Humans evaluate decisions using their current utility function, not their future utility function as a potential consequence of that decision. Using my current utility function, wireheading means I will never accomplish anything again ever, and thus I view it as having very negative utility.
It’s often difficult to think about humans’ utility functions, because we’re used to taking them as an input. Instead, I like to imagine that I’m designing an AI, and think about what its utility function should look like. For simplicity, let’s assume I’m building a paperclip-maximizing AI: I’m going to build the AI’s utility function in a way that lets it efficiently maximize paperclips.
This AI is self-modifying, meaning it can rewrite its own utility function. So, for example, it might rewrite its utility function to include a term for keeping its promises, if it determined that this would enhance its ability to maximize paperclips.
This AI has the ability to rewrite itself to “while(true) { happy(); }”. It evaluates this action in terms of its current utility function: “If I wirehead myself, how many paperclips will I produce?” vs “If I don’t wirehead myself, how many paperclips will I produce?” It sees that not wireheading is the better choice.
If, for some reason, I’ve written the AI to evaluate decisions based on its future utility function, then it immediately wireheads itself. In that case, arguably, I have not written an AI at all; I’ve simply written a very large amount of source code that compiles to “while(true) { happy(); }”.
I would argue that any humans that had this bug in their utility function have (mostly) failed to reproduce, which is why most existing humans are opposed to wireheading.
I would argue that any humans that had this bug in their utility function have (mostly) failed to reproduce, which is why most existing humans are opposed to wireheading.
Why would evolution come up with a fully general solution against such ‘bugs in our utility functions’?
Take addiction to a substance X. Evolution wouldn’t give us a psychological capacity to inspect our utility functions and to guard against such counterfeit utility. It would simply give us a distaste for substance X.
My guess is that we have some kind of self-referential utility function. We do not only want what our utility functions tell us we want. We also want utility (happiness) per se. And this want is itself included in that utility function!
When thinking about wireheading I think we are judging a tradeoff, between satisfying mere happiness and the states of affairs which we prefer (not including happiness).
So, people who have a strong component of “just be happy” in their utility function might choose to wirehead, and people in which other components are dominant might choose not to.
PlatypusNinja’s point is confirmed by the fact that addiction happens with regards to things that weren’t readily available during the vast majority of the time humans evolved.
Opium is the oldest in use I know of (after only a short search), but it was in very restricted use because of expense at that time. (I use “very restricted” in an evolutionary sense.)
Even things like sugar and fatty food, which might arguably be considered addictive, were not available during most of humans’ evolution.
Addiction propensities for things that weren’t around during evolution can’t have been “debugged” via reproductive failure.
1) I can’t tell if it’s much older than others. The estimates I can gather (Wikipedia, mostly) for their length of time mostly points to “at least Neolithic”, so it’s not clear if any is much older than the others. In particular, the “since Neolithic” interval is quite short in relation to human evolution. (Though I don’t deny some evolution happened since then (we know some evolution happen even in centuries), it’s short enough to make it unsurprising that not all its influences had time to propagate to the species.)
2) On a stronger point, alcohol was only available after the humanity evolved. Thus, as something that an addiction-protection–trait should evolve for, it hasn’t had a lot of time compared to traits that protect us from addiction to everything else we consume.
3) That said, I consciously ignored alcohol in my original post because it seems to me it’s not very addictive. (At the least, it’s freely available, at much lower cost than even ten kiloyears ago, lots of people drink it and most of those aren’t obviously addicted to it.) I also partly ignored cannabis because as far as I can tell it’s addictive propensity is close to alcohol’s. I also ignored tobacco because, although it’s very addictive, it’s negative effects appear after quite a long time, which in most of humanity’s evolution was longer than the life expectancy; it was mostly hidden from causing selective pressure until the last century.
1) I can’t tell if it’s much older than others. The estimates I can gather (Wikipedia, mostly) for their length of time mostly points to “at least Neolithic”, so it’s not clear if any is much older than the others. In particular, the “since Neolithic” interval is quite short in relation to human evolution. (Though I don’t deny some evolution happened since then (we know some evolution happen even in centuries), it’s short enough to make it unsurprising that not all its influences had time to propagate to the species.)
2) On a stronger point, alcohol was only available after the humanity evolved. Thus, as something that an addiction-protection–trait should evolve for, it hasn’t had a lot of time compared to traits that protect us from addiction to everything else we consume.
Um, alcohol was the most common method of water purification in Europe for a long time, and Europeans evolved to have higher alcohol tolerances.
Not sure if this helps your point or undermines it, but it seems relevant.
Most people prefer milder drugs over harder ones, even though harder drugs provide more pleasure.
I think that oversimplifies the situation. Drugs have a wide range of effects, some of which are pleasurable, others which are not. While “harder” drugs appear to give more pleasure while their effects are in place, their withdrawal symptoms are also that much more painful (e.g. compare withdrawal symptoms from cocaine with withdrawal symptoms from caffeine).
While “harder” drugs appear to give more pleasure while their effects are in place, their withdrawal symptoms are also that much more painful (e.g. compare withdrawal symptoms from cocaine with withdrawal symptoms from caffeine).
This doesn’t hold in general, and in fact doesn’t hold for your example. Cocaine has very rapid metabolism, and so withdrawal happens within a few hours of the last dose. From what I hear, typical symptoms include things like fatigue and anxiety, with anhedonia afterwards (which can last days to weeks). (Most of what is referred to as “cocaine withdrawal” is merely the craving for more cocaine.) By contrast, caffeine withdrawal often causes severe pain. Cocaine was initially believed to be quite safe, in part as a result of the absence of serious physical withdrawal symptoms.
Amphetamine and methamphetamine are probably the hardest drug in common use, so hard that Frank Zappa warned against them; withdrawal from them is similar to cocaine withdrawal, but takes longer, up to two weeks. Sometimes involves being depressed and sleeping a lot. As I understand it, it’s actually common for even hard-core speed freaks to stay off the drug for several days to a week at a time, because their body is too tired from a week-long run with no sleep. Often they stay asleep the whole time.
By contrast, in the US, alcohol is conventionally considered the second-”softest” of drugs after caffeine, and if we’re judging by how widespread its use is, it might be even “softer” than caffeine. But withdrawal from alcohol is quite commonly fatal.
Many “hard” drugs — LSD, nitrous oxide, marijuana (arguably this should be considered “soft”, but it’s popularly considered “harder” than alcohol or nicotine) and Ecstasy — either never produce withdrawal symptoms, or don’t produce them in the way that they are conventionally used. (For example, most Ecstasy users don’t take the pills every day, but only on special occasions.)
Well, I said most existing humans are opposed to wireheading, not all. ^_^;
Addiction might occur because: (a) some people suffer from the bug described above; (b) some people’s utility function is naturally “I want to be happy”, as in, “I want to feel the endorphin rush associated with happiness, and I do not care what causes it”, so wireheading does look good to their current utility function; or (c) some people underestimate an addictive drug’s ability to alter their thinking.
It’s worth noting that the example is an Experience Machine, not wireheading. In theory, your current utility function might not be changed by such a Better Life. It might just show how much Better it really is.
Of course, it’s clearly unethical to use such a device because of the opportunity cost, but then the same is true of sports cars.
Humans evaluate decisions using their current utility function, not their future utility function as a potential consequence of that decision. Using my current utility function, wireheading means I will never accomplish anything again ever, and thus I view it as having very negative utility.
It’s often difficult to think about humans’ utility functions, because we’re used to taking them as an input. Instead, I like to imagine that I’m designing an AI, and think about what its utility function should look like. For simplicity, let’s assume I’m building a paperclip-maximizing AI: I’m going to build the AI’s utility function in a way that lets it efficiently maximize paperclips.
This AI is self-modifying, meaning it can rewrite its own utility function. So, for example, it might rewrite its utility function to include a term for keeping its promises, if it determined that this would enhance its ability to maximize paperclips.
This AI has the ability to rewrite itself to “while(true) { happy(); }”. It evaluates this action in terms of its current utility function: “If I wirehead myself, how many paperclips will I produce?” vs “If I don’t wirehead myself, how many paperclips will I produce?” It sees that not wireheading is the better choice.
If, for some reason, I’ve written the AI to evaluate decisions based on its future utility function, then it immediately wireheads itself. In that case, arguably, I have not written an AI at all; I’ve simply written a very large amount of source code that compiles to “while(true) { happy(); }”.
I would argue that any humans that had this bug in their utility function have (mostly) failed to reproduce, which is why most existing humans are opposed to wireheading.
Why would evolution come up with a fully general solution against such ‘bugs in our utility functions’?
Take addiction to a substance X. Evolution wouldn’t give us a psychological capacity to inspect our utility functions and to guard against such counterfeit utility. It would simply give us a distaste for substance X.
My guess is that we have some kind of self-referential utility function. We do not only want what our utility functions tell us we want. We also want utility (happiness) per se. And this want is itself included in that utility function!
When thinking about wireheading I think we are judging a tradeoff, between satisfying mere happiness and the states of affairs which we prefer (not including happiness).
So, people who have a strong component of “just be happy” in their utility function might choose to wirehead, and people in which other components are dominant might choose not to.
That sounds reasonable.
Addiction still exists.
PlatypusNinja’s point is confirmed by the fact that addiction happens with regards to things that weren’t readily available during the vast majority of the time humans evolved.
Opium is the oldest in use I know of (after only a short search), but it was in very restricted use because of expense at that time. (I use “very restricted” in an evolutionary sense.)
Even things like sugar and fatty food, which might arguably be considered addictive, were not available during most of humans’ evolution.
Addiction propensities for things that weren’t around during evolution can’t have been “debugged” via reproductive failure.
Alcohol is quite old and some people believe that it has exerted selection on some groups of humans.
What sort of selection?
Selection against susceptibility to alcohol addiction. I don’t think anyone has seriously proposed more specific mechanisms.
I agree that alcohol is old. However:
1) I can’t tell if it’s much older than others. The estimates I can gather (Wikipedia, mostly) for their length of time mostly points to “at least Neolithic”, so it’s not clear if any is much older than the others. In particular, the “since Neolithic” interval is quite short in relation to human evolution. (Though I don’t deny some evolution happened since then (we know some evolution happen even in centuries), it’s short enough to make it unsurprising that not all its influences had time to propagate to the species.)
2) On a stronger point, alcohol was only available after the humanity evolved. Thus, as something that an addiction-protection–trait should evolve for, it hasn’t had a lot of time compared to traits that protect us from addiction to everything else we consume.
3) That said, I consciously ignored alcohol in my original post because it seems to me it’s not very addictive. (At the least, it’s freely available, at much lower cost than even ten kiloyears ago, lots of people drink it and most of those aren’t obviously addicted to it.) I also partly ignored cannabis because as far as I can tell it’s addictive propensity is close to alcohol’s. I also ignored tobacco because, although it’s very addictive, it’s negative effects appear after quite a long time, which in most of humanity’s evolution was longer than the life expectancy; it was mostly hidden from causing selective pressure until the last century.
Um, alcohol was the most common method of water purification in Europe for a long time, and Europeans evolved to have higher alcohol tolerances.
Not sure if this helps your point or undermines it, but it seems relevant.
Most people prefer milder drugs over harder ones, even though harder drugs provide more pleasure.
I think that oversimplifies the situation. Drugs have a wide range of effects, some of which are pleasurable, others which are not. While “harder” drugs appear to give more pleasure while their effects are in place, their withdrawal symptoms are also that much more painful (e.g. compare withdrawal symptoms from cocaine with withdrawal symptoms from caffeine).
This doesn’t hold in general, and in fact doesn’t hold for your example. Cocaine has very rapid metabolism, and so withdrawal happens within a few hours of the last dose. From what I hear, typical symptoms include things like fatigue and anxiety, with anhedonia afterwards (which can last days to weeks). (Most of what is referred to as “cocaine withdrawal” is merely the craving for more cocaine.) By contrast, caffeine withdrawal often causes severe pain. Cocaine was initially believed to be quite safe, in part as a result of the absence of serious physical withdrawal symptoms.
Amphetamine and methamphetamine are probably the hardest drug in common use, so hard that Frank Zappa warned against them; withdrawal from them is similar to cocaine withdrawal, but takes longer, up to two weeks. Sometimes involves being depressed and sleeping a lot. As I understand it, it’s actually common for even hard-core speed freaks to stay off the drug for several days to a week at a time, because their body is too tired from a week-long run with no sleep. Often they stay asleep the whole time.
By contrast, in the US, alcohol is conventionally considered the second-”softest” of drugs after caffeine, and if we’re judging by how widespread its use is, it might be even “softer” than caffeine. But withdrawal from alcohol is quite commonly fatal.
Many “hard” drugs — LSD, nitrous oxide, marijuana (arguably this should be considered “soft”, but it’s popularly considered “harder” than alcohol or nicotine) and Ecstasy — either never produce withdrawal symptoms, or don’t produce them in the way that they are conventionally used. (For example, most Ecstasy users don’t take the pills every day, but only on special occasions.)
Well, I said most existing humans are opposed to wireheading, not all. ^_^;
Addiction might occur because: (a) some people suffer from the bug described above; (b) some people’s utility function is naturally “I want to be happy”, as in, “I want to feel the endorphin rush associated with happiness, and I do not care what causes it”, so wireheading does look good to their current utility function; or (c) some people underestimate an addictive drug’s ability to alter their thinking.
Addiction is not simply “that was fun, lets do it again!”
Addicts often want to stop being addicted, they’re just akraisic about not taking the drugs or whatever.
It’s worth noting that the example is an Experience Machine, not wireheading. In theory, your current utility function might not be changed by such a Better Life. It might just show how much Better it really is.
Of course, it’s clearly unethical to use such a device because of the opportunity cost, but then the same is true of sports cars.