell themselves to be happy and then be happy. If that were the case, psychiatrists would be out of business and wars would be a thing of the past. The same must be true then for synthetic emotions; for the emotions to be real, they cannot be programmed into a machine. However, machines, robots, formal systems, whatever the name, none of them can function on their own without being programmed. Thus, by the simple “P-Q” argument, if emotions cannot be programmed, and automatic formal machines cannot function without programs, then automatic formal machines, or robots, cannot feel emotions.

Reminders:

Post 2 due Thursday night @ midnight

Screening this Thursday at 9pm in the main lounge. Film TBA.

Use this thread to talk about the movie today, The Great Robot Race. Also, here’s the website for the 2007 Urban Grand Challenge. Any substantive comments you make in this thread about the robot race will count towards your comments.

Also,

Here is a picture of a human robot that was created several years ago. I encourage you to look at the picture. When I first saw it, the picture was in a newspaper of the two in a subway. I could not differentiate the two.

http://www.crunchgear.com/2007/04/27/when-robots-attack-what-does-the-future-hold/

Here is a video showing the movements and abilities the creators were able to manipulate mechanically on this robot to make him appear more human:

and if that doesn’t work, here’s the link: http://www.youtube.com/watch?v=RksP_gAqSh0

Now in terms of believing what something is based on “what it seems,” which I believe is a Descartes-ism, here is a video from youtube that a human appears to be a robot based on how he moves and the sounds emitted. Ignore the first thirty seconds.

And here’s the website in case that doesn’t work:http://www.youtube.com/watch?v=SR4eBe5FNwU

To know the difference between human or robot, must we see the “object” being viewed lagging? Or rather, must it be imperfect about it’s movement in order to determine how “real” the being is?

See this video to understand what I’m talking about:

Or if that doesn’t work here’s the website: http://www.youtube.com/watch?v=qqdluIN2Wjo

My final example is an excerpt from Bicentennial Man, the movie with Robin Williams if you can recall…

Website: http://www.youtube.com/watch?v=z5YMEwX2-88

Now that all that fun stuff is out of the way, let’s focus on what Haugeland has to say about Cognitive Science. What I have learned thus far from our philosophy class (or at least the one concrete thing that I have learned) is that every word counts in philosophical writings. A close reading of philosophy forces you to question what does that word really mean? What are the intentions of the philosopher based on his use of ___ word? Initially Haugeland states, “The basic idea of cognitive science is that intelligent beings are semantic engines —in other words, automatic formal systems with interpretations under which they consistently make sense.” He builds on this further by stating, “with universal hardware, any semantic engine can in principle be formally imitated by a computer if only the right program can be found.” These statements suggest that a) we need to investigate further as to what he means by “intelligent,” b) that human beings are semantic engines, and c) that a computer can “imitate” “any semantic engine” such as humans as long as the right programming has been done. So the examples I placed in this post are considered imitated semantic engines. Can we imply that these imitations are in fact intelligent?

Later on in the article, Haugeland takes the “devils advocate” role on this view of semantic engines. He claims, ” people just aren’t semantic engines…no semantic engine can be genuinely intelligent.” That buzzword intelligent has got my mind reeling. Is artificial intelligence what he infers by the use of the word intelligent? Or is it rather the type of intelligence that comes with Aristotle’s meaning of intelligence or rather his definition of wisdom?

Haugeland cites two strategies for arguing that cognitive science is misconceived. The first being the “Hollow Shell” and the latter being the “Poor Substitute” method. The first strategy suggests that it doesn’t matter how well you program a semantic engine, it can’t really understand anything because it’s lacking this “X” factor. The second, or Poor Substitute, method suggests that such engines are even incapable of acting or “imitating” the notion of understanding. He cites three different possibilities for this “X” factor: consciousness, original intentionality, and caring. The first, consciousness, relates to what was said by a philosopher we already studied. (I can’t remember who it was.) It was something about this internal almost magical quality (help anyone?). It was basically what we would call our “essence.” The argument here is that a semantic engine (which for the remainder of my post I’m going to regard as a human robot for simplicity sake) is incapable of understanding because it lacks a consciousness. What’s interesting is that no one has a specific understanding of consciousness, so it can’t really be stated that human robots are incapable of this quality or rather that they lack it.

The notion of Original Intentionality is just a fancy way of stating that things only have meaning because we give it to them. This brings up the concept that words and symbols and numbers and gestures etc only have meaning based on what we (as humans) have assigned to them. How do we know that human robots don’t have this “X” factor? As Haugeland states, “It seems that original intentionality must depend on whether the object has a suitable structure and/or dispositions, relative to the environment.” That being said, I challenge whether or not such a “suitable structure and/or disposition” can be programmed? Afterall, when we raise children, aren’t we essentially programming them to have just that? I believe that in terms of human beings, original intentionality is programmed by the individuals that raise us or impact us greatly in the way we view our environment and understand the world around us. Haugeland also discusses that a “perfect robot…would seem to act on its own opinions, defend them from attack, and modify them when confronted with counter-evidence–all of which would suggest that they really are the robot’s own opinions.” I find that hard to believe since afterall, a robot’s responses would be programmed. THUS, these reactions/actions would be programmed by a human. This would infer then that these responses of sorts are not the “robot’s own opinions” but rather the creator’s or the author’s opinions.

The third X factor just makes me feel like the combination of these three elements describes wisdom rather than intelligence or cognitive science (but wisdom in the eyes of Aristotle). As Haugeland explains, “the intuition is that a system could not really mean anything unless it had a stake in what it was saying–unless its beliefs and goals mattered to it. Otherwise, it is just mouthing noises, or generating tokens mechanically.”

Reading this article made me think of Bicentennial Man and the human-made robot that I cited earlier in the post. All of those really seem to be human. It’s either the way they look, act, talk, respond (via gestures) that makes them appear human. IF a robot can be programmed to fulfill the three X factors described by Haugeland, does that mean that they are in fact human or rather that they are semantic engines? When watching Bicentennial Man, it’s so easy to accept the notion that a robot could feel, doubt, react (and all those terms used by Descartes). Are we in control of our senses? Or do they control us? I bring this up because I can accept that Robin Williams is a robot in the beginning part of the movie until he has “skin” and thus looks like a human. I can accept that the guy in the second movie clip is a robot until I see him move freely after his presentation.