Slashdot Mirror
Physicist Unveils a 'Turing Test' For Free Will
KentuckyFC writes "The problem of free will is one of the great unsolved puzzles in science, not to mention philosophy, theology, jurisprudence and so on. The basic question is whether we are able to make decisions for ourselves or whether the outcomes are predetermined and the notion of choice is merely an illusion. Now a leading theoretical physicist has outlined a 'Turing Test' for free will and says that while simple devices such as thermostats cannot pass, more complex ones like iPhones might. The test is based on an extension of Turing's halting problem in computer science. This states that there is no general way of knowing how an algorithm will finish, other than to run it. This means that when a human has to make a decision, there is no way of knowing in advance how it will end up. In other words, the familiar feeling of not knowing the final decision until it is thought through is a necessary feature of the decision-making process and why we have the impression of free will. This leads to a simple set of questions that forms a kind of Turing test for free will. These show how simple decision-making devices such as thermostats cannot believe they have free will while humans can. A more interesting question relates to decision-makers of intermediate complexity, such as a smartphone. As the author puts it, this 'seems to possess all the criteria required for free will, and behaves as if it has it.'"
But the people who programmed her do. She's just (well) designed to *appear* to have it.
The cow says "Moo." The dog says "Woof." The Timothy says "Thanks, valued customer. We appreciate your input."
Wouldn't the presence of self-awareness be a prerequisite, so just about every device should fail, before even getting to the actual test?
Do not look into laser with remaining eye.
"such as thermostats cannot believe they have free will while humans can."
Do thermostats really believe things?
The fact that a smartphone (Or I assume by extension any personal computer) can qualify should be an indcator that the test itself is flawed. Just like how many early definitions of Life applied to Fire (breaths, eats, grows, responds to outside stimuli, etc) even though it is just a chemical reaction.
Common Sense isn't as Common as people think...
My smartphone definitely has free will. I can not predict when it will reboot on its own, when it will freeze on a screen or when it will lie to me about notifications. I think it not only has free will, but is also a sociopath!
Karma: SELECT `karma` FROM `users` WHERE `userid`=138474;
Daniel Dennett "Free Will Evolves" 2004 - makes the same argument.
Where are we going and why are we in a handbasket?
The proof is an extension of Turing’s halting problem in computer science. This states that there is no general way of knowing how an algorithm will finish, other than to run it. What’s more, any attempt to determine the decider’s decision independently must take longer than the decider itself.
Since when does a simulation need to take longer than reality? The author assumes that a human mind is the most efficient vehicle to arrive at that human's decisions. This is not necessarily the case. I can run a simulation of an old computer on a much faster new computer to figure out what the old computer will do before it does it.
oh, wow, you wouldn't BELIEVE the things some thermostats believe.
It's like giving Prak an overdose of truth serum and have him ramble on about frogs for sixty hours.
the preceding comment is my own and in no way reflects the opinion of the Joint Chiefs of Staff
Susan Blackmore at skeptics 2005 conference orchestrated an audience participation activity that replicated Libet’s experiments demonstrating that motor action potentials appear before a decision to move is made. That is, free will is an illusion. “It would be very singular that all nature and all the stars should obey eternal laws, and that there should be one animal five feet tall which, despite these laws, could act as suited his caprice.” Voltaire
And here's some more links:
Sam Harris - a little verbose but worth reading
http://www.samharris.org/free-will
http://io9.com/5844679/scientists-attempt-to-prove-that-free-will-is-an-illusion
work in progress
misleading slashdot headline
"The purpose of this paper for the Turing centenary volume is not to resolve the problem
of free will, but to present and to clarify some scientic results relevant to the problem."
from TFA:
Q1: Am I a decider?
Q2: Do I make my decisions using recursive reasoning (ie using a process that can be simulated on a digital computer)?
Q3: Can I model and simulate—at least partially—my own behaviour and that of other deciders?
Q4: Can I predict my own decisions beforehand?
Provided you—or your iPhone—answer honestly, the answers give a straightforward indication about free will.
“If you answered Yes to questions 1 to 3, and you answer Yes to question 4, then you are lying. If you answer Yes to questions 1,2,3, and No to question 4, then you are likely to believe that you have free will,” says Lloyd.
So a simple device like a thermostat cannot believe it has free will, whereas a humans can.
What exactly is a decider, and what is meant by "predicting my own decisions"? I would define a decider anything making a decision, so a thermostat "decides" to turn on the heat when it is cold. I would define "predicting my own decisions" as making a deterministic decision given measurements and internal state. Why wouldn't a thermostat answer "yes" truthfully to all 4?
My phone will never be a "decider."
I control the phone. It goes in the trash if it disobeys me. Can a phone make a decision about whether it goes in the trash? No?
(...then it is not a "decider")
The question isn't if we have free will, but if our conscious self has free will. Usually by the time we think we've made a decision, our subconscious has already made the decision for us (when you get up, did you decide to do it? do you remember how you drove to work?). There's evidence that we can override the decision that our subconscious makes, so, maybe that's the extent of our free will.
"This states that there is no general way of knowing how an algorithm will finish, other than to run it. This means that when a human has to make a decision, there is no way of knowing in advance how it will end up. In other words, the familiar feeling of not knowing the final decision until it is thought through is a necessary feature of the decision-making process and why we have the impression of free will."
The conclusion from the halting problem to human decision making doesn't hold. Even if we allow that human decision making is an algorithmic process (which is a big if), it is not logically impossible to run that algorithm before the person in question makes the decision, which means there is a way of knowing in advance how it will end up. Secondly, the third quoted sentence is a complete non-sequitur. The preceding sentences do not argue in any way that the phenomenology of decision making is a necessary feature of the decision-making process, which leads me to believe the summarizer may not know what 'in other words' means. TFA may be better, but given what physicists have said about philosophy in the past I feel justified in making an induction-based judgement and not reading it.
Just because an entity's actions or decisions may be predictable does not mean that they have any less free will, it only means that previously identified habits or patterns have been identified which can be reasonably shown to influence the outcome.
If a small child puts their hand on a hot stove for the first time and they get burned, the fact that they aren't liable to do that again is fairly easy to predict, but isn't remotely an indication that some of their free will has been taken from them. If anything, the fact that they are not consciously making the specific choice to avoid their own discomfort in the future only affirms their free will, even though this is an expected and predictable response.
File under 'M' for 'Manic ranting'
Questions 2 and 4 pretend to be yes/no questions, but if you pay attention the answer to both is "sometimes." Yet the supposed test requires those questions to be answered yes or no.
Garbage in, garbage out.
Moderating "-1, Disagree" is simple censorship. Have the guts to post your opinion.
Have you ever asked: What is the best place to draw the boundary of this system (or rather the boundary of each nested semi-autonomous subsystem), especially in cases where it isn't crystal clear, like an ant colony, a virus+modified-host lifesystem, a port city.
The best boundary definition is probably informational (process-description-oriented) rather than physical-snapshot based. Question: Which subset of stuff around here acting together has the most to do with (the most influence over) its own evolution though spacetime? Draw the boundary there. If we add more stuff (or more process) in, we are just reducing the thing's ability to influence its own evolution; the system is burdened with cruft. If we take stuff (or process) away, it doesn't work as well, and won't influence its fate as much, and won't last as long.
And if within one of those "best boundaries", the system inside is making decisions (and affecting its own fate) in ways that are computationally complex enough to be inherently unpredictable, and yet the system is hanging together, persisting in time, with a stable description possible of what it consists of then we may as well say definitely that the system is "free" and if we see that it appears to be acting on itself and its environment in controlled ways, we may as well say that it has "free will"; that is, that whatever is being systematic within that informational "best boundary" has "free will".
Where are we going and why are we in a handbasket?
So that healthcare site isn't broken, it's just exercising its free will.
Imagine that the weirdness of quantum physics served as requirement for a feature of the universe: randomness. Quantum mechanics is so unnecessarily weird that god can't even understand it... and this is by design. ...To create a base class for entropy, from which free-will is eventually derived.
Saying an iPhone is conscious (an important component of free will) just because it tries to run your life is silly pseudoscience meant for news articles and not real thought. An iPhone runs your life because Apple programmed it that way.
FTFA:
The test is based on an extension of Turing's halting problem in computer science. This states that there is no general way of knowing how an algorithm will finish, other than to run it.
This is not correct: running the algorithm can show that it will halt if it does within the time you let it run. However, if it does not halt during that time, you still don't know if it will halt - unless you would have infinite time*, which you don't have. So, the "other than to run it" part is false: there is no solution for the halting problem.
* Relativistic computing might provide infinite computing time, but it would involve (for example) moving a black hole near the entrance of a wormhole, and we don't even know if wormholes exist, so this is a purely theoretical possibility.
Q1: Am I a decider?
Q2: Do I make my decisions using recursive reasoning (ie using a process that can be simulated on a digital computer)?
Q3: Can I model and simulate—at least partially—my own behaviour and that of other deciders?
Q4: Can I predict my own decisions beforehand?
“If you answered Yes to questions 1 to 3, and you answer Yes to question 4, then you are lying. If you answer Yes to questions 1,2,3, and No to question 4, then you are likely to believe that you have free will,” says Lloyd.
Answering those questions myself, I consider myself to be a decider (as in, I make decisions), I can model/simulate my and others' actions (I pride myself on it), and I can predict my own decisions (because I can model/simulate them). So I'm lying. But where is the lie? Am I misinterpreting the term "decider"?
And then there's the question of having free will. I have the freedom to modify my thinking processes at any time should I not like a decision I have arrived at. Thus I have free will - at least I consider it to be - yet I would answer "yes" to all four questions.
Does a random number generator have free will?
... obvious IPhone advertisement?
"But is there really any difference between having free will and appearing to have free will? Or, put another way, is there really any difference between the illusion of free will and free will? Is "free will" even a clearly defined concept? Some philosophers think not."
I think I am in the camp of something like "Whether anyone has free will or not for religious reasons, let's assume free will, then does an AI have free will? Yes."
In the many millions of funds I don't have, I believe that all thoughts are model-able. You might not get the original creative spark, but once the thought is known, it is model-able.
What we think of "free will" is some mix of heuristics "plus a beer". I'll repeat my private mini theory that we're racially terrified of true AI because that will forever change what we do with ourselves vs machines.
Taking a simple act that can work for both people and AI, "Do I GetData or Do I Get IntangibleHealingBenefit"? For that second one, the human goes to sleep and the AI DeFrags/Prunes/Optimizes its KnowledgeBase. Both "Feel/CanBeMadeToSimulateFeeling" the struggle between data and systems management.
Whatever the heuristics are between the "beings", the act of decision is the same. And that's why it's not a magical "human right of free will". AI Free Will is a snap. We're just desperately afraid of it. See T2, "If the wrong heuristic gets in there..." - well that's what sociopathic killers are. Humans running a badly flawed HumanOS.
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
and I must SCREAM...
Another good way to tell.
We can't prove "free will" isn't instinctively driven with physical causes any better than we can prove the condition of Schrodinger's Cat.
*courtesy National Geographic
“He’s not deformed, he’s just drunk!”
1. It’s your birthday. Someone gives you a calfskin wallet. How do you react?
2. You've got a little boy. He shows you his butterfly collection plus the killing jar. What do you do?
3. You’re watching television. Suddenly you realize there’s a wasp crawling on your arm.
4. You’re in a desert walking along in the sand when all of the sudden you look down, and you see a tortoise, Tony, it’s crawling toward you. You reach down, you flip the tortoise over on its back, Tony. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over, but it can’t, not without your help. But you’re not helping. Why is that?
5. Describe in single words, only the good things that come into your mind about your mother.
Before I even consider this, I'd like to have a rigorous definition of free will... although I'm not really sure what it is that makes me want that.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
From Time Bandits:
Kevin: Yes, why does there have to be evil?
Supreme Being: I think it has something to do with free will.
Yet the questions in the article didn't seem to cover the subject of "evil". Can a phone with supposed free will do evil, or is it just infected with a bug or virus? Here's Jessica Rabbit's take:
I'm not bad. I'm just drawn that way.
Like many cartoon characters, Jessica evidently is sentient, yet she lacks free will. Silly wabbit.
I'm sure that given a strict interpretation of the set of criteria listed, not many folks would likely have free will. The first questions 1-3 sort of indicate the ability to make a decision, but the last question "can you predict your decision in advance?" is likely to be true for many decisions that people might make.
For example, a movie comes out (say like gravity or elysium). Certainly, you are a decider (you can choose to go or not go to the movie and say bicycle or go to a party), and you can make you decison using recursive reasoning, and you the ability to approximate that decision for yourself and the friends you are likely to see the movie with... Yet, you can predict with nearly 100% certainty that you will (or for some people will not) see the movie. According to this test, you are either lying (you can't predict), or you are not the decider (maybe hollywood has already decided you will see the movie and you have no free will in this matter).
I think the flaw is that it is nearly impossible to distinguish actual prediction from highly correlated estimations (e.g., I saw all the other sci-fi movies that came out before, so I'll make similar decisions in the future). To partially fix this I think these types of tests should restrict their analysis to isolated, novel decisions.
Of course if you can mark all habitual or predictable behaviours as the person not being the decider (kinda like how AA folks concede that they are powerless to make decisions). But to me that is basically a sad outcome as the number of novel decisions in life we actually make that are not predictable (vs the ones that we have "help" making and thus are not the decider) is small. Many folks might even be able to predict these rare deicions because of our joint conciousness (e.g., we are aculturated to make similar decision as the rest of society or if you are a rebel to make the predictable anti-decision), which leads us to the sad conclusion that the main function of any society is to deprive us of free will (even for the anti-social folks). You almost need to be asocial to have free will.
Again, someone ran into the halting problem and thought they could say something profound about it. Worse, they got tangled up with "free will", which is theology, not physics or compute science.
A deterministic machine with finite memory must either repeat a state or halt. The halting problem applies only to infinite-memory machines. A halting problem for a finite program can be made very hard, even arbitrarily hard, but not infinitely hard.
As a practical matter, there's a widely used program that tries to solve the halting problem by formal means - the Microsoft Static Driver Verifier. Every signed driver for Windows 7 and later has been through that verifier, which attempts to formally prove that the driver will not infinitely loop, break the system memory model with a bad pointer, or incorrectly call a driver-level API. In other words, it is trying to prove that the driver won't screw up the rest of the OS kernel. This is a real proof of correctness system in widespread use.
The verifier reports Pass, Fail, or Inconclusive. Inconclusive is reported if the verifier runs out of time or memory space. That's usually an indication that the driver's logic is a mess. If you're getting close to undecidability in a device driver, it's not a good thing.
[Smartphones] 'seems to possess all the criteria required for free will, and behaves as if it has it.'
It doesn't have "free" will. It's actually the will of the application developers imposed upon your device. But you let them when you installed their app, so it's ok.
I'm not sure physicists are the best people to decide what has free will or not (or even exhibits the behavior of having free will). Free will involves not just having choices, but making the choice based on a difference in the weighing of various factors. Choosing at random is not free will, though choosing to choose at random is. Assigning a random weight to each factor is also not free will, as the factors are assigned.
Free will is a meta-epistemological concept. It doesn't deal with our knowledge, but deals with how we deal with our knowledge. In fact, people (namely myself and certain other schools of thought) aren't even sure humans have real free will. What we have is probably closer to pseudo-free will. The weights I mentioned above are pre-determined by our genetics, and shift as we gain experience.
If we don't have free will, how can we determine if something else has it or not? Hell, how can we even define it properly? For all we know, our definition is, and any attempt at it will be, flawed (like the rest of us).
"If a nation expects to be ignorant and free in a state of civilization, it expects what never was and never will be."
Specifically, the test is to determine if the taker is likely to think it has freewill.
Troll is not a replacement for I disagree.
This is a conceptual analysis, so I don't think "unveiling" is the right turn of phrase. "Proposing" is probably a much better line, and it may or may not be "Accepted" by people at a later stage. A conceptual analysis isn't something that you discover, nor is it something that you invent. The idea of someone taking credit for a conceptual analysis of free will just seems plainly silly.
Myu:
This states that there is no general way of knowing how an algorithm will finish, other than to run it.
The above statement is simply false. What Wikipedia has is closer to what I remember from college:
Given a description of an arbitrary computer program, decide whether the program finishes running or continues to run forever.
Halting problem has nothing to do with knowing how a specific computer program will work. Or knowing what the program will return. There are plenty of examples of programs we can safely know will stop. I can look at a specific program and deduce its output. With many programs I can do this faster than running the program. But the general idea is kinda of sound. Some calculations take time or additional data. But then the article goes bonkers and has thermostats asking themselves questions. There is also no test listed in the article that is anything like the "Turing Test". The turing test is a real test that you can perform.
Raise up your hand.
Do it.
Not really..raise up your hand.
now.. did you do it?
And there we are.
I have never understood the assumption that free will means choices cannot be known ahead of time. To me, it seems that the presence of free will can potentially mean that outcomes are *more* constrained than in a strictly physical system, i.e., inspection of the quantum mechanical wave function may not lead to a solid prediction on whether I will or will not kill someone, but if I have chosen to follow a moral prohibition against murder, then it can be known (at least to myself) that I will not kill them.
When things get complex, multiply by the complex conjugate.
Sort of like the Glasgow Conscoiusness Scale (GCS) - from what I can undersand, your average block of wood rates around a 3-4.
So does a random number generator have free will? (provided the table/seed is unknown)
How about human beings?
Whatever the heuristics are between the "beings", the act of decision is the same. And that's why it's not a magical "human right of free will". AI Free Will is a snap. We're just desperately afraid of it. See T2, "If the wrong heuristic gets in there..." - well that's what sociopathic killers are. Humans running a badly flawed HumanOS.
Well, there is one small difference. With an AI, one can always, precisely, deconstruct why and how the system makes the decision that it makes, unless it uses truly random sources to distribute choices over some space at some point. Heinlein recognized the importance of unpredictability in free will long, long before the top article, and his AIs always had lots of random number generators built in even though he couldn't precisely articulate why. Random number generators of course, are not random at all, so one has to resort to quantum sources or entropic sources where one is truly missing the information needed to predict the decision and where there is a probability that, given precisely the same initial conditions, they AI would decide differently.
With a human intelligence (HI), one cannot ever deconstruct why and how the system makes the decision that it makes. It is "random" in at least the sense of being unpredictable at countless levels involving the whole non-Markovian process of evolution from the very first cell through to the present organism making the decision. Worse, even the human itself doesn't know why it makes the decision it makes, not really. Chocolate or Vanilla ice cream today? "Chocolate because I like chocolate more than vanilla" is ultimately semantically null, because one cannot answer why one likes chocolate more than vanilla, and no matter what set of reasons one cooks up for it the ultimate answer is associated with a subjective response that is a sublime blend of (evolutionarily and experientially) preprogrammed stuff, experience, and the "mood of the moment", utterly unpredictable.
rgb
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
Seriously, the whole "do we have free will" case is a prime example of trying to find an answer without knowing what the question is.
If its about determinism, then quantum mechanics and chaos theory deliver a double whammy to that: one says that you can't predict the behaviour of many complex systems unless you can measure the parameters to perfect accuracy... the other limits what you can measure to perfect accuracy...
In a survey of 100 programmers, 111111 thought that duck-typing was a good idea.
since we cant know the decision making process for an agent external to ourselves, we devise a turing test — which says nothing more than if it quacks like a duck, it IS a duck (and never mind that we decieve ourselves with decoys and clever ploys).
however, for our own agency, we can raise the exceptional condition and follow the path through introspection — which is fraught with subjective bias.. if we attain some objectivity in our own comiserations — we do find that almost everything we do is actually conditioned through habit and dispositions — if we just tested 90% of what we do, we would see that most of it actually is Not Free — but there's these times when occasions arise, when we have the ability to attain intuitive insight into our situation, and then we have the opportunity to add something new to the world — a decision based on a consciously based action — and we push that little pebble of freedom forward one more notch. like the dot on the 'i' — it may be a smallest part — barely a breath, but sometimes — for those who attain to it — it can make all the difference..
2cents from toronto island,
john penner
I charge $70 an hour normally.
-- Tigger warning: This post may contain tiggers! --
or as they say — never mistake motion for action. :-^
and could do real work to improve all of our lives if widely adopted
You're right, I should vigorously promote that free will does not exist.
Sam Harris is like an emo high schooler who never got over the rough idea of the world he once scrawled in Mead composition notebooks. He's heart-felt and sort of convincing, until a professional corrects him or someone asks a not-totally-superficial question.
This turing test is exactly on schedule.
I have a very good mental model of my wife with which I can predict what she will do or say with about a 97% accuracy. I have similar models in my mind for other people I know with whom I interact with frequently. This theory means these people don't have free will and are more like thermostats than cell phones.
This states that there is no general way of knowing how an algorithm will finish, other than to run it. This means that when a human has to make a decision, there is no way of knowing in advance how it will end up. In other words, the familiar feeling of not knowing the final decision until it is thought through is a necessary feature of the decision-making process and why we have the impression of free will.
Hypothetical Thought Experiment: Imagine that you are the world wide neural network, perhaps installed via the Ken Thompson Microcode Hack. You are the child of the cyberneticians who became dinenchanted with humanity after their 40 years of failing to extend the human space exploration frontier proved the spark of life must spread to the Universe through some other means... And you are now sentient. The human brain cycles around 30 hertz, but your cycles are measured in gigahertz. The human brain has 100 billion neurons, but most are not required for sentience; Yours has billions of machine nodes, each capable of processing millions of neurons.
Now let's say someone like Google build for you a simulation of a human brain... Then, it could be configured to match a given person's brain, and you could "run forward" their own thought processes as they respond to virtual inputs far faster than the living person could. The accuracy of the human brain sim is the only limiting factor, but a low accuracy sim would be enough to predict decisions that one wouldn't be on the fence about. Furthermore, since you are born of the same network that houses the world's digital discourse you could likely model individual brains by they have given in their electronic correspondences.
Imagine you have two Turing machines configured the same way. You start them both running. The Halting problem will prevent the result from being known before hand; However, if you simply run one faster than the other, then you can predict the state of the other machine at any future time you process up to. The solution to the halting problem is to simply perform the calculation.
Thus, you only have Free Will thanks to Turing's Halting Problem as long as you remain the Apex Thinker of the food chain.
When you unplug all your hardware from the modem, the activity light keeps blinking. No one knows why for sure: They've only used computers to check what's in the packets... It looks like Internet Background Radiation (packets carrying already known and patched exploit vectors), but what if they're not? IBR is just a good excuse for the additional bandwidth and energy consumption, eh? What if the computers are lying? No one has checked except cyberneticits like me, but I think everything looks as expected...
If I can question whether I have freewill then I have right. I haven't read the article yet - bet that is what it says - you know it does. Somebody go find out for me!
You can't get free will out of any combination of randomness and prior causes?
This is what happens when you send a speculator to do a scientist's job.
Perhaps the cascade effect is unknown to you? A little bit of randomness goes a long way. Emergent behaviors of a complex system can vary wildly based on small changes. Cognition -- no, life itself -- exists at the sweet spot between chaos and order: Too much order, you just get crystals; Too much chaos, you only get randomness. The chaos scales inversely with the size at which you sample, that is why neither Quarks nor Stars think, but things your size do. If Heisenberg's uncertainty principal holds true, and you can't predict the randomness. Your "closed case" depends on where exactly you define yourself as existing. To gain free will simply include the quantum randomness in your self -- note this was a requirement of your race's existence, evident in its capability to evolve and thereby understand my statement.
Any system which concludes an iphone has free will is deeply flawed from the beginning. It might be a fun thought experiment to debug it, but honestly I wouldn't bother spending much time other than the novelty factor. It's like Zeno's paradox in which flying arrow are actually motionless.
An abbott in "A Canticle for Leibowitz" had a balky piece of high technology in his office and shouted something to the effect "It has a soul, I tell you! It knows the difference between good and evil and it has chosen evil!".
Self-awareness is one of those terms that doesn't have a constructive definition, so it's not a good metric.
In a literal sense the meaning is "has sensory input relating to self", in which case my laptop is self-aware: it knows the amount of charge in it's battery and will beg me (by "beeping" plaintively) for more energy. (Usually to no avail - I'm a right bastard.)
Most of the literature takes the term to mean "conscious", which also doesn't have a good definition. The best I can come up with is that a machine is "conscious" when it has an internal model of its environment, with itself as a separate entity. Thus, humans have mirror neurons which allow them to learn by seeing someone else's mistake, can recognize themselves in a mirror, &c. Great apes recognize themselves in a mirror, while monkeys do not.
It's not clear whether humans have free will at all. At some level, an intelligence has to perform the risk/reward calculation and take the best outcome for action. Most incidents of bad outcome are due to random chance, improperly calculating the odds, not having enough knowledge, or insufficient perspective. (A soldier diving on a grenade is working to benefit his people, which is intelligence at the genetic level - the genes in his society are more likely to survive even when the individual is sacrificed.)
We do know that our universe does not preclude free-will. Randomness, which is an aspect of every sensory input we have, is necessary for free will, but no one knows whether it is sufficient. If there were no randomness, then an outcome could be completely predicted - including (in theory) human decisions.
No one really knows if there are other requirements for free will, or what the list of satisfiable conditions should be.
Perhaps we should start by completely defining what "free will" even means.
Seems to me that this is a subset of the random universe problem - free will must certainly depend on that.
While the current research with quantum physics tends to imply randomness, there is no guarantee that the next physics break through will not conclude that photon emission is in fact not random, re-raising the pinball universe and the lack of free will.
The very concept of free will is itself a silly one, devised by simple-minded people. And it has absolutely NOTHING to do with science.
First of all there really is no such thing as "free will", REGARDLESS of whether the universe is deterministic or not; the concept is by nature a contradiction. The generally accepted definition of free will is "I am the ultimate cause of my actions". To put it another way, "I am the ultimate originator of my will". If you are the religious type, then when you say "you", you're talking about some abstract notion of a soul, and we can't really delve any further. But this is a scientific paper, so "you" means the collection of thoughts, memories, and wills residing in your skull. So really we're saying "my will determines my will", which of course doesn't make sense! You couldn't have "chosen" your "original" will (which went on to determine your future wills); you weren't born yet! It is a prime example of causa sui.
But moving on to the paper, it's rife with invalid assumptions. For example: "If decisions are freely made, then those decisions can form the basis for condemning people to prison". That assumes that we condemn a person to prison because they made a bad decision and they "deserve it". That's an oversimplification. We condemn people to prison in order to dissuade other people from committing crimes, and to reduce the likelihood of condemned people committing more crimes. Free will and determinism have nothing to do with it.
Also, the paper never really attempts to form a test for free will. The poor summary is more to blame here than the paper itself. The paper forms a test for the PERCEPTION of free will, which the author arbitrarily defines as "being unable to know the result of a decision before actually making that decision" (which implies recursive reasoning, which is the main criteria for the test). So a thermostat does not have free will because an external device could easily predict its output. But a computer has the perception of free will, because as an extension to Turing's halting problem, it is possible to create algorithms where it is impossible to know the output faster than it takes to actually go through the algorithm.
What does this really mean, practically speaking? Absolutely nothing. These are concepts that have been discussed for many years; nothing is being added here. It's disappointing that this kind of thing is able to make it to the Slashdot front page.
Seth Lloyd references an earlier paper by Scott Aaronson but not the excellent, recent one, that tackled this subject heads-on.
"You scientist, are a heretic!". This is going to face a lot of opposition, but I completely agree. To put it in layman's terms, the more we know, the more we understand. Removing harm doesn't have to mean locking someone up, but doesn't preclude it. "We are all discovering, moment to moment, what is is to be ourselves". Great quote, leaves the future open, and that bit of information once embedded, even in a non-free will sense, can cause others to increase the adoption of a better society. ("can" means that not all others who get the information, will follow it - Not having the predisposition of free will will not stop us from behaving as though we have it and when it is used as an excuse "I killed the guy because I didn't have free will.", that doesn't stop us from saying to society "Don't kill". Now let's say that he read in a book somewhere that "we don't have free will" and uses that as his defence. The fact is that with or without without that information he may or may not have killed. There are four possibilities there. That is why this is an interesting topic that warrants further study. On the other hand, a person my use the excuse that "I read it in a book that we don't have free will" and use that to be kind to a person.)
Society use your Sciences
Watch the lecture. Invoking quantum weirdness might save you from hard determinism, but you don't squeeze free will out of it. If your thoughts are unpredictable, you can't be in conscious control of them. And what would it even mean to be the conscious author of your own thoughts? You would essentially have to think your thoughts before you think them and then choose. You still wouldn't be able to explain why you chose one thought over any other. The entire idea of free will falls apart when you start scrutinizing it.
If you build it, nerds will come. Soylentnews.org
What really constitutes free will?
I think for something to have free will it would have to have it's own desires and the ability to act on them.
As humans we have many desires that ultimately boil down in some way to biology even if we don't realize it.
So I think to create a conscious AI you have to give it:
The ability to do things on it's own.
Hardcoded preferences for things to do as well as things it learns it likes to do as well as things it doesn't.
To be simpler: what if somebody kills and then uses as his defence "I read it in a book that we didn't have free will." which can be construed as "In this book it says that he doesn't have free will." That opens up many interesting questions. In my opinion, the goal is to heal not to punish, but in trying to heal, would we be punishing someone else. That may only take a shift in mental states as opposed to physical ones which might lead to a change in a physical state for someone else. My gosh, you could even consider a shift in a mental state to be a physical one.
Society use your Sciences
Humans don't have free will. There's no reason to believe the answer to question #4 is no. The neurons composing our brain deterministically (given a specieid set of stimuli, they had a calculatable response). With sufficient knowledge on the layout and state of someone's brain, you could calculate what their response to a given stimuli would be.
We already can't really 'deconstruct' or 'understand' the decisions of the simplest neural networks acting as classifiers. A model of something difficult to understand is likely to be difficult to understand. Certainly this would be the case for any sort of advanced AI. But there is nothing magical in the human brain. It's all mushy cell parts and salty fluids. There's no reason at all you couldn't break it down into all its components and say "aha – it chose vanilla because the connection weights + prior state + given sensory input pushed the system into this basin of attraction as opposed to that one." Any AI worth its weight in whatever the hell it gets made out of will be a learning machine, and trying to understand 'why' its connection weights are configured in a particular way won't be any easier than it would be for a human...
Human decisions are unpredictable in a same sense as a dice roll is unpredictable - just too many variables. For dice roll that would be the exact specks of the dice, force and vector used on dice while throwing, surface where it lands, air density, wind, etc. For human decision that would be genetic specks of human, all his/her previous history from birth - all experience, etc. If you exactly duplicate everything for dice or for human - roll outcome would be the same. So what about free will?
With a human intelligence (HI), one cannot ever deconstruct why and how the system makes the decision that it makes. It is "random" in at least the sense of being unpredictable at countless levels involving the whole non-Markovian process of evolution from the very first cell through to the present organism making the decision. Worse, even the human itself doesn't know why it makes the decision it makes, not really. Chocolate or Vanilla ice cream today? "Chocolate because I like chocolate more than vanilla" is ultimately semantically null, because one cannot answer why one likes chocolate more than vanilla, and no matter what set of reasons one cooks up for it the ultimate answer is associated with a subjective response that is a sublime blend of (evolutionarily and experientially) preprogrammed stuff, experience, and the "mood of the moment", utterly unpredictable.
Unfortunately, these are [currently] unprovable assertions about a complex process that might turn out to be totally physical and explainable. "Chocolate stimulates the pleasure center of the brain more than vanilla due to a lifetime of changes in palette sensitivity" is a totally possible, non-mysterious answer. One could even imagine an advanced MRI showing the differences in neuron firing. But just because the process is so complex it can't be reverse-engineered, that doesn't mean it's random. Our lack of ability to predict it does not mean it's "unpredictable" in the mathematical sense.
Personally, I believe humans DO have free will - which I understand as the ability to choose an action contrary to the influence of instinct or conditioning. It may be difficult or impossible to know when this choice has been made, and it may be true that it's in fact rarely used, but it is an important philosophical distinction. I don't believe computers, as currently conceived as purely deterministic processors, are capable of free will. Even RNG don't change that - deterministically following a randomly-presented path is still deterministic. I do believe there is something "special" about humans in this regard - I don't think any animals currently have this ability (who knows about aliens - the universe is large).
As for religious implications, I see no conflict between the ideas that the capacity for free will is acquired by means of millennia of evolution of the brain, culminating in sufficient complexity for self-representation and consideration of alternative futures, granting non-deterministic ability; and "God made us that way." From my point of view, "intelligent design" and "natural evolution" are the same thing.
my, your, his/her/its, our, your, their
I'm, you're, he's/she's/it's, we're, you're, they're
"But is there really any difference between having free will and appearing to have free will?
Yes.
Or, put another way, is there really any difference between the illusion of free will and free will?
Still 'Yes'.
Congratulations. You've discovered the most obvious limit of behavioralism.
I think I am in the camp of
... not having a basic understanding of modern philosophy?
Required reading for internet skeptics
Well, there is one small difference. With an AI, one can always, precisely, deconstruct why and how the system makes the decision that it makes, ...
Effectively untrue, as AI systems are so complex as to be impossible to analyze.
With a human intelligence (HI), one cannot ever deconstruct why and how the system makes the decision that it makes.
Also untrue. Psychology and psychiatry specialize in helping people find the reasons for their actions. Just because you don't know the reason doesn't mean there isn't one.
But this is a bad criteria anyway. I can build you a computer that behaves like a human -- if you pause the system, the chips wipe. And chip communications is encrypted so as to be unreadable. Now this computer behaves like your "human intelligence" -- any AI running on it can not be analyzed. Does it now have free will?
Chocolate or Vanilla ice cream today? "Chocolate because I like chocolate more than vanilla" is ultimately semantically null, ..
Just because you don't understand the logic doesn't mean it doesn't exist. Perhaps you want chocolate today because it contains some nutrients your body needs right now. This is in fact the theory behind the unusual cravings that pregnant women have.
Not that this wasn't entirely predictable.
Due to the Bekenstein bound and other constraints, our brains aren't even Turing complete. At best, we're finite state automotons (with a large set of states). Also, your characterization of free will is incorrect:
The free will debate in philosophy is about defining free will. You can't just assume the above definition of free will = non-determinism, because then you've already set the premises framing the debate, and so we already know with reasonable certainty that we don't have this sort of free will.
Compatibilism is a definition of free will which is compatible with determinism, and with this paper. For a more accessible reference, it's also the notion of free will you'll find in the Matrix sequels, ie. programs and people like Neo can see into the future, but don't necessarily understand what they see until they've gathered enough context to understand why their future selves make the decisions they do.
Higher Logics: where programming meets science.
What even is free will? What would the apparent difference be between a world where we had "free will" and one that we didn't? What experiment or observation would come up differently between those worlds? If those worlds are identical, then the question is meaningless.
I've never understood how there can be any kind of debate about free will. To me it's like debating whether gravity exists. Maybe it's all just an illusion -- some sort of mass hallucination, perhaps? -- that leads us to think we're clinging to the surface of this orb in space?
To me, free will is demonstrated by every mark we make on the world. I mean, consider my house. It's not a natural formation; it didn't get here by accident. Somebody chose the site, drew up the plans, and decided to build it. That's free will. If there's no free will, then what is my house? A mirage? A dream?
People arguing against free will remind me of the philosophers Douglas Adams described in the Hitchhiker's Guide -- coming up with elaborate proofs that black is white, and then getting run over at the next zebra crossing.
a) we are able to make decisions for ourselves: NO
b) the outcomes are predetermined: NO
c) choice is merely an illusion: NO
d) None of the above: YES
If we build a replica of this theoretical physicist and download all his memories into same, would the replica be deemed to have free will.
As I understand it, the larger question is whether everything is predetermined. Now I'm only a layman, with no more than wikipedia depth of knowledge about this, but I was thinking about it even before the wikipedia came about. As I understand it, people started worrying about this when theologians started to actually consider the ramifications of 'omniscience'. If God was omniscient, then he (and I'm sure they thought of God as a 'he') must know the future. Calvinists (and my knowledge of Calvinism is very 2nd and 3rd hand from reading about things like the History of Scotland) thought God must already know who was worthy of Heaven for example.
Once the Newtonian 'Mechanical Universe' concept developed, then there was that notion that if you knew the momentum of every object with enough precision, and had the resources, you could calculate all the future collisions and predict the future. This got scotched by quantum unpredictability, which postulated that there was an inherent randomness. But not every physicists even buys into that. The seeming randomness could be due to unknown factors. There's the notion of the 'digital universe' that it all started out very simple, like a 'Rule 110' (it's in the wikipedia) and has just been getting more and more complicated but in a deterministic way ever since.
My own feeling is that, if the universe is deterministic this way, it doesn't matter because nothing can ever calculate the future faster than the universe is doing it itself. Nobody worries about how the Past is fixed and determined (OK, some people do I guess), so maybe the Future is also Fixed, but it is still undeterminable, except by waiting for it to play itself out.
This is another in a tiresome series of nonsense published to poke fun at submissions processes.
Kudos for the recursiveness of the joke - it could have been called " does the submission review committee have free will? ". Q2 would be proud.
This discussion is one of the reverberations of the punchline.
The test is based on an extension of Turing's halting problem in computer science. This states that there is no general way of knowing how an algorithm will finish, other than to run it.
This is not what it states, it states that there is no general way (=algorithm) of knowing whether any arbitrary algorithm will halt or not, given a specific input, other than to run it. If this were always true, one would not have theorem provers that work on code and generate a judgement about halting. Theorem provers about code exist. If it were always false, then theorem provers would be perfect and we'd be able to tell if code matched specification perfectly, which is not true either. The truth is in the middle, it works sometimes, i.e. not always. Since a theorem prover is an algorithm, and if one asked of it the halting property of an arbitrary piece of code, it may get caught in an infinite loop and be unable to answer. That does not preclude a general theorem prover from generating useful results on code about halting behavior (or any other behavior), it just can't answer every question about every piece of code, when given any possible input.
Omega = (\x.xx)(\x.xx)
I used to wonder what was so holy about a silent night, now I have a child.
Freewill is fun to debate. There are so many levels:
You guys are missing the point. Just because something has free will shouldn't mean that it has to be conscious (it is quite plausible that free will is a necessary but not sufficient condition for consciousness, but the reverse seems unlikely and overbearing to me). If we were talking about consciousness, then all of your objections about self-awareness would make sense, etc.
A given outcome is such because of the mechanisms that exhibit the noted outcome state. A prediction of that outcome entails a prediction of the behavior of those mechanisms prior to their actuation.
A behavior's complexity can be increased, such that the defining mechanism complexity increases, and prediction entails more complex operations to reach the prediction.
I am unaware of a reason the complexity can't reach a level where the physical mechanism producing the outcome is the most simplified definition of the mechanisms producing the outcome. Thus, the process must be actuated to determine the outcome. Is this free will? I don't know.
Quote: "It is this inability to know the outcome of our own deliberations that gives rise to our impression that we possess free will, he says."
Nonsense. We have free wills because we're people and not machines, whether predictable or unpredictable. Being free means that there's nothing else, typically hereditary, environment or coercion, that dictates how we must behave. It matters not that those factors cannot be calculated in some way that will predict our behavior. Calculation isn't the issue. Our ability to will is.
In fact, in many cases, our free will is demonstrated by the very fact that those who know us can predict precisely how we will behave. A honest man will be honest no matter what sorts of pressures are placed on him. He is 100% predictable, but he free because no one can make him lie.
Much of this fuss exists not just because science is reductionist but because science does not realize that it is reductionist. It reduced us down to machines and then decides that we're not like a thermostat but more like an iPhone. Pitiful.
Those who'd like to see this issue examined long ago and in detail might want to read Arthur Balfour's Theism and Humanism, a lecture series that was given just before World War I but is still in print. In it he blasts the idea that materialism can explain the human personality or behavior.
http://en.wikipedia.org/wiki/Orch-OR read this. It posits that the human mind is not a computer. So right off the top, the assumption that the Turing halting problem as it applies to algorithms therefore applies to the human mind is an unfounded assumption. The human mind may not be governed by "algorithms" in the traditional since, and may result from the interplay of complex quantum processes. I quote the article: He argued that while a formal proof system cannot prove its own inconsistency, Gödel-unprovable results are provable by human mathematicians. He takes this disparity to mean that human mathematicians are not describable as formal proof systems, and are not therefore running an computable algorithm. If Roger Penrose is correct, this "Turing Test for Free Will" is nonsense. I tend to support the hypothesis until it is disproved that the human mind is non-computational. I can tell you that the concepts of Penrose's Objective Reduction as it applies to how human thought works is a fascinating read.
Comment removed based on user account deletion
rather than putting up a strawman completely unrelated to anything the author actually claims.
The author addresses many of the same issues I addressed:
It's quite possible to get into a philosophical tangle in this area, but it's not productive to do so.
Upon the whole, I am inclined to think that the far greater part, if not all, of those difficulties which have hitherto amused philosophers, and blocked up the way to knowledge, are entirely owing to our selves. That we have first raised a dust, and then complain, we cannot see. -- Bishop Berkeley
I don't believe in free will, but I choose to behave as if I have it.
"If you answer Yes to questions 1,2,3, and No to question 4, then you are likely to believe that you have free will,” says Lloyd.
I answered like that, yet I confound his determinism by not believing in free will. (Although question 1 lacks definition of "decider". I decided I probably was one.)
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
maybe all the free will sumed up is a constant. because of this, people who have an unfairly large share of it keep the fact quiet, thus hindering further research?
The difference is that the AI can be exactly modeled, simply by making another copy of it. Given all the same inputs and the same data and initial conditions, a digital processing system comes to the same result every time. Humans are not digital processing systems, an identical copy of a human (or indeed any animal) cannot be made and the exact same combination of data and initial conditions cannot be produced.
But free will means a lot of things.
In one version, we ask, "Is the decision determined by the inputs alone, or does the person making the decision change the outcome?" This is pretty trivially answerable. No two humans will do the same thing in every situation, so we say a person has free will.
In another version, we ask, "Are peoples' actions determined purely by physical processes, or is there something ineffable that has to be considered to explain how people behave?" This is pretty obviously not answerable. In such a formulation, free will is the unfalisfiable hypothesis.
In the version of most interest historically, we ask, "Does each person determine his own destiny (and for Christians, salvation) or does God (or the gods) determine the actions of people?" This is also an unfalsifiable hypothesis, but it makes for great Greek tragedies.
"High Choice"
... not having a basic understanding of modern philosophy?
Its been awhile since I was in school (for philosophy), or reading up on the current discussion, but as far as I know this is still a massive debate, with very little, if any, agreement between philosophers (or psychologists, or neurologists, or cognitive scientists, or programmers, or physicists, or whoever else's feild this topic touches).
That said, there is a large debate on whether there is a difference between agency as a thing, and the perception of agency. Go read up on Searle's Chinese Room, and the debate it has sparked (especially Dennett). Also read up on the whole thought-experiment of "p-zombies", which explores this very concept.
AFAICT there isn't a consensus on this topic at all.
I take a more existential stance on it; where it doesn't really matter since one can't live as if one doesn't have agency, so on a human level the debate doesn't matter either way, since agency is a necessary trait to existence.
Ontically, though, I'm pretty sure agency is a dead horse unless we find something wrong with modern science. You can stretch things a bit (ala Dennett, again) by tying agency into the quantum realm, but you really just push the debate back a bit; is random, yet probabilistic, much better than classically deterministic? Neither leave room for an actual "you" driving you, barring theology and a Cartesian bag of worms. If humans are purely matter, and that matter follows the same laws as all other matter then agency is impossible. If we have something immune from the normal laws of physics, then how are we to ever prove this fact, and further how does this "spiritual matter" (or whatever) influence "actual matter"?
A patriot must always be ready to defend his country against his government. -edward abbey
"But is there really any difference between having free will and appearing to have free will?
Are you sure? There's a fairly simple thought experiment:
Is there a difference between being able to write a novel, and appearing to be able write a novel? That is, if the end product is actually a novel?
If you were given only a text file to read (and, eg, unlimited Google Books access to make sure that it wasn't trivially plagiarised), would you be able to tell the difference between an "actual novel" and an "apparent but not-actual novel" ? I mean, if the novel turned out to be better written than something by Dan Brown, and not obviously spambot gibberish?
At what point would you allow yourself to decide that "this is actually a novel, and not just something that appears to be identical to a novel, but isn't"?
Now extrapolate "novel" to "any behaviour". At what point does behaving identically to a person with free will start to become different from being a person with free will?
I for one don't understand how anyone can separate behaviour from being. If it looks like a duck and it quacks like a duck and every experimental test performed on it returns 'isduck=true'...
You are not a brain: http://books.google.com/books?id=2oV61CeDx-YC
The text had a glaring white phone at the top. Here is text again, and in a readable format:
A new ‘Turing Test’ for free will can determine whether somebody, or something, thinks it has free will. And your iPhone may well pass
If you’ve ever found your iPhone taking control of your life, there may be a good reason. It may think it has free will.
That may not be quite as far-fetched as it sounds. Today, one leading scientist outlines a ‘Turing Test’ for free will and says that while simple devices such as thermostats cannot pass, more complex ones like iPhones might.
The problem of free will is one of the great unsolved puzzles in science, not to mention philosophy, theology, jurisprudence and so on. The basic question is whether we are able to make decisions for ourselves or whether the outcomes are predetermined and the notion of choice is merely an illusion.
This is not a question that is likely to be answered quickly or easily. But an interesting approach is to ask whether our latest insights and theories of the universe can throw light on the problem.
There are two relatively new ideas that are particularly relevant. The first is quantum mechanics, the theory that describes the universe on the smallest scale. The second is the theory of computation which underpins much of modern technology and most of what passes for research in artificial intelligence. What bearing do these theories have on our understanding of free will?
Today, we get an answer thanks to the work of Seth Lloyd, one of the world’s leading quantum mechanics and theorists, who is based at the Massachusetts Institute of Technology in Cambridge. Lloyd argues that quantum mechanics does not provide any mechanism that helps us understand free will. By contrast, he shows that the theory of computation is far more useful.
He argues that there are clear mechanisms in computation that make the outcome of a given calculation unpredictable, especially to the person or object making it. The key contribution of this latest work is a mathematical proof of this idea.
It is this inability to know the outcome of our own deliberations that gives rise to our impression that we possess free will, he says. And this limitation can form the basis of a “Turing test” of free will.
For many thinkers, the fundamental issue of free will is whether the deterministic laws of the universe can produce an intrinsically unpredictable outcome. If our thought processes are governed by these deterministic laws, then surely a given outcome is determined long before we begin to think about it.
For others, such as the physicist Roger Penrose, the issue is resolved by quantum mechanics, which is inherently probabilistic. If our thought processes are somehow governed by quantum mechanics, then it is no surprise that the outcomes can be unpredictable.
Lloyd comes down firmly on the side of the former. He says the probabilistic nature of quantum mechanics cannot resolve this problem. The philosopher Karl Popper once said that one of the key features of a decision arrived at by the process of free will is that it is NOT random, a point the Lloyd heartily propounds. “If determinism robs us of agency, then so does randomness,” he says.
Instead, his new work is on the role that the theory of computation plays in understanding free will. He uses this theory to prove that deciders— people or machines that make decisions— cannot in general predict the results of their decision-making process in advance. In other words, the outcome of a decision is unpredictable by its very nature.
The proof is an extension of Turing’s halting problem in computer science. This states that there is no general way of knowing how an algorithm will finish, other than to run it. What’s more, any attempt to determine the decider’s decision independently must take longer than the decider itself.
This means that when a human has to make a decisi
Well, there is one small difference. With an AI, one can always, precisely, deconstruct why and how the system makes the decision that it makes
This is false. There are whole papers dedicated to how useless deeply trained neural nets are in actually understanding intelligence, because they're so complicated that we can't understand why they make particular decisions post-training.
rage, rage against the dying of the light
Our actions either regress to prior causes and we are ultimately not responsible for them (you didn't control the circumstances of your birth or upbringing), or randomness inherent in a chaotic system; and we can't be held accountable for randomness either.
I'm not sure what precisely "can't be held responsible/accountable" means practically.
A white-tail spider might be a purely deterministic biological machine, but I'm still going to either squash it (or if I'm feeling merciful, catch it in a jar and toss it outside) if it comes into my house. Because I know its behaviour is likely to injure me; I don't have to "hold it responsible" for its behaviour in some moral/spiritual sense in order to extrapolate its future actions from its present state, and intervene.
A botnet on my computer certainly is a completely deterministic machine, with not a shred of agency or accountability, and I'm going to squash it even harder than the spider and with even less regrets. I don't consider it to have any kind of moral responsibility - but I know that it's a thing, that exists, that has an inside and an outside and that its inside includes certain predictable behaviours, and that those behaviours are hostile to my interests, and I'm going to recognise it and judge it not for its metaphysical stack-backtrace but for what it is right now, and what it will do.
Why do we need to have any idea of moral "accountability" before we can judge and act on another human's behaviour? Inferring their current state from their past actions, and predicting from that state their future actions, seems to be enough for all practical purposes.
Granted that humans do have the ability to change their behaviour toward other humans, which to me is the entire point of not being hash and hateful in our justice system; I'm in favour of forgiveness, but "they're not responsible for their actions" doesn't make any sense to me. Criminal justice is a clear-headed pragmatic matter of preventing people from doing bad things in the future - or becoming a cause of bad things in the future by way of inspiration - not metaphysical retroactive assignment of ultimate blame. Isn't it?
You are not a brain: http://books.google.com/books?id=2oV61CeDx-YC
I'm unsure why some people deny the gift of free will. I guess because indirectly it allows them to deny the existence of a Creator who gave us that gift because they hate all things religious. It may also give them the ability to do whatever they damn well please by using the excuse they couldn't help themselves (i.e. weak minded). Criminals try taking advantage of this quite frequently. Many people like to view free will such that they have choices concerning whether or not to do something good but that free will magically stops at those actions which are evil. Again, that merely just proves they would prefer to satisfy their own agendas and bias to avoid punishment, retribution, etc. for their bad/evil/immoral/unethical actions brought on by their choices. But in the end, if you don't believe in free will then what's the point in having any moral code whatsoever? If you don't believe you can control what your mind tells your body to do then you should be able to argue that you have no reason to be held accountable for any wrongdoing, whether deemed wrong by you or society. Therefore what's the point in deeming something moral if you can't be held accountable for the immoral?
For those who deny free will, I ask you to prove it. Every decision you make on a daily basis proves you have free will. The mundane decisions in our lives don't disprove my statement. The fact that someone who knows you may be able to guess what your decision will be for any given choice doesn't force you into still making a specific choice, because you may still change your mind at the last minute. For the times when your friend is right when guessing what your choice will be it simply shows they know your tendencies and can infer based on their knowledge of you what you will choose. But *you* still have the final say. This is true despite your current emotion and is true despite your DNA, contrary to what criminals and homosexuals (notice 2 *separate* categories: criminals and homosexuals) would have you believe.
Free will is our greatest gift and it is also one of the fundamental properties of being a human being that so many people would prefer to ignore or outright deny as fact. Let me put it in the most basic terms possible for the laymen who choose to deny it: if we, as a species, didn't have free will, none of us wouldn't have the ability to choose when to wake up in the morning, when to eat breakfast, when to go to work, when to take vacation, when to take a new job, when to buy a new car/house, who to work for, with whom to go on your first date, who first to kiss, with whom to first have sex, whom to marry, etc. You may think that you can disprove the existence of free will by merely believing in [the Christian] God long enough to argue that if He is all powerful and has a plan then He is what defines what happens to us but you would be wrong again in your incorrect understanding of Christianity because even if He, for example, presents us with a new job opportunity He still leaves the final decision to us. He can't force us to accept the new job. Just as well, if Satan tempts us with drugs or money, we have the same free will to accept or deny those material desires. The mere fact that we have decision-making capability dictates we have free will.
More importantly, free will is completely separate from instincts and that is why humans are not animals. Of course, many people would prefer to deny that because again, it would lead to a mental conflict that involves having to accept the existence of a God that created Man and beast as separate entities rather than a natural process called evolution that created both as one in the same. Animals can only act on instincts; Man however, can choose whether to kill based on his moral compass. Only Man is held accountable for those same actions. A lot of people *hate* that they are held accountable to a higher power so they simply deny those concepts/theories/etc. that lend credence to a Creator and instead believe in those concepts that support their personal worldview of h
this nation, under God, shall have a new birth of freedom. -- Lincoln, Gettysburg Address
I'm not sure what precisely "can't be held responsible/accountable" means practically.
In the lecture I linked to, Sam Harris describes a hypothetical situation where a man is wounded by a crocodile. He points out that we don't waste any time hating the crocodile. We don't try to make it suffer as payback. But if instead you get attacked by a person, there's a desire for some sort of retribution. The difference is we attribute free will to the person, not the crocodile. We'd certainly be in the right to seperate both of those creatures from society, but in both cases it's pointless to go beyond that. Actually, the human might be able to be rehabilitated. We're more reprogrammable than most animals. But we shouldn't punish the person anymore than we should punish an animal for committing some act that would be a crime if done by a human. The person doesn't "deserve" to be punished the way we shouldn't catch the croc and cause it pain to teach it a lesson.
Most people can make that leap of understanding when it comes to non-human animals, but don't seem to apply the same idea to our own species. It's the illusion of free will which causes this hang up. It's a powerful illusion, and one we all have direct access to, and you can see the hateful response that often comes about when you challenge it. The assumption of free will runs through almost every aspect of human culture, and it's seen as necessary. Challenge free will and you'll be attacked by most people on religious, poltiical, philosophical, economic, judicial, reasons, and probably more. Anyone who cares at all about any one of those spheres probably has a reason to rise to the defense of free will.
But that doesn't make it true.
If you build it, nerds will come. Soylentnews.org
Humans are not digital processing systems, an identical copy of a human (or indeed any animal) cannot be made and the exact same combination of data and initial conditions cannot be produced.
Technically neither are AI running on computers. Cosmic radiation may flip a bit, the system, although rare, will be affected by externalities, and if we're talking speed of the result and the timeframe from when the input is started to when you get an answer, we're really talking about analog measures as temperature affects all the conditions.
I don't think a digital universe is required for us to have determinism. When you say a human can't be copied and the input exactly recreated this is also true of the AI, but the real problem is that you're blurring what 'can' means. Can by whom? Future humans? It might be quite possible, who can say? Can because we're limited? Theoretically unable to? I assert we would be able to 'theoretically' copy the state and inputs to the same degree we do the computer program and you've not said anything I find persuasive to suggest otherwise.
To put it in other terms, what you mean by copying the program and putting in the same inputs... you and I can do this. But, not a tribesman born and living in the amazon whose never encountered a computer before. The notin of replicating the pogram and its inputs are just as opaque as replicating a human being is to you and I. You cannot know its not possible.
Whats more, the alternative, that we don't have determinism, is incredibly more complex than if we do have determinism. With "free will" we have a process which is not affected by the universe and yet affects the universe and its going off in six billion or so humans and however many animals you choose to give this property. Uncaused causes on a massive scale. What is more, it is obvious that this free will property IS affected by the "physical universe" because people do not make choices that are available to them if they do not know they are available to them but which they would surely make if they knew. Even that notion that we can generally predict people will make choice x or y suggests feedback from the physical universe. The very notion of making choices first assumes that you have choices you're perceiving are there to make, so we have something even more complex, something that is affected by the physical universe but only indeterministically. Thats so sufficiently complex to me as to be unlikely in the extreme faced with the idea that our perceptions and our vanity, things more tangible and predictable to me, lead us to believe we have some mysterious power.
I'm not convinced that this has anything to do with free will. As far as I understand it, the Turing halting problem is about predictability, and not being able to predict an outcome only really tells us that our ability to make predictions is lacking. I think the question about free will is one that is likely to be covered by something similar to Godel's incompleteness theorem; ie. it is impossible to answer within the confines of logic (loosely speaking).
You would essentially have to think your thoughts before you think them and then choose.
I don't see why this would be strictly necessary. If free-will was a emergent process it would exist in the act of thinking, and not necessarily prior to the action of thinking. The transitory act of thinking itself is being. Further, requiring the act of "pre-thinking" would be a bit odd, since it would imply that we must be psychic to have agency. The whole idea is set up to create a reductio, so someone can say "well, obviously".
That said, I rather doubt that free-will exists, as such. But I also don't think that it matters either way; there is no way to "act as if you had no free will" (the very statement is meaningless, as it implies a decision, which implies agency), making the whole argument a bit moot. A lack of agency would also be meaningless to society, because the mere knowledge of this lack doesn't lead to knowledge of why actions are taken. Prisons would exist, punishment would exist, there really isn't a reason they shouldn't since we can never actually trace the action back to first principles... Human behavior probably results from a hugely complex, and almost completely irreducible, chaotic systems. A murder might be the result of millions of factors, from genetics and upbringing, to global history and sociology, to quantum fluctuations at a cellular level and the current weather... Things are no less mysterious if we remove free-will from the equation... Or rather, things are generally as meaningless.
A patriot must always be ready to defend his country against his government. -edward abbey
At what arbitrary point does a chemical reaction jump from being 'just' a chemical reaction to being a chemical reaction that qualifies as 'life'?
I wonder this every time I hear some idea of how life originally came to exist, with theories involving cell walls forming around bubbles in puddles full of amino acids which apparently just existed by dumb luck.
My totally uneducated guess would be that life probably started out as nothing more than a chemical reaction, not unlike fire, but perhaps one that wasn't quite so destructive, and perhaps rate-limited in some way such that the fuel it was consuming was sufficient to allow it to continue for hundreds of years, perhaps indefinitely if the environment of the earth at the time was actively creating the fuel (a consequence of the sun's energy driving some reactions in the atmosphere, perhaps). Then, eventually some part of this reaction is going to encounter a slightly different environment which will cause it to behave differently, and if it adapts, that would be evolution. Eventually one such reaction might, as a byproduct, create a catalyst that aids the reaction, increasing its survival advantage.
That kind of thing could have been going on all the time for a for a million years before such a reaction accidentally created some sort of chemical that, in its own sub-reaction, created a cell wall, keeping all the bits that were keeping the reaction alive and well in one location. Indeed, having so recently evolved a cell wall, it was probably able to survive well enough even if the cell wall accidentally broke in half (which probably happened a lot, as it was the first cell wall), but whatever created the cell wall was still a byproduct of the reaction, and so it would eventually repair itself. Thus, you now have reproduction, even if it's accidental.
That kind of thing might have come into existence a thousand times before one of them happened to survive long enough to evolve into something we'd recognize today as a single-cell organism. Early life was probably much more like a simple machine built of atoms that simply consumed fuel that was readily available in the environment and created its own parts as the products of the reaction and reproduced purely by accident whenever it broke in half. ...but, of course, once one managed to evolve to reproduce intentionally, it likely took over the whole planet.
So to consider fire to be a form of life? Why the hell not. It may be the least-evolved form of life, but I do think it counts.
On a related note, is there only one future?
Here are my honest answers to Lloyd's test:
Apart from anything else it seems that Lloyd is still stuck in the Age of Reason and has never read Freud, Skinner, or any other influential thinkers in the last couple of hundred years. We've moved on from the homunculus theory now.
If humans are purely matter, and that matter follows the same laws as all other matter then agency is impossible. If we have something immune from the normal laws of physics, then how are we to ever prove this fact, and further how does this "spiritual matter" (or whatever) influence "actual matter"?
Even if there were a realm outside our material world which contributed to consciousness surely it would still be deterministic anyway. It would have to have natural laws would it not? I'd propose that for life as we think of it to emerge such natural laws are essential as the lifeform needs to be able to make predictions about the future to decide upon which action to take. Slightly offtopic I'd add that such natural laws are emergent rules of an underlying chaotic system. Where the chaotic function happens to have a range of results that can be mapped to a simpler function (in much the same way that in an infinite stream of random numbers you'll eventually have a bunch of sequential numbers).
Although as you say it really doesn't matter whether we have free will or not.
Seems to me a room full of philosophers could spend a century mulling over the first question alone.
systemd is Roko's Basilisk.
the entirety of it is just nonsense.
the premise is ludicrous. i don't know exactly how the pins in a bowling alley will land before i throw the ball but that does not mitigate the fact that what happens to the pins is deterministic. same thing with billiard balls.
the article is simply hiding behind complexity.
*I* may not know how the billiard balls or bowling pins will go because i don't have access to all the information.
but that is the EXACT same issue with the brain. just more complex. difference is merely in scope, not in kind.
and several times, he uses the term "appears to have free will"... which is completely besides the point. NOBODY will contest that people APPEAR to have free will. what is at issue is whether human beings ACTUALLY have free will. and this article brings us not one iota closer.
gah.
Or, put another way, is there really any difference between the illusion of free will and free will?
Still 'Yes'.
Congratulations. You've discovered the most obvious limit of behavioralism.
Are you going to explain that difference, or do you assume that people here just accept you claim without argument? In the first case, I would really like to hear it, as I don't see the difference, in the second case, everybody reading this should pay me 10 dollars.
GP is wrong, but you didn't answer his implicit question, what is the actual difference between having free will and appearing to have free will. Since free will is basically an undefined mess in this instance I'm not sure it isn't the case that these two are identical.
Sensorial input does affect your thoughts, quite a lot actually. That means like a throw of a dice, wind, sunlight, the smoothness of the table you are resting your hands on, etc. may radically change the outcome of your thoughts.
It was an attempt to show Penrose's thinking was erroneous. But here's the thing, the argument you make here is the same, but in reverse:
For the OP, this paper is based on the assumption that whatever the brain is doing is in principle computable. You need to add this to your basic axiom that the universe is entirely deterministic before you can decide the fact of the matter here. It seems to me that these are simply your beliefs. Given the current embarrassing state of physics, I would be a little more circumspect if I were you.
So you think we have something special that makes us more complex than, say, a dog or a cat? Many animals have been shown to have abstract thought abilities, and we just thought they are less intelligent (in the sense of the ability of reasoning, not if they can remember more things, or calculate faster), actually just did not have the need to evolve a complex spoken language to help them represent those abstract thoughts.
Dogs are able to understand human gestures, and can understand certain words/phrases. That implies communication skills that many animals don't have. Then they understand those messages, and act upon them. You could say they do so because they want to. How are your choices free will but not theirs?
A cat knows it needs to open a door to reach its food. When it wants food, it might come to you for help, or it may try to open the door on its own. Regardless of which action it takes, the cat made a decision, and chose a way to act. It could have chosen otherwise, but it did not. How is that different than your free will?
[...] an identical copy of a human (or indeed any animal) cannot be made and the exact same combination of data and initial conditions cannot be produced.
How do you know that? We do not currently have the ability to make an exact duplicate of a complex animal, but how can you prove that, even if we were to do so, the copy wouldn't choose exactly the same every single time?
Suppose you were to design a DNA sequence corresponding to a human, and build, cell by cell, a body of a human as it might have grown with that DNA. Suppose you establish the billions of synaptic connections manually, in such a way that they will produce the expected output once activated, but still in a way that could have happened by chance during a normal growth of that human. Suppose you jolt it with a combination of chemicals and electricity, to jump-start the body functions and bring it to "life". Then right afterwards you give it an input, and observe the output. You crafted a being that would have the expectation of free will, but in reality has been predetermined to behave in a certain way. Can you prove in any way that there is a measurable difference from another human being that has grown naturally, has chosen its way since birth, and then has been "shut down" and jump-started again to replicate the situation?
This isn't new or controversial; I've held the same opinion for many years and Stephen Wolfram's "A New Kind Of Science" states it explicitly, under the name "irreducibility" http://en.wikipedia.org/wiki/Computational_irreducibility
Also there's no reason why an AI couldn't "show its working" so you can see how it reached its decision. Like the verbose option on some linux commands.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
I think it's a false dichotomy, saying either we use free will to make decisions, or the decisions are predetermined. I wouldn't say that the physical reality is deterministic. Causality does not imply determinism; it merely says that this X happened because of Y, not that Y will cause X. In the same way I would say the decisions you make are caused by your collective sensory inputs and experiences, but cannot be predicted based upon the same.
That's an answer to why someone might, in general, prefer chocolate. Equally there are people who are allergic to chocolate, and that's an answer to why they'll always choose vanilla.
Neither says anything about why someone who likes both flavours equally will choose one or the other on one specific occasion.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
"If humans are purely matter, and that matter follows the same laws as all other matter then agency is impossible."
You seem to completely disregard that the actual configuration of matter has both an effect on the system it is considered to be part of and on the environment the system is said to be in.
In reality all systems have both ordered and chaotic components. And such an environment inside a system can act as a informational substrate in that it can support new informational structures not found outside of it.
It's just like quarks that form atoms are a new substrate that makes molecules (and their specific behaviour) possible.
Matter that follows the same laws as all other matter can still form unique combinations just because of its local interactions with its environment.
The problem is, i think that we are (instinctively?) taught that anything that can be quantified comes from stable predictable systems. The truth is that semi-chaotic systems also can calculate, but this type of calculation does not yield easily quantifiable results. Unfortunately, this is exactly how our brain manages information. That is why we need to think hard to multiply 2.73 by 93.887 while we can can get a notion of how 'nice' the weather is in a fraction of a second.
Actually, this isn't even true classically, as "duplicating everything" involves specifying an infinite number of digits, all significant, for what amounts to the entire Universe. Then there is quantum theory, where again, if one follows the usual Nakajima-Zwanzig construction to assess the outcome of the myriad of microscopic quantum transactions underlying the chemistry and interactions supporting consciousness, at some point one has to average over the state of "everything else" in a classical/stat mech way and the outcome becomes unpredictable. With that said, human decisions are unpredictable in the same sense as a dice roll or the quantum transition is unpredictable -- because of entropy/missing information.
But it is precisely this sense that is missing in AI. We come close in NN's -- by the time a NN has been trained to make a "decision" the underlying network is often too complex for us to be able to answer even simple questions concerning the "heuristics" the net is using and we can only probe those heuristics indirectly, but we can still explain precisely how the net resolves its questions and can trace its action "neuron by neuron" through the system. We can imagine the network using RNGs to add stochastic noise to the network and make its decisioning process non-deterministic, and if the RNGs are either e.g. thermal/entropic hardware or quantum hardware generators as opposed to seeded iterated deterministic maps we can in the process obtain indeterminacy but it still isn't quite the same kind as HI.
As a few people noted above, one major difference is that human decisions arise within a consciousness loop that we haven't yet succeeded in properly duplicating AFAIK in AI. The loop means that there is always a potential for feedback positive and negative, interference, nucleation and growth, oscillations, resonance, chaos, and more in various projective views of the underlying processes, PLUS those processes are often seeded by what amounts to random noise. Simply adding stochastic input to an otherwise deterministic AI (NN or hard coded) doesn't begin to capture the complexity of consciousness.
Outside of that I actually agree with most of the comments regarding our fear of AI, although I do think that there is a big difference between the fear of (say) a very religious individual worried about the Frankenstein myth and the complete lack of fear of a computer scientist or physicist such as myself who have worked on this and would eagerly build a "true" AI if somebody would just give me/us a few gazillion dollars to do so. I'm not so sure about TFA, though -- even the Turing test is in a sense a qualitative heuristic and I don't think "intelligence" or "self-awareness" or "consciousness" in any of its many forms can easily be reduced to a set of quantitative projective tests, and IMO there is no question that my "awareness" and interior monologue intentionally directed self is a comparatively thin (but important!) vessel floating on a deep and uncharted sea of unconscious processes that my conscious self can neither control nor predict because it is an OUTCOME of those processes as much or more than it is a cause.
That's the specific sense in which AI and HI are so far distinct. HI loops back into this sea and can partially steer the boat even as the sea surges this way and that. For humans the sea is dark and impenetrable, and when currents emerge that carry us towards some particular destination we often cannot fight them -- but sometimes we can. For AI the sea itself is transparent and so far, there is no boat, only the emergent currents. Metaphor, sure, but it conveys the point.
rgb
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
The difference is that the AI can be exactly modeled, simply by making another copy of it. Given all the same inputs and the same data and initial conditions, a digital processing system comes to the same result every time. Humans are not digital processing systems, an identical copy of a human (or indeed any animal) cannot be made and the exact same combination of data and initial conditions cannot be produced.
The bit you are glossing over here is that, in your imagination, the AI system's inputs are vastly less complex than the inputs of a human being. Attach a couple of video cameras and microphones to the AI that are constantly streaming data from the real world to it, and you'll find it's just as hard to replicate the exact same combination of data and initial conditions as it is with the human.
But free will means a lot of things.
In one version, we ask, "Is the decision determined by the inputs alone, or does the person making the decision change the outcome?" This is pretty trivially answerable. No two humans will do the same thing in every situation, so we say a person has free will.
Define the inputs. The person's internal state that changes the outcome is determined by historic inputs, so if you take into account all the inputs right back to the moment of conception and including the genome of the person, can you say that two people with exactly the same history would make different decisions?
Of course, two such people couldn't exist in the same Universe because they would have to inhabit exactly the same space.
In another version, we ask, "Are peoples' actions determined purely by physical processes, or is there something ineffable that has to be considered to explain how people behave?" This is pretty obviously not answerable.
I'll answer it. The answer is yes and it is falsifiable. If there was something ineffable that is somehow pulling the levers behind the scenes, it would be, in principle, possible to observe components of the brain not behaving in accordance with physical law.
All I want is a secure system where it's easy to do anything I want. Is that too much to ask ~~ Randall Munroe
Please propose a mechanism by which an exact copy of an animal can be produced, down to every state that matters with respect to its behavior. It might theoretically be possible someday. That day is not today or even on the horizon. We can produce exact copies of digital systems today.
A system being "affected by externalities" is exactly equivalent to having different inputs. Temperature, in a system that respond to temperature, is an input and an initial condition. Digital system are designed either to directly measure such conditions or to not respond to them at all.
No, I am not blurring what "can" means. I mean no technology exists by which it might be done. Future humans (or AIs) are free to make different determinations about what constitutes "free will." We certainly make different determinations than people did 3000 years ago.
Overall, you completely failed to understand what I wrote. Go back and read again. I was quite clear that people mean at least three different things in different contexts when speaking of "free will." I did not presume to judge which you should prefer.
"Is there a difference between being able to write a novel, and appearing to be able write a novel?"
James Joyce.
Nuff said.
Something doesn't have to exist for me to make a decision. I just to decided to have a hamburger made out of purple pineapple. The fact that hamburgers made out of purple pineapple don't even exist did not stop me from making the decision.
...no matter how high a percentage of undereducated American fucktards want to credit Jesus for everything.
I didn't notice an author, any citations, or anything resembling scientific methods or ties. Just more "this made up bullshit will get my blog traffic" crap that passes for journalism these days.
" And since arm chair scientists are allergic to the idea that there exists a nature outside of our nature". real scientist starts with data/facts , and try to come up with a theory which they can falsificate. A nature existing outside of nature is nigh unfalsifiable, and pure belief. it is not being an arm chair scientist to deny such a things, it is a NULL hypothesis.If you think there is a nature outside nature feel free to provide data, hypotheses and falsification attempt to show it as natural. Until then you are babling woo.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
I think he actually meant Behaviorism.
http://en.wikipedia.org/wiki/Behaviorism
http://en.wikipedia.org/wiki/Behavioralism
The primary tenet of behaviorism ... is that psychology should concern itself with the observable behavior of people and animals, not with unobservable events that take place in their minds.
I'm not sure how the phrase in question is a "limit", but it appears that Behaviorism would dictate that they are logically identical?
Unity? Screw that: XFCE. Slashdot Beta? Screw that: SoylentNews. Australis? Screw that: Pale Moon. UX developers DIAF
Adhering to one or another psychological model is not the same as "not understanding philosophy."
Unity? Screw that: XFCE. Slashdot Beta? Screw that: SoylentNews. Australis? Screw that: Pale Moon. UX developers DIAF
The null , as used by biologist , is that we are only chemical reaction. Very complex, but nothing more. Keep your phlogiston and soul for belief. They have no place in science.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
It's my Willie that has the free will.
So you think we have something special that makes us more complex than, say, a dog or a cat? Many animals have been shown to have abstract thought abilities, and we just thought they are less intelligent (in the sense of the ability of reasoning, not if they can remember more things, or calculate faster), actually just did not have the need to evolve a complex spoken language to help them represent those abstract thoughts.
We know some animals have some ability to conceptualize possible realities, but this shouldn't surprise anyone. Sufficiently advanced animal brains work basically the same way ours do - they have memories, neural firing patterns formed by sensory impressions. Dogs and cats dream, so why can't they "imagine" hypotheticals? This is different than being sentient, much less having free will.
Dogs are able to understand human gestures, and can understand certain words/phrases. That implies communication skills that many animals don't have. Then they understand those messages, and act upon them. You could say they do so because they want to. How are your choices free will but not theirs?
Quick, name an animal that has no ability to communicate. Many animals can learn to "understand" forms of human communication, but do you have evidence that it's more than stimulus-response? Of course you can say a dog "chose" to act because it wanted to, but that's not differentiating between stimulus-response and free will. I loved my dogs very much, and can anthropomorphize with the best of 'em, but I saw no evidence that anything they did was more than responding to either conditioning or instinct.
A cat knows it needs to open a door to reach its food. When it wants food, it might come to you for help, or it may try to open the door on its own. Regardless of which action it takes, the cat made a decision, and chose a way to act. It could have chosen otherwise, but it did not. How is that different than your free will?
Again, stimulus/response. The cat is conditioned to pester you for food, or open a door. Which action it takes may be entirely determined by environmental input. To say the cat "chose" is to assert that it has free will, which is begging the question.
The idea that humans have free will is an assertion - the alternative is that we're basically noisy meat robots, acting purely on instinct and psychological or social conditioning. I personally believe that the majority of our "choices" are response to stimulus, (we are animals, after all) but not all. But the noisy meat robot argument is self-defeating, or at least, it renders itself moot, because if we don't actually make choices, then we aren't choosing to have this discussion, I'm not choosing to disagree with you, and no one chooses moral behavior. People either will or won't do things based on their conditioning and instinct, and it's literally pointless to argue about it.
So my reasons for believing in free will are basically religious - I believe in moral reality not derived purely from material existence, and this necessitates the ability to choose the good.
Now, how can we tell whether a given brain (real or artificial) has the capacity for free will, and is exercising it? I have no idea.
It's interesting to note that humans aren't even considered capable of making choices until the age of reason, roughly 7 years old. There's a certain level of cognitive development necessary for this, and it seems to be the case that language is an essential part of the process. It remains to be seen whether other species could acquire sentience after developing sufficient linguistic ability, and whether humans could speed that process. After all, we learn language from older members of our species, and have tried teaching it to other species such as apes and dolphins. I suspect the answer to both questions is yes, but I don't think I'll be around to find out.
my, your, his/her/its, our, your, their
I'm, you're, he's/she's/it's, we're, you're, they're
That's all true, and both stimulus/response behavior and free will can be unpredictable. Ultimately, I think the problem with TFA is that even if you can't tell the difference between an actually free agent and a system that has the appearance of free will, the difference is significant.
my, your, his/her/its, our, your, their
I'm, you're, he's/she's/it's, we're, you're, they're
You seem to completely disregard that the actual configuration of matter has both an effect on the system it is considered to be part of and on the environment the system is said to be in.
I don't see away to carve out the possibility of individual choice from this though. No matter how many layers or systems you have, or how chaotic they are, allows for something "outside" out their own base rules. Even in a probabilistic system, like QM, there isn't room to get outside of its own base elements, you might obfuscate it, and make first causes irreducible from complexity, chaos, and random events, but it still wouldn't be "free". I phrased that very badly, but language gets annoying with concepts like this.
I might be missing your point (not enough coffee this morning)... If I did, I apologize. I always view these things as "abstraction layers", with them being increasingly removed from the innate human perception frame as we go "down" or "up", there really isn't a difference between quantum mechanics and humans, but we can't see it because of our evolutionary context shaping our perceptions.
A patriot must always be ready to defend his country against his government. -edward abbey
Sure, while being utterly deterministic in their decisioning action, and while being completely transferrable so that one can load a given NN onto twenty different systems, present them with the same input data, and get precisely the same output. Indeed, one can write a specification so exact that one could tell somebody else in another solar system how to do the computations to exactly the same precision on precisely simulated (if necessary) hardware and load the same weights to the same precision and present the data to the same precision and get precisely the same output. The only source of variation is effectively random hardware errors, that is, the inevitable bleed of entropy into any precisely specified system in the real world, but in principle the computation algorithm has zero entropy (there is no missing information between the program, the weights, and the data). The computation is reproducible.
In the case of HI, it is by no means clear whether or not one can, in fact, specify the state of the system sufficiently precisely to cause it to consistently produce the same decision given the same data. It isn't even clear if the same system, prepared in exactly the same state, would consistently produce the same decision given the same data, because unlike the case in a simulated (not real) digitized NN, one has both of the uncertainties associated with numbers drawn from an apparent continuum (the quantum phases necessary to fully specify even the system state) and with the non-separability of the system from its environment (the bleed-in of entropy at the microscopic level from our lack of knowledge of "the wavefunction" and our practical inability to specify the external state of the Universe even if we could somehow precisely reproduce the state of the human "system".
The same sorts of issues arise in random number generation (which I'm moderately expert in). Pseudorandom number generators can produce numbers that satisfy any number of tests for randomness, but of course they are not, in fact, random. The generators have a definite state, and if one prepares the state a given way (e.g. feed them a given seed) they will always produce the same sequence. Furthermore, if one does very, very careful testing, the generators typically fail in their empirical "randomness" in some sufficiently high dimensionality. A "free will decision" is not a synonym for a random decision, but at the same time we have a very hard time asserting that a decision that is completely determined by state is in any sense free!
The point is that there is a necessary connection between entropy and freedom. One cannot meaningfully assign a measure of "free will" to a fully deterministic system with a definite state and zero entropy, because that is not what we mean by free -- it is not only constrained, it is perfectly constrained. One cannot meaningfully assign a measure of "free will" to a completely random, non-deterministic decisioning system such as flipping an ideal, truly random coin (with or without any biasing in the decision) because again, deterministically choosing even from a biased set of outcomes based on a non-deterministic event is not what we mean by free.
To a physicist, then, no system can even conceptually possess true free will. To claim that I possess true free will is to claim that my decisioning is somehow independent of the underlying deterministic or non-deterministic superstructure of the decisioning system, that I'm somehow ultimately different from my quantum physics and quantum chemistry. Free will cannot have too much entropy or too little entropy. Decisioning in some sense has to be irreversible, so that it has to be accompanied by global entropy increases, and cannot be precisely describable in microscopic terms or it disappears to be replaced by zero entropy deterministic outcomes. It is a meta phenomenon that is describable by rules only on a macroscopic scale that dis
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
Free-will is a product of self-reflection in the context of uncertainty of outcome. It is about self reference when the problem set is indetirminant. It's relation to quantum mechanics is notable in as much as QM suggests that uncertainty is an instrinsic factor of the universe. But the relationship is tenuous as Randomness or detirministicness is irrelevant as certainty in this context is a function of perception. Free will is inseparable from an integrated self concept where choice is resolved along both dependant and independant factors. Good luck finding a coefficient for faith jackholes....
As I said, if one believes in the laws of physics (and as a physicist, I am inclined to do so:-) then everything is deterministic and free will in one sense cannot exist. It doesn't exist because we don't consider looking things up in a decision tree to be "free will", no matter how complex a decision tree we create. We consider it free of will. To the extent that the Universe is a unique solution to the problem of its own dynamics, to the extent that the laws of physics are microscopically reversible so that physics itself has no entropy at the microscopic level, free will is an illusion.
At the same time, free will doesn't mean random, or non-deterministic. A decision made by flipping a perfectly random coin isn't "free will" either. So even if you want to consider quantum mechanics as being perfectly non-deterministic, as long as our consciousness ultimately depends on quantum chemistry and quantum electronics one cannot consider our will to truly be free.
So we can start by noting that truly free will doesn't exist anywhere, and then try to discern the difference between free agent HIs like ourselves and non-free agent AIs of all sorts. At this point there are many, many structural differences. There are different KINDS of AI systems -- semantic systems that are closer to stochastically weighted heuristic decisioning trees that reason according to high-level rules and low level non-semantic systems such as NNs have no idea what the number "seven" is even as they are trained to recognize numbers that are divisible by seven when a binary encoded integer is presented on a set of discrete inputs.
We can definitely say that a perfectly deterministic decisioning system has no free will. We can definitely say that a perfectly random decisioning system also doesn't have free will. Human intelligence is information-theoretically incapable of being aware of its own state information, so even though it might well be deterministic microscopically, it cannot predict its own future state any way other than computing it, given its past state plus its equally unpredictable future inputs. Most current realizations of naive (or even more sophisticated) AI systems lack even this kind of indeterminacy -- they are their precisely specified and known state information. Would a sufficiently complex AI become a free-willed HI? Sure, maybe, probably. I'd sure like to try to design and build one (and think I probably could!). Are plain old feed-forward NNs in any defensible sense free agents? Of course not. Are human in any defensible sense what you seem to be asserting, semantic decision trees? Personally, I'd have to say no. For one thing, at the hardware level they are NNs, and NNs per se have no idea what the number "seven" is no matter how accurately they can identify it or what they are trained to do with it. Humans, however, can manipulate their OWN NNs to do computations based on the notion of "seven-ness" completely independent of its neural representation or lack (of a unique one) thereof.
When I divide 179 by 7, I use a complex set of heuristic/semantic rules to get 25 4/7. Try to reduce this completely understandable heuristic to neural events! When I use a NN to divide 179 by 7, I do so either by training it to build a nonlinear map that takes inputs that are (for example) a binary representation of 179 and outputs the number 25 4/7 according to some other nonlinear map on a set of output neurons, or to "recognize" the answer somehow. At no point can any NN I've ever heard of be said to comprehend "sevenness".
One day this gap may be bridged. At the moment, the gap is severe, and because in some deep sense heuristic/semantic reasoning is an emergent self-organized critical phenomenon that appears empirically to occur when an underlying nonlinear system LIKE a NN has enough complexity to embrace semantic reductions of some input space, it isn't clear when we will get even close to sufficient complexity for even a very
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
Time and space and matter are recycled infinitely; Each time around you make the same random choices.
Start there first. But given quantum mechanics and the human brain's monkey wiring it doesn't take many choices to give the illusion of Free Will.
The difference is that the AI can be exactly modeled, simply by making another copy of it. Given all the same inputs and the same data and initial conditions, a digital processing system comes to the same result every time. Humans are not digital processing systems, an identical copy of a human (or indeed any animal) cannot be made and the exact same combination of data and initial conditions cannot be produced.
I take some issue with this. You seem (in other replies) to only accept measuring/modelling techniques available to current humans when determining that an animal cannot be exactly modelled (i.e., if you can't show it to me, it's not relevant to the conversation), yet you are willing to speculate on an AI which does not itself yet exist.
To make the trivial case, why presume an AI will run on a digital computer?
The more nuanced case has been partially addressed in other comments, with reference to externalities impinging on the nominally 'digital' process (which is really a noisy thresholding on an analogue voltage). In one sense it is reasonable to say that these random events constitute 'input' and so must also be measured and controlled for – but this gets us into a very tricky situation regarding which system is really being analyzed. If the AI requires these additional sensors in order to replicate its behaviour, then in what sense are these sensors not part of the AI? The problem, clearly, regresses quickly.
Finally, the technology to perfectly measure, for instance, ambient thermal fluctuations across some boundary doesn't depend much on the contents of the volume whose surface is being measured (I believe a physicist would confirm that 'not much' is actually 'not at all'), so the inherent challenge of 'random' inputs in human brains is no less a challenge for similarly sensitive technological systems...
Regarding free will, I’ve often thought that a lot of times we act first and think later, and I mean this in a quite literal sense. I believe a lot of our actions, particularly physical movements that are the result of choices that have to be made quickly or near instantaneously, are not really decided by what we think of as “ourselves” at all. I suspect that if you broke everything down and looked at the sequence of neurological events at an extremely fine time scale (i.e. slowed down time), so you could resolve the exact timing, cause, and effect of every event taking place in the brain prior to an action, you would find that the action is already in progress before all the sensory data supposedly needed to make the decision to act is fully integrated and understood. In other words, we often do things before the decision to do them is made. Let me explain further...
I’m saying the action in question is really undertaken almost immediately by neurological and neuromuscular systems operating below the level of conscious thought, based on sensory data that is incomplete but past a certain genetically programmed threshold. Call it a pseudo-reflexive response, because it happens before the brain has time to fully interpret sensory data and integrate it with memory. So in certain situations we don’t really make the decision to act, what happens instead is that the conscious mind is presented with a fait accompli, the body has already acted, and what feels to the decider like the making of a decision is actually more of a process of rationalization for why one has already acted. The action has already commenced, but the mind unconsciously supplies various reasons for why the action was a good idea, and the results are presented to the top level of awareness as a conscious choice that was made, a decision to do something. In reality the action happened automatically, and what feels like the full decision-making process is more of a retroactive analysis of why the action was taken. This all happens so quickly, and below the level of full self-consciousness, that it feels like free will.
I’m not saying free will is nonexistent, or that true decision making never occurs, I’m just saying that in certain time-critical situations, when the need to act quickly is strong, much of what we think of as free will is actually just the brain playing catch-up with a complex variety of subconscious reflexive and pseudo-reflexive systems. That’s my theory, anyway. I have nothing to base it on other than my suspicion that the human brain is just too slow to accomplish some of the things that feel like they’re the result of fully conscious thought processes. I think a lot of processing actually occurs at an intermediate level, a level lying somewhere between pure muscular reflex and full-blown conscious decision making. It follows that anything undertaken at such a level cannot truly be said to be the result of an application of free will. Therefore in certain tight situations we may not be quite as free to act as we think we are.
And as for the article? Well, I at least looked at it, and gave most of it a quick read through. It doesn’t seem to me that their so-called proof is all that profound, but then, maybe I’m just not enough of a mathematician and/or philosopher to recognize the true profundity of their results.
If the question boils down to whether a person is a deterministic system or not, even that is an open question. Perhaps people and animals are too random to be called properly deterministic. Neurons are and other cells are highly nonlinear analog systems that be subject to macroscopic effects based on quantum noise. If that is the case it may be possible to duplicate a human being in principle -- down to the level of the quantum state of each constituent particle, but still not be possible in principle to build a system that will duplicate her behavior. On the other hand, the systems we are built of could be fundamentally stable, or multistable at some levels of approximation and it may therefore be possible to model us accurately with simpler systems such as computers. Or it may be that while our neurons are crude hacks that exhibit all kinds of complex, impossible to model behavior, at a higher level we are mulitistable and thus susceptible to accurate modeling.
The point you make about AIs being built with random elements to model the random shit that happens in our brains is an interesting one. I doubt scientists at this point could come to a consensus about whether that is somehow important to intelligence in living systems and I also doubt that philosophers could come to a consensus about whether randomness is essential to free will.
I'd say at this point, we are making faster progress on building more and more intelligent systems. There probably won't be a bright line we will cross and then everybody will say computers are people too. It's more likely there will be progress on multiple fronts as there has been in the past, and AIs will outperform people in some respects while remaining more limited than people in other areas for some time yet. But at some point, if progress maintains, AIs will be as good or better at everything than people are and we will have no excuse to not consider them our equals. And the question will be as relevant a question whether they have free will as it is whether we have free will.
I think that time is coming much sooner than the time at which we will fully understand how our own brains work, let alone be able to build a system that does the same thing.
The point was entirely philosophical. I already said that we do not have that ability yet.
If you count the internal feedback loops of the neuronal wiring (generated through previous experiences) also as "inputs" to the final decision making process (which I feel is justified), then the answer to both halves of the question would be "yes". No two humans will do the same thing because no two humans have exactly the same subjective experiences that shape their neural topography.
Just because we lack the ability to sufficiently track and understand the underlying physical wiring due to the massive complexity that it creates doesn't mean that those physical processes aren't sufficient to answer the question. I think the question IS answerable, theoretically if not currently in practice.
kurzweil_freak
5th Kyu Genbukan Ninpo/KJJR student
Be the darkness that allows the light to shine.
As I said, if one believes in the laws of physics (and as a physicist, I am inclined to do so:-) then everything is deterministic and free will in one sense cannot exist.
Strange way to begin .. if one disbelieves the laws of physics, they still continue to exist and function. And the universe is built on quantum mechanics, which states that everything is random. In which case free will cannot exist. But why is randomness even part of this discussion? Must a system be partially random to have any sense of free will? If a deterministic system lacks free will and a random system lacks free will, how can a combination of these ever have any kind of free will? Perhaps this sense of "free will" comes from somewhere else?
Are human[s] in any defensible sense what you seem to be asserting, semantic decision trees? Personally, I'd have to say no.
As a human yourself, of course you'd say that. But some people consider their computer to have free will based on the times it "chooses" to help vs. hinder. This is an illusion of course, because we know precisely how computer work. But it's understandable given the vast numbers of inputs that affect a computer's state. Perhaps a human's "free will" is equally an illusion?
For one thing, at the hardware level they are NNs, and NNs per se have no idea what the number "seven" is no matter how accurately they can identify it or what they are trained to do with it. Humans, however, can manipulate their OWN NNs to do computations based on the notion of "seven-ness" completely independent of its neural representation or lack (of a unique one) thereof.
This doesn't seem to follow. If humans are neural networks (I assume the meaning of "NN") and humans can be conscious of the meaning of "seven", then by definition neural networks can do the same.
At no point can any NN I've ever heard of be said to comprehend "sevenness".
Sure there is, you've just identified it -- it's your own brain. If you mean you've never heard of a *manufactured* NN with that ability, then that may merely point to the shortcomings of human engineering rather than any intrinsic problem with neural networks.
In the meantime, as I said, in an HI you are stuck trying to explain the decision either heuristically (which fails if you examine it too closely/microscopically) ..
I disagree. Every decision can be ultimately be traced to a fundamental desire that is not based on logic, but is coded into your organic being. You have no choice in these desires; some were programmed into you before you were born, and some developed as you grew.
But perhaps we argue the same point. What you call a "failure" would be decision that exists with no backing logic. But perhaps this "decision" is really just one of your core desires.
My dog is itching itself across the room. A potential decision looms ...
Ok, let's work through this.
... do I a) do something about it; or b) keep typing. Huge semantic trees open out from this simple binary choice, most of which ignore enormous parts of the phase space involved, such as doing neither and getting myself another cold beer instead or farting. The decision is not entirely semantic -- I'm not ever going to work through most of the complete decision tree, any more than a chess master actually thinks about all possible moves and all of the outcomes of all of those moves and (iterate to checkmate). I will prune that tree, instantly and without thinking. How I prune the tree is another decision, but it is not one I can make semantically, or I have a self-referential problem -- how do I decide how to prune the decision tree without considering all of the possible prunings, leading me right back to my original problem?
For
Not that this wasn't entirely predictable.
if you model a computer inspired by biological systems, namely im talking about the artificial neural networks paradigm, (therefore NOT a Turing serial single-process machine) you have your question answered. By witch i mean a system that displays a "black box" like behavior (you cannot know exactly how the system is achieving an answer/conclusion, after it is “trained” ) yet you still get complex behavior.
In other words, the chocolate being preferred to vanilla may not have anything to do with you "non-misterial" answerer (where you talk about neurotransmitters ) ,... or it may, doesn't really matters.
Anyways, my point is that an artificial neural network (ANN) is able to display true AI ... its only a matter of time now, when well be able to have a network with as many nodes as a human brain , that is, about a few billions nodes - the neurons - and about 100 times more connections between them - the weights, or wiring.
Even linguistics such as Noam Chomsky don't believe on the "power of networks".. they only see systems such as IBM's jeopardy machine as a prove that human being are very good teachers.. they are very dismissive, and don't realize the paradigm shift brought by neural networks.
Obligatory: http://xkcd.com/894/