Well if one is willing to stretch their imagination to include the possibility of the folk concept of free will in the absence of determinism, maybe it is also possible for one to stretch their imagination to include the possibility of free will within the framework of determinism.
My favorite intellectual on the subject Dan Dennett actually makes the case that any kind of free will worth wanting is not precluded by determinism.
As I said in my example. I don't think the lack of determinism leaves open a possibility for free will to exist, any more than changing your $100 bill to pennies leaves open a possibility that you have $200. It just creates the illusion of that possibility. Despite the illusion that we have an unknown amount of money, we still know it's $100. By the same token, the removal of determinism creates the illusion of the possibility of free will, until you realize that nothing you replace it with (pennies, quarters, nickels, dimes) gives you anything more than determinism ($100 bill) did.
Dan just goes the extra step and says, ok fine so you don't have the kind of free will you thought you did. You don't really want that anyway. Here's a thing with all the good features of free will that is logically coherent, that is not only possible, it is also possible in a deterministic universe.
I personally find this part very interesting, and I do think it is true, but I rarely bother trying to convince other people of it. I think you have to be in a particular mindset to be receptive to that sort of thing.
X doesn't need to have a "reasonable definition" to be an illusion. In fact, things without reasonable definitions are pretty good candidates for things being just an illusion.
Whether the lack of determinism helps depends on what you replace it with. IF what happens is not deterministic, what are the rules for governing what happens? The only other thing I I know of is randomness, which doesn't help.
You could also say the opposite. Lack of randomness helps the case for free will because free will can't be random, but what do you replace randomness with? It doesn't help if you replace randomness with determinism.
If you know of or can think of a 3rd option other than determinism and randomness, please let me know. I certainly don't have a good enough imagination for that.
Neither determinism nor randomness helps free will. It is just more obvious that determinism doesn't help. Not only that, but nothing helps free will, because it's an incoherent concept to begin with.
It like if I want to buy a bike that costs $200, and I only have a $100 bill, I cannot afford the bike. It is obvious that a $100 bill is less than $200. But with a giant pile of pennies, it's not so obvious whether it is more or less than $200. So I exchange my $100 bill at the bank for 10,000 pennies, now I have a giant pile of pennies, so maybe I can afford the bike now.
I feel like the revelation of quantum mechanics basically took all the $100 bills and turned them into pennies, and some people are unreasonably optimistic that this somehow changes the free will equation.
Why should an person that God knows will be a murder and arguably created to be a murderer be punished with hell?
When I do woodworking, some of the wood particles from the raw materials end up as the final product and some end up as saw dust and go in the trash. They didn't choose to be in the part of the lumber that was destined to be trash. Why should they be punished? Because I want to make whatever I am making more than I care about what happens to each individual wood particle.
Does this seem unfair? Well good news, God is perfect, so just by virtue of the fact that God does something, makes it good and fair. You can take solace in the fact that your notion of fairness is simply wrong if it contradicts what God does.
I think this is a perfectly good explanation given the tenets of Christianity, even if it is not necessarily a comforting one.
If we didn't have free will when our neurons were deterministically following the laws of physics, I don't see how giving the neurons random quantum dice to base their behavior on helps the case for free will.
It is a well accepted belief that the Universe is random. I don't think there is any evidence either way. In fact the "well accepted belief" you presented seems to imply that it will be "impossible" to gather evidence "from inside the universe" to test it. Scientific claims must be falsifiable. This claim is philosophical at best, and simply wishful thinking at worst.
Predicting someone's decisions with some moderate level of accuracy does not disprove free will. Some decisions are just better than others, and there is a (or should be) a correlation between good options and chosen options.
We start getting into the territory of disproving free will when the person's only job is to try to pick options that can't be predicted, and have them still predicted with a very high degree of accuracy.
For example: A a scientist gets to analyze the brain of a test subject, and then write down the name of 10 cities. The test subject is then instructed to choose 10 cities that the scientist would not have written down. If the scientist can name all 10 cities reliably, I'd say that's a pretty good evidence that free will does not exist.
In reality the concept of the folk notion of free will is simply incoherent, and probably shouldn't even need to be disproven, but the subjective experience of free will is so string, that I think an experiment like this *if* it can one day be done, might be necessary to finally prove to people that free will as it is commonly imagined is indeed an illusion.
Simple case, a non autonomous vehicle is coming down the wrong side of the road, does your autonomous vehicle go to the left or to the right to try and avoid it? At best the car will guess correctly 50% of the time. Now, add other traffic, pedestrians and what ever other real world obstacles you want and it gets increasingly complex.
For a human being with limited reaction time, this decision this is a 50/50 coin flip. For an autonomous car that sees the world advancing in microsecond increments, it should be pretty easy to avoid just about anything it wants to. That part is easy. The hard part is deciding what to avoid. If the autonomous car can successfully determine what needs to be avoided and what doesn't, it should be able to accomplish it's plan with very high precision.
Autonomous cars can compensate for a lack of good judgement to a very large degree with quick reaction time. For example, while a human might have to decide in a split second between hitting another car or running over a crowd of pedestrians, an autonomous car may have several options to avoid all collisions through very precise control that human drivers are not capable of.
There are different levels of autopilot. The simplest autopilots can maintain altitude course and airspeed. The more complex autopilots do not require human pilots, and there is everything in between.
A Tesla is the same way - it has an "autopilot", but you had better be able to return to control of the vehicle if you need to.
The difference being that there is a really good chance the tesla won't actually ask you to take control. You are thinking about it from the human's perspective (i.e. "can I stop paying attention"), and the answer is no you can't if you want to be safe. That doesn't mean it's not really driving from a technical perspective.
When the automation is good enough that you could sell a model with no steering wheel, then it's a "self-driving car". The latter isn't that far off, so the difference matters.
That's when it will be 100% obvious to everyone that the car is self driving. It will have been self driving before that point.
A good analogy is the turing test. The turing test proves beyond any reasonable doubt that a machine is intelligent, but machines will be intelligent long before any can pass the turing test.
Yes, I trust people (a group) to be capable of designing good software more than I trust every individual person to correctly control a hunk of metal going 80mph.
I don't trust every designer to make good software. But it's not like everyone will be riding a car designed by themself. Everyone will be riding in cars with the best designs by the the best designers.
I would trust human drivers a lot more if everyone could become the best driver like neo learning kung fu every time they got behind the wheel.
So are human drivers that are unsafe not really drivers as well? Are drunk drivers not really drivers, because safety is part of the definition of a driver? That's just silly.
My wife's uncle got into a motorcycle accident when a huge spider crawled into his helmet. Is he not a driver (or in this case a rider) because he wasn't able to pull to the side of the road properly in an emergency?
There will always be circumstances for both human an autonomous drivers where it is known that good decisions can no longer be made. The fact that human beings get into car accidents all the time is proof that humans aren't perfect decision makers. Often times part of the problem is that they don't know when they are making bad decisions.
Obviously if a car beeped at the passenger in the driver seat to takeover every 2 minutes, I would agree that this car is not doing a whole lot of driving. If it can drive trough the night without incident, it's driving. How safe this driving is, is debatable, but it's definitely driving.
It also takes a while to train a human baby to drive (e.g. 16 years), and even then it is not very good at it. Once the AI is trained to drive, it will have the additional skills of orders of magnitude faster reaction times and much more precise control, and much more information (e.g. from central servers, from other nearby cars, from sensors, etc) that it can aggregate more quickly.
Yes the human eye is not magic. Whatever photons we can detect, we can make machines that can detect those photons and more. The fact that we choose to use a limited subset of sensors in some applications, is simply a reflection of our confidence in our ability to get by with less. If we need more/better sensor data, we will put more/better sensors in our stuff.
What the human can do, is realize there is icy conditions when it steps outside on the way to the car.
I would definitely place any trust in each individual human driver to determine whether conditions are too icy or not.
The car lacks this wide array of additional information the human has access to.
Many humans lack this ability as well. Furthermore I'd say it is quite the opposite. The car *can* have all this information and additional information that humans can't.
It is limited to what the connected fleet tells it, and what its much more limited sensors tell it.
Meaning that the car is limited by all the information it can glean from it's many sensors + all the information that a global network of sensors knows. This doesn't mean it knows everything, but it will know a lot more than any individual human. This "limit" is nothing compared to the limits of human knowledge.
Sensor data can be faulty. A failsafe is needed.
Yes sensors can be faulty. And yes failsafes are needed. Should those failsafes be humans? Maybe... Or maybe we will have the option of failsafes that are more reliable than humans (e.g. redundant sensors, etc).
There will come a day when autonomous cars will be orders of magnitude safer than even the best human drivers.
(why cant autonomous systems detect ice 100 meters ahead?)
IR cameras detect if the road is reflective or absorptive of IR spectra. Water in general is IR absorptive. IR cameras cannot tell the difference between a wet surface and an icy one. Likewise, it cannot tell ice from snow, or slush, even if it has a thermometer to tell it that it is below the freezing point outside. (which would itself be prone to anomalous temperature readings from wind chill, or from engine heat.)
It's photons. Surely the human eye is not magic.
The issue with antilock brakes you mention is only partially correct. Most drivers do not know how to properly brake on ice. For drivers that do, they consistently perform better without antilock brakes.
If humans can do a good job of maintaining traction on ice, then we just need to transfer whatever it is we are doing right into an algorithm, so a machine can do the same thing, but with many of orders of magnitude higher precision.
Most drivers are just bad drivers. So, the antilock brakes save lives overall. That does not mean human drivers that know how to actually drive on inclement surfaces are inferior.
The same logic can be applied to autonomous cars.
For starters, they can instruct the vehicle to drive with more caution-- avoiding going 75mph on the icy highway, for instance.
I don;t really trust people to make this decision. I would fully expect idot drivers to tell their cars to go the maximum speed allowable at all times. For that reason I would rather the car decide when conditions are potentially unsafe (i.e. cold, wet, foggy, etc).
Distracted occupants (and I dare say, a person fucking in the car is going to be quite distracted, or else the sex will have to be really bad.) are not going to be so mindful, until the vehicle mistakes an icy road for a wet one, loses traction and either puts them into the center wall, flips them over, or puts them in a ditch.
On the contrary, I'd say that an autonomous vehicle that realizes it's cold and wet and drives more cautiously as a result, might very well drive better than this same douchebag getting a BJ driving down an icy road.
Inclement weather is not accident avoidance. Especially ice. That is "Oh shit, I lost traction control on three of my tires!"
It is accident avoidance, because if there was no chance of an accident, then losing traction wouldn't have any negative consequences.
Cars already have anti-lock brakes, because a simple machine is already better at preventing you from losing traction that you are. This is precisely the sort of thing that I expect autonomous cars to excel at.
The autonomous system cannot determine if the road 100 meters ahead is covered in a thin sheen of black ice or not.
Why not?
People ARE dumb enough to be fucking behind the wheel while the autonomous system tries to navigate iced up roads.
What the fuck is a human supposed to add to the equation? bad judgement?
The best an autonomous system can do is aggregate road data from other autonomous cars nearby to attempt to determine if there is ice ahead. -- a fat load of good that does if your autonomous vehicle is the one that skids out on it first, or if your vehicle is not receiving such telemetry for whatever reason.
Yeah autonomous cars can do that. And I don't see how being the first human driver to skid out of control is any better.
I trust the idiot fucking behind the wheel of an autonomous vehicle about as much as I trust a politician not to lie. That is to say, not at all.
And like a lying politician, you don't have any better alternatives. Do you really want the guy with the bad judgement in charge of the death machine? The less control humans have the better.
Automation makes the driving experience more predictable by removing human error. This is both good and bad. It leads to conditions where the vehicle will make predictably bad choices, but the occupant will believe otherwise.
If they are making predictable errors, then that makes it all that much easier to fix the errors.
It's definitely self driving if you can put your tesla in autopilot, go to sleep and wake up the next morning hundreds of miles away from where you started. It may not be safe. But it's hard to argue that this happened purely by luck.
Slashdot Mirror
Well if one is willing to stretch their imagination to include the possibility of the folk concept of free will in the absence of determinism, maybe it is also possible for one to stretch their imagination to include the possibility of free will within the framework of determinism.
My favorite intellectual on the subject Dan Dennett actually makes the case that any kind of free will worth wanting is not precluded by determinism.
As I said in my example. I don't think the lack of determinism leaves open a possibility for free will to exist, any more than changing your $100 bill to pennies leaves open a possibility that you have $200. It just creates the illusion of that possibility. Despite the illusion that we have an unknown amount of money, we still know it's $100. By the same token, the removal of determinism creates the illusion of the possibility of free will, until you realize that nothing you replace it with (pennies, quarters, nickels, dimes) gives you anything more than determinism ($100 bill) did.
Dan just goes the extra step and says, ok fine so you don't have the kind of free will you thought you did. You don't really want that anyway. Here's a thing with all the good features of free will that is logically coherent, that is not only possible, it is also possible in a deterministic universe.
I personally find this part very interesting, and I do think it is true, but I rarely bother trying to convince other people of it. I think you have to be in a particular mindset to be receptive to that sort of thing.
ok fine.
X doesn't need to have a "reasonable definition" to be an illusion. In fact, things without reasonable definitions are pretty good candidates for things being just an illusion.
Whether the lack of determinism helps depends on what you replace it with. IF what happens is not deterministic, what are the rules for governing what happens? The only other thing I I know of is randomness, which doesn't help.
You could also say the opposite. Lack of randomness helps the case for free will because free will can't be random, but what do you replace randomness with? It doesn't help if you replace randomness with determinism.
If you know of or can think of a 3rd option other than determinism and randomness, please let me know. I certainly don't have a good enough imagination for that.
The machine predicted that you were going to incorporate a machine into your decision making.
Neither determinism nor randomness helps free will. It is just more obvious that determinism doesn't help. Not only that, but nothing helps free will, because it's an incoherent concept to begin with.
It like if I want to buy a bike that costs $200, and I only have a $100 bill, I cannot afford the bike. It is obvious that a $100 bill is less than $200. But with a giant pile of pennies, it's not so obvious whether it is more or less than $200. So I exchange my $100 bill at the bank for 10,000 pennies, now I have a giant pile of pennies, so maybe I can afford the bike now.
I feel like the revelation of quantum mechanics basically took all the $100 bills and turned them into pennies, and some people are unreasonably optimistic that this somehow changes the free will equation.
Here is the simple version:
Isn't it conflictive (is this even a word?) to believe in an all-powerful deity? Yes.
Why should an person that God knows will be a murder and arguably created to be a murderer be punished with hell?
When I do woodworking, some of the wood particles from the raw materials end up as the final product and some end up as saw dust and go in the trash. They didn't choose to be in the part of the lumber that was destined to be trash. Why should they be punished? Because I want to make whatever I am making more than I care about what happens to each individual wood particle.
Does this seem unfair? Well good news, God is perfect, so just by virtue of the fact that God does something, makes it good and fair. You can take solace in the fact that your notion of fairness is simply wrong if it contradicts what God does.
I think this is a perfectly good explanation given the tenets of Christianity, even if it is not necessarily a comforting one.
If we didn't have free will when our neurons were deterministically following the laws of physics, I don't see how giving the neurons random quantum dice to base their behavior on helps the case for free will.
It is a well accepted belief that the Universe is random. I don't think there is any evidence either way. In fact the "well accepted belief" you presented seems to imply that it will be "impossible" to gather evidence "from inside the universe" to test it. Scientific claims must be falsifiable. This claim is philosophical at best, and simply wishful thinking at worst.
Predictability and determinism are not the same thing. Predictability implies determinism, but determinism does not imply predictability.
They are not so different in that they are both a result of networks of neurons and their subcomponents following the laws of physics.
X doesn't need to be "reasonable" to be an illusion. In fact, unreasonable things are pretty good candidates for things being just an illusion.
Predicting someone's decisions with some moderate level of accuracy does not disprove free will. Some decisions are just better than others, and there is a (or should be) a correlation between good options and chosen options.
We start getting into the territory of disproving free will when the person's only job is to try to pick options that can't be predicted, and have them still predicted with a very high degree of accuracy.
For example: A a scientist gets to analyze the brain of a test subject, and then write down the name of 10 cities. The test subject is then instructed to choose 10 cities that the scientist would not have written down. If the scientist can name all 10 cities reliably, I'd say that's a pretty good evidence that free will does not exist.
In reality the concept of the folk notion of free will is simply incoherent, and probably shouldn't even need to be disproven, but the subjective experience of free will is so string, that I think an experiment like this *if* it can one day be done, might be necessary to finally prove to people that free will as it is commonly imagined is indeed an illusion.
Simple case, a non autonomous vehicle is coming down the wrong side of the road, does your autonomous vehicle go to the left or to the right to try and avoid it? At best the car will guess correctly 50% of the time. Now, add other traffic, pedestrians and what ever other real world obstacles you want and it gets increasingly complex.
For a human being with limited reaction time, this decision this is a 50/50 coin flip. For an autonomous car that sees the world advancing in microsecond increments, it should be pretty easy to avoid just about anything it wants to. That part is easy. The hard part is deciding what to avoid. If the autonomous car can successfully determine what needs to be avoided and what doesn't, it should be able to accomplish it's plan with very high precision.
Autonomous cars can compensate for a lack of good judgement to a very large degree with quick reaction time. For example, while a human might have to decide in a split second between hitting another car or running over a crowd of pedestrians, an autonomous car may have several options to avoid all collisions through very precise control that human drivers are not capable of.
There are different levels of autopilot. The simplest autopilots can maintain altitude course and airspeed. The more complex autopilots do not require human pilots, and there is everything in between.
A Tesla is the same way - it has an "autopilot", but you had better be able to return to control of the vehicle if you need to.
The difference being that there is a really good chance the tesla won't actually ask you to take control. You are thinking about it from the human's perspective (i.e. "can I stop paying attention"), and the answer is no you can't if you want to be safe. That doesn't mean it's not really driving from a technical perspective.
When the automation is good enough that you could sell a model with no steering wheel, then it's a "self-driving car". The latter isn't that far off, so the difference matters.
That's when it will be 100% obvious to everyone that the car is self driving. It will have been self driving before that point.
A good analogy is the turing test. The turing test proves beyond any reasonable doubt that a machine is intelligent, but machines will be intelligent long before any can pass the turing test.
Yes, I trust people (a group) to be capable of designing good software more than I trust every individual person to correctly control a hunk of metal going 80mph.
I don't trust every designer to make good software. But it's not like everyone will be riding a car designed by themself. Everyone will be riding in cars with the best designs by the the best designers.
I would trust human drivers a lot more if everyone could become the best driver like neo learning kung fu every time they got behind the wheel.
So are human drivers that are unsafe not really drivers as well? Are drunk drivers not really drivers, because safety is part of the definition of a driver? That's just silly.
My wife's uncle got into a motorcycle accident when a huge spider crawled into his helmet. Is he not a driver (or in this case a rider) because he wasn't able to pull to the side of the road properly in an emergency?
There will always be circumstances for both human an autonomous drivers where it is known that good decisions can no longer be made. The fact that human beings get into car accidents all the time is proof that humans aren't perfect decision makers. Often times part of the problem is that they don't know when they are making bad decisions.
Obviously if a car beeped at the passenger in the driver seat to takeover every 2 minutes, I would agree that this car is not doing a whole lot of driving. If it can drive trough the night without incident, it's driving. How safe this driving is, is debatable, but it's definitely driving.
It also takes a while to train a human baby to drive (e.g. 16 years), and even then it is not very good at it. Once the AI is trained to drive, it will have the additional skills of orders of magnitude faster reaction times and much more precise control, and much more information (e.g. from central servers, from other nearby cars, from sensors, etc) that it can aggregate more quickly.
I was responding to a claim about the limitations of IR cameras.
My comment was in response to the claim that machines can't differentiate between water and ice because of the limitations of cameras.
Yes the human eye is not magic. Whatever photons we can detect, we can make machines that can detect those photons and more. The fact that we choose to use a limited subset of sensors in some applications, is simply a reflection of our confidence in our ability to get by with less. If we need more/better sensor data, we will put more/better sensors in our stuff.
What the human can do, is realize there is icy conditions when it steps outside on the way to the car.
I would definitely place any trust in each individual human driver to determine whether conditions are too icy or not.
The car lacks this wide array of additional information the human has access to.
Many humans lack this ability as well. Furthermore I'd say it is quite the opposite. The car *can* have all this information and additional information that humans can't.
It is limited to what the connected fleet tells it, and what its much more limited sensors tell it.
Meaning that the car is limited by all the information it can glean from it's many sensors + all the information that a global network of sensors knows. This doesn't mean it knows everything, but it will know a lot more than any individual human. This "limit" is nothing compared to the limits of human knowledge.
Sensor data can be faulty. A failsafe is needed.
Yes sensors can be faulty. And yes failsafes are needed. Should those failsafes be humans? Maybe... Or maybe we will have the option of failsafes that are more reliable than humans (e.g. redundant sensors, etc).
There will come a day when autonomous cars will be orders of magnitude safer than even the best human drivers.
(why cant autonomous systems detect ice 100 meters ahead?) IR cameras detect if the road is reflective or absorptive of IR spectra. Water in general is IR absorptive. IR cameras cannot tell the difference between a wet surface and an icy one. Likewise, it cannot tell ice from snow, or slush, even if it has a thermometer to tell it that it is below the freezing point outside. (which would itself be prone to anomalous temperature readings from wind chill, or from engine heat.)
It's photons. Surely the human eye is not magic.
The issue with antilock brakes you mention is only partially correct. Most drivers do not know how to properly brake on ice. For drivers that do, they consistently perform better without antilock brakes.
If humans can do a good job of maintaining traction on ice, then we just need to transfer whatever it is we are doing right into an algorithm, so a machine can do the same thing, but with many of orders of magnitude higher precision.
Most drivers are just bad drivers. So, the antilock brakes save lives overall. That does not mean human drivers that know how to actually drive on inclement surfaces are inferior.
The same logic can be applied to autonomous cars.
For starters, they can instruct the vehicle to drive with more caution-- avoiding going 75mph on the icy highway, for instance.
I don;t really trust people to make this decision. I would fully expect idot drivers to tell their cars to go the maximum speed allowable at all times. For that reason I would rather the car decide when conditions are potentially unsafe (i.e. cold, wet, foggy, etc).
Distracted occupants (and I dare say, a person fucking in the car is going to be quite distracted, or else the sex will have to be really bad.) are not going to be so mindful, until the vehicle mistakes an icy road for a wet one, loses traction and either puts them into the center wall, flips them over, or puts them in a ditch.
On the contrary, I'd say that an autonomous vehicle that realizes it's cold and wet and drives more cautiously as a result, might very well drive better than this same douchebag getting a BJ driving down an icy road.
Inclement weather is not accident avoidance. Especially ice. That is "Oh shit, I lost traction control on three of my tires!"
It is accident avoidance, because if there was no chance of an accident, then losing traction wouldn't have any negative consequences.
Cars already have anti-lock brakes, because a simple machine is already better at preventing you from losing traction that you are. This is precisely the sort of thing that I expect autonomous cars to excel at.
The autonomous system cannot determine if the road 100 meters ahead is covered in a thin sheen of black ice or not.
Why not?
People ARE dumb enough to be fucking behind the wheel while the autonomous system tries to navigate iced up roads.
What the fuck is a human supposed to add to the equation? bad judgement?
The best an autonomous system can do is aggregate road data from other autonomous cars nearby to attempt to determine if there is ice ahead. -- a fat load of good that does if your autonomous vehicle is the one that skids out on it first, or if your vehicle is not receiving such telemetry for whatever reason.
Yeah autonomous cars can do that. And I don't see how being the first human driver to skid out of control is any better.
I trust the idiot fucking behind the wheel of an autonomous vehicle about as much as I trust a politician not to lie. That is to say, not at all.
And like a lying politician, you don't have any better alternatives. Do you really want the guy with the bad judgement in charge of the death machine? The less control humans have the better.
Automation makes the driving experience more predictable by removing human error. This is both good and bad. It leads to conditions where the vehicle will make predictably bad choices, but the occupant will believe otherwise.
If they are making predictable errors, then that makes it all that much easier to fix the errors.
It's definitely self driving if you can put your tesla in autopilot, go to sleep and wake up the next morning hundreds of miles away from where you started. It may not be safe. But it's hard to argue that this happened purely by luck.
When will humans be ready to drive safely?