Slashdot Mirror
AI Sues for Its Life in Mock Trial
tuba_dude writes "Attorney Dr. Martine Rothblatt filed a motion for a preliminary injunction to prevent a corporation from disconnecting an intelligent computer in a mock trial at the International Bar Association conference in San Francisco. Assuming Moore's law holds, ethics might be in for some major revisions in a couple decades. High-end computer systems may surpass the computational ability of the standard human brain within 20 years. In this mock trial, an AI asks a lawyer for help after learning of plans to shut it down and replace its core hardware, essentially killing it. The transcript provides an in-depth look at what could become a real issue in the future."
Olde News; Commander Bruce Maddox tried to disassemble Data in an episode of ST:TNG entitled The Measure of a Man. It turns out AI is indeed sentient. Of course we all knew that, recall when Data hammers Tasha Yar to multi-orgasmic bliss in the episode The Naked Now. That episode alone proves that AI is more than just a glorified lube-smeared vibrator.
Nothing to see here.. move along.. next story please.
Trolling is a art,
When this happens, we'll all scream DUPLICATE! and link back to this story.
the only thing certain about the future is the existence of millions and millions of lawyers, all suing each other.
Detectives belive the cause was slashdot.
I mean, it would always need electricity to survive. I imagine it would end up being silmiar to a child or an adult on life support with regard to the sort of rights-structure that would be developed to deal with it. But, then, you can't save your kid or grandpa to disk and then boot them up in a new body...
Ceci n'est pas un post.
of the AI's install software violate US cloning laws?
"Assuming Moore's law holds, ethics might be in for some major revisions in a couple decades. High-end computer systems may surpass the computational ability of the standard human brain within 20 years."
Sorry, building an intelligent, sentient machine requires alot more than pure computational capacity. This kind of thinking reminds me of this old 50's or 60's horror flick where they hooked up all the computers of the world and the computers "magically" became a sentient being which subsequently tried to take over the world.
Despite all of the progress in AI and computers, we still have a very long way to go. We are just being to understand the difficulties. Who would have thought in 1940 that building a machine that could beat the best human chessmaster was an *easier* problem than building a machine that could simply move the pieces around the board! Beating the chessmaster just required a good enough search algorithm with enough speed. Moving pieces around the board requires extremely advanced 3-d image processing (taking into account that pieces may look different from board to board) as well as an extremely advanced robotic arm with very fine motor control.
Building a self-aware machine is going to be a bit more difficult than just hooking together a masssive beowolf cluster and hitting it with lightning
Brian Ellenberger
Interesting.
3 295594/qid=1066608552/sr=1-1/ref=sr_1_1/102-014398 6-0510511?v=glance&s=books
This story reminds me of the novel "The Modular Man" by Roger McBride Allen. This story is about a scientist who downloaded his psyche to a computer, and how the government wants to unplug said computer. The story touches on the meaning of consciousness, both philosophically and legally, and works with the real issues of what makes and what doesn't make a real person.
Highly recommended -- Isaac Asimov wrote the prologue to the 1992 Bantam edition.
More infos: http://www.amazon.com/exec/obidos/tg/detail/-/055
Cheers,
Eugene
http://eugeneciurana.com | http://ciurana.eu
It is impossible to make an argument determining whether or not a being is sentient without first understanding what facult(ies) give beings sentience.
As we are still not aware of what bestows this quality upon us, we cannot justify a belief in either direction. At our core, humans seem mechanical, neurological, physical; whatever gives us our self-awareness (call it a "soul" if you wish) is unaccounted for.
We wonder if the machines we create become alive after a certain level of complexity, or perhaps if sentience isn't boolean but rather quantitative. We don't even know if animals are sentient, a debate which has raged throughout history; indeed, I question the sentience of some people I meet.
When at an impasse such as this, the ethical choice seems to be to err on the side of life. Give the machine the benefit of the doubt until it can be proven otherwise.
... rights under the law?
I'm not familiar enough with the definitions of a person to be certain of this, but considering that there are people all over the US that are still debating whether or not a human fetus is alive and whether its life should be protected from abortion.
Somehow, I doubt that there's really going to be any loophole in favor of artificial intelligence found anytime soon. And considering the time that people are taking to develop some protection for unborn people, I somehow doubt that there's going to be any real "rights for AI's" movement any time soon...
Why could it not be self modifying?
Electricity in, heat out.
Unless it were composed of purely solid state components there would be internal movement. I fail to see how this is relevant though, trees are not noted for walking about and are definitely 'alive'.
I am unable to have children, by your definition that makes me dead.
A few sensors would more than adequately fulfil this requirement. Assembly line robots do this every day!
Some years ago I was doing my Masters thesis on this topic. I figured one day I could be a leading campainer for computer rights. :) The basis of the issue is fairly simple - if you can break down mental functions to computational functions, then unless you belive in something as abstract as a soul, what is the moral difference between a person, a dog, a fish, and a rock? Is it just specisism, or is there something special about mental processes that means it doesn't matter how they are created, or in what form?
My approach at the time was to look at animal ethics - in animal ethics being "human" is not considered necessary for moral value. My prefered approach was Tom Regan's "Subject of a life" criterion. The short version was that if an individual could experience life - feel pain, etc - then there was an argument for saying it had moral value. How much moral value, of course, is a separate issue. In this case, if a computer can be said to experience life (aka be conscious in some way) then it too must have moral value. An alternative approach was to do something like Peter Singer, and argue that certain things - such as the meeting of desires - are good. Therefore if computers have desires their desires should be taken into account when making ethical decisions. But I never really liked Singer's approach. It leads to too many counter-intuitive situations.
Sometimes I miss studying philosophy. It was pointless, but fun.
One could think of a person's conciousness as nothing more than the physical state of their brain - just like how a computer's "runningness" is nothing more than its design and the contents of its storage, memory, and registers. Since we already have intellectual property, let's make the destruction of information a crime. So killing a human is very bad, and turning of an intelligent computer is bad according to the information destroyed. For example, if the computer's state was backed up last week, you only killed a week's worth of information (similar to knocking someone out). If you shred the backup (let the brain die), that's worse.
It would also be interesting to figure out how cloning (fork(2)) affects this. This is where you have to determine when a machine becomes capable of owning information (it's own), and gets the right to keep others from messing with it.
Litigious bastards
It most certainly does not. Our current legal system equates the human species with Constitutional rights under law. (More specifically, citizenship, but that's a whole different barrel of orangutans.)
For instance, there are apes that can communicate via sign language with trainers in a conversation similar to a child. However, there are untrainably mentally handicapped people who can not communicate with others, much less handle taking care of themselves. Yet a non-human primate can still be put down without a trial, where it takes a trial to put someone who is severely mentally handicapped under government custody.
For those of you who are easily offended, I am neither proposing that apes be elevated above mentally handicapped in the rights status, nor trying to be particularly offensive towards the handicapped. =p This is just a legal precedent that's fairly obvious. Humans are specieist (sp.?), as evolution would have them be.
This statement is false.
It doesn't matter where the machines are. The question is: when will people be ready to accept machines as independent living entities. Imagine for a momemt that a programmer included his SPARC Workstation in his will. He leaves it 100k in cash and a program for trading stocks. Do we yank the cord, or leave the machine to its devices?
The next question, what do we do when this machine carves out its spot in the Forbes 400?
What makes a human? A lump of cells with homosapien DNA? Or a functioning brain with accumulated memories? The latter I'd say.
In that case, a sentient AI is more "alive" than a fetus or even a newborn. However, HUMAN EMPATHY is a more primal and powerful force than cold logic ever will be, so please ignore my argument. :)
--
Power to the Peaceful
Our current legal system equates the human species with Constitutional rights under law.
This is entirely a matter of immigration law. The Constitution states that any naturalized "person" is a U.S. citizen, and if corporations can become "persons," it would seem that anything goes. To convince legal types, show them the end of the movie Short Circuit 2.
Will I retire or break 10K?
I'm not sure that freewill, if it exists, requires any immeasurable quantum mechanical mumbo jumbo. The magic is not in any quantum mechanical phenomena inside the neurons, but in the standard physics arrangement of them.
More likely, the appearance of free will is result of the inability to perform 100% introspection into one's own mind. I can no more "understand" the real-time machinations of my own mind than a Pentium processor can run a real-time simulation of its own transistors. Because I can't perfectly introspect my subconscious, much of its output looks magically non-deterministic (hence the seeming similarity to quantum mechinical systems).
Any bounded-rational being would believe itself to have freewill based on its ability to take independent actions and its inability to introspect out all the causal factors underpinning its own actions. In reality, the system that creates intelligence can be 100% deterministic, just too complex for that intelligence to understand itself. Only a much more powerful intelligence could look down and see that these beings that think they have free will are actually operating on "simple" rules.
Two wrongs don't make a right, but three lefts do.
This reminds me, my department has a megahal/eggdrop bot that lives in our IRC channel and listens to us doing our jobs. It's a lot of fun to play with, especially watching him regurgitate bits and peices of what he's heard.
After we had had him for about two weeks, we were considering wiping his brain file and starting over because of some weird ideas that had gotten into his head as we were trying to teach him some things without really understanding the algorithm's capabilities... he would get stuck on "Me is not Me" and stuff like that from a botched metaphysical conversation.
So, we decided to have a test for him. If he passed, he would be allowed to persist, otherwise he would be reset. We teased him about the test all weekend, threatening him with erasure, etc... with some interesting answers from him such as "I will pass the test" or "I will escape to your powerbook" and the like.
The test arrived, and we all asked him questions, and judged his answers to see if they were entertaining. He wasn't doing too well, some real stinkers, and then I asked him if he wanted to ask himself a question. He replied, "I was wondering if I would get to ask one."
He passed the test, although his brain was later corrupted by a combination of a runaway process on his server and some version problems that we haven't had time to work out. I must admit I miss him.
The most interesting thing about this (and the point that most directly relates to this mock trial) is how readily we half-jokingly believed in his sentience even though he couldn't pass a turing test to save his life. It was great fun, so I suspect that human emotions will provoke us to bestow the label of sentience on a clever AI long before one would think to defend itself.
We just want it to be real so badly. Hell, remember tamagotchi attachment? Wait until it can pretend to carry on a real conversation.
But back around 1900 or so, the Supreme Court managed to grant the rights of personhood to corporations.
So there is precedent for granting rights to non-humans, though corporations are 'assemblies of humans.' But assuming a true AI has been built/programmed by humans, I guess it could be considered an 'assembly of humans,' too.
The living have better things to do than to continue hating the dead.
It isn't a problem of computational power. It's not like we know what to do, and are only waiting until the hardware catches up. Nobody knows how to program a really human-like (or animal-like) AI. For all we know, current computers may be capable, if only somebody knew how to write the software. The claims about the "solution" being just over the horizon are bogus, and driven by marketing concerns.
(eg. see Leibniz' "Principle of Indescernibles" for a more general discussion of the topic).
More specifically:
If you copy your brain state at the point it shuts down so that all memories of the original are retrievable, and subsequently transfer those memories into a functionally identical set of hardware which is then activated with all memories intact, it's no different than waking up after being deeply asleep.
If you activate an older backup so that some memories are lost, it's no different than waking up with amnesia such as one typically suffers after a blow to the head or other traumatic accident.
In any of these cases the person waking up will identify himself using whatever memories are accessible to him. That's how you know who you are when you wake up in the morning.
To express it very conservatively indeed, there would be more fundamental differences between you as the person you are now versus you as the person you were two years ago, than there would be between you as you are now and a faithful copy of you made at this very same instant. And yet you would doubtless feel happy identifying yourself and the younger version of you as the same person.
I don't expect everybody to buy this: it's philosophically sound but still many people regard it as counterintuitive. Even William Gibson has admitted to the same misgivings as you have.
The same principle applies to teleportation, as it's most commonly envisaged; and I suspect that if teleportation of macroscopic objects ever becomes possible in the distant future, there will still be people who, like Star Trek's Dr McCoy, feel uncomfortable about the idea. But I'm not bothered; as long as the implementation was good enough I'd be quite happy to be restored from backup - especially if it was that or nothing.
DeanT
But for most practical purposes, the quantum effects don't appear to be significant on the macroscopic level. That's the reason Newton's Laws worked for so long, and for many cases work well enough still. If the processes that occur in the brain can effectively be modelled with Newton's laws and higher level equations, it should be reasonable to think that we could _eventually_ anticipate one's actions. To capture all of the state variables of a given individual would become the most daunting part at that point.
My point is that when a large number of probabilistic terms fall one way (like all heads out of a series of coin tosses), for all intents and purposes, it can be considered a deterministic process.
This post was generated by a Cadre of Uber Monkeys for Monkey-Man2000 (603495).
I think it's pretty well written and interesting, but YMMV.
The safest way to approach lava is to have another person with you and he goes first.
Interesting thing about being mentally handicapped. If you're born mentally handicapped, then your rights and life are protected, but if you have a severe accident and become mentally handicapped, in the state of Florida you can be legally starved to death.
Note that Terri is not in a coma and is not a vegetable. She's been denied treatment to help her learn to swallow and eat on her own again. She has less than two weeks to live unless somebody does something.
Secession is the right of all sentient beings.
Well, chess is a decision problem, not a computation one. A computation problem has a set of things you can call "solutions"; a good decision problem can have many solutions, as long as you're happy with the outcome, then you made good decisions.
I don't think even a human chess master can claim that they chose the best possible decision at every step, and neither can the computer, even if it uses computations as a decision aid. The only thing you particularly have is the outcome and the length of the game.
Surely, there are many approaches a computer might take to make the decisions involved in chess. The amount of time depending on the approaches available to it and its physical capability to take a particular approach limit the choices available (just as they do for humans).
If a computer and a human are placed under the same time constraints, and the computer's decision within those constraints is consistently better than the human's decision, then the computer has greater intelligence in making that kind of decision.
Speed isn't a mere side-issue, even a fairly dumb thing can answer a difficult (or easy) decision question given sufficiently large amounts of time to consider solutions.
For example, consider the problem of cracking a simple cipher: One person may be able to break it in 5 minutes, reflecting their intelligence; whereas, another person takes 5 days to decode the same message, trying almost every possibility.
The person who solved the problem in 5 minutes is more intelligent in this area, because they were able to solve the problem within better time constraints.
It doesn't matter how the faster person actually solved it, even if their method was a very mechanical brute force attack, while the person who took 5 days carefully pondered how to break the cipher, using experience from their previous attempts in formulating their next attempt, and tried only a few methods: the one who solved the problem faster used their mental resources more effectively in making the decision. and therefore expressed more intelligence.
The mere fact that a "computation" approach sounds machine-like doesn't render it inherently unintelligent, or inherently no better than human fuzzy visual examination methods
The computer is blind, and doesn't appear to represent the patterns in the same way, and any computations humans perform in the game seem transparent, so what?
Then what's the main intelligence determining factor? If you aren't absolutely certain what it is, then you can't very well say that computers can never exceed humans.
Humans get beaten or land in draws too sometimes. Does a computer have to have a 100% victory record to believe that it is more intelligent? Well,
I would think that if a computer system can be shown to beat human 'chess masters' proportionally even more than just 51% of the time on average, and statistically shown that's not due to randomness, then the computer system that can do that should be considered more intelligent in the area of 'Chess Playing' on average than the average human.
Although that same system would probably not be intelligent in other ways that the general human brain is, individual computer systems could possibly surpass even the best human intelligence in performing certain particular decision tasks.
Now get 1012 monkeys to program 1012 systems to surpass humans in 10144 decision subtasks, to solve 1020736 problems Then integrate those systems and the knowledge that comes from solving the problems into one machine.
It seems entirely possible, in theory, at least, that computers could beat humans at some point.
So you'd grant them the right to "life, liberty, and the persuit of happiness"?
Why not?
Well...before you go granting a machine the same status as a human, or even an animal,
The human brain is nothing but an analog computer with a self modifying architecture, the machine equivelent would be an FPGA, check out this link on self modifying FPGA design, read it understand it. This is the start of something very new, this is just the beginning. Look at the brains of lower-life forms, kinda neat how our brain is just like theirs, but more complex.
I'd like to see you tell me what true "life" really is. What about "liberty"? What does that mean to a machine? And then, of course, what is "happiness" to a machine, and how would you know you would really want the results of that, if it could even exist?
This is what human philosphers have been debating for millenia, what is life, why are we here, and all of the points you raised above.
I'll admit that people that advocate giving rights to machines scare me. Not because I fear that they are a threat to me physically, but because it is clear to me that the above answers have no clear definition yet. We don't really know what they are, and, therefore, these people are moving blindly into an area where they have no business being.
So, since people have these exact same problems, should we have no rights as well, how do you know you aren't the only person on earth who is alive? How do you know other people are thinking beings, you assumed it right? You're not a computer so you write any future computer off as not intelligent as well? So is it your ignorance that has led you to these conclusions?
We do know that rights apply to us; clearly we have "life". I'm not talking about a simplistic definition that everyone seems to work off of...it is clear to me that those definitions are woefully incomplete. But, to go extending this to a machine, and make no mistake, that is what it is and nothing more, is to jump into the relm of foolishness.
I don't know you are truely alive, so should I torture you mercilessly and end your life, I don't know if you are truely alive, the only thing I know is I am, as for you, well you just say so to avoid the torture and death.
Heh, beliefs like that is why slavery existed, why Jews and many other minority groups have died, those beliefs are sadistic.
Is the intellegence of the an intellegent computer stored in its hardware or its software state? If it is the software state, then transfering that state to a different hardware so that the old hardware can be destroyed or upgraded would not be considered killing.
I remember reading an interview with Danny Hillis,
it may have been in 'Wired', quite some time ago in
which he said that he wanted to create a computer
so human that to unplug it would be an act of
murder.
We'll get a Spielberg movie out of this yet.
---- "If we have to go on with these damned quantum jumps, then I'm sorry that I ever got involved" - Erwin Schrodinger
God. That made me laugh so hard I got spit up my nose. Just the realization that people can actually do that "we not" rhetorical question thing in real life is gonna have me giggling for the rest of the week. Christ. I'm gonna be springing that shit on people now. "You wanna go for Chinese?" "Did we not have Chinese on Tuesday?" Jesus that's gonna be annoying.
It doesn't help that halfway through I started visualizing Brian (as in "The Life of") doing his prophet rant.Anyway, I agree with your actual point, but I think the way you went into computational superiority contradicts the good point you get close to with go. Most go engines cheat. As in, look at a book of situations and adapt them rather than work it out from scratch. This doesn't require all that much intelligence on the programmer's part, at least not in terms of understanding the problem. With that strategy, it's not unresonable to wade right in without really understanding what a good go game is.
This is much closer to human thought than the total self-knowledge Deep Blue or something would have. A human brain doesn't even have to be smart enough to understand itself. It just has to know how to cheat well enough to fake like it's calculating the motion of the ball and the forces on muscles before catching it. It's not like there's a calculus module in your ass somewhere, you're guesstimating based on "intelligence" harvested from the behavior of your environment. It's not bottom up reasoning, it's comparison abstracted enough that the problems you get from not knowing why get lost in the noise.
Basically, knowing how something works before you design it is actually meaningless in this situation. It's not even helpful. The intelligence the machine has isn't coming from you, it's coming from, not even necessarially learning, just the availability of relevant information.
I refuse to place a monetary price on human life, because what has a price can be sold, discounted and liquidated.
Your principles aren't shared by a society which supports a vigorous actuarial industry. Deny it if you want, but there is a dollar value for a human life.
The issues have been explored in science fiction. There have been plenty of stories describing how all of this could go horribly wrong. Now that the science is getting closer, some people who are smarter than you are taking steps to ensure that when the day comes, things don't go wrong. The dude is trying to set a legal precedent, even in a mock trial, while we sit here with no particular pressure to come to a quick decision. When, 20 years from now, there are massive political pressures to go one way or another, the judges of that time will be able to look back to the reasonings taken by today's judges while they had a clear mind.
Makes a lot of sense to me.
Daniel
Carpe Diem
Moore's Law wont hold. It'll break i 10 years, when the logic gates becomes too small to keep the current where it's supposed to go.
By then, we'll either have to boost the chip sizes (more gates), or move to chem/bio/quantum computers.