Experience Machines

Will Wilkinson criticizes Richard Layard concerning Nozick's "experience machine":

Will Wilkinson / The Fly Bottle: Value Monism & Public Reason: More Layard Flogging: Here's what [Layard] says about Nozick's famous experience machine:

If offered the chance, asks Nozick, would you plug in? Of course, many people would not, for all sorts of reasons. They would not trust the machine to deliver what it promised, so they would prefer to keep their real autonomy. Or they might have obligations to others that they could not perform if they were inert. And so on. Thus this is a weak test case, especially because it describes a situation so far from our reality that we have almost become a different animal.

That the machine perfectly delivers as promised is stipulated. Inability to entertain the counterfactual--to actually conduct the thought esperiment--is not an argument against it. And 'obligations to others they could not perform'? Well, yes. This is precisely the sort of thing people might worry about because people generally think they ought to meet their obligations, regardless of the hedonic payoff. That's part of Nozick's point, dipshit. If Layard was honest, he would bite the bullet and say, yes, plug in. And if there was an experience machine for each of us that would maximize the hedonic quality of our experience, then we would be obligated individually and collectively to forgo a real life of actual action and actual engagement, and instead climb into our pods on the Matrix pod farm, and dream sweet virtual dreams until we die. If Layard will not deign to explain to us why, despite our deep sense of revulsion, we ought to see this scenario as the happiest of all possible circumstance, he cannot expect us to acquiesce to his Benthamite Philosopher Technocrat fantasy.

I find this very interesting.

Wilkinson doesn't seem to have a clue about just how much time we already spend plugged into "experience machines."

I, for example, just spent an hour and a half lying largely motionless on my bed, plugged into a low-tech experience machine. Although my physical body was motionless and largely unstimulated, the experiences that my mind was having as I was plugged in was of going back in time to May 1979, of teleporting across an ocean to a seminar room in King's College, Cambridge, and of listening to thinkers like Michael Ignatieff, J.G.A. Pocock, Donald Winch, and Nicholas Phillipson expound their respective takes on political economy and the Scottish Enlightenment. The experience wasn't "real"--I wasn't really in King's College, it wasn't really 1979, there was no conference I was present at, et cetera. But the fact that I was jacked into a low-tech experience machine rather than present at a conference did not make my experience "bad" or second-rate or not worthwhile. In fact, it was a much better experience than I would have had at the real conference: I could fast-forward through the boring talks; I could pause and rewind the action to make sure I grasped important points. And I could speed things up.

I read, after all, perhaps five times as fast as people typically talk--ten times as fast as if they are not well prepared, and using lots of filler and recapitulation as they try to figure out what to say next.

It's much more pleasant to lie on one's bed in the cool California evening and read Istvan Hont and Michael Ignatieff, eds. (1983), Wealth and Virtue: The Shaping of Political Economy in the Scottish Enlightenment (Cambridge: Cambridge University Press: 0521312140) http://www.amazon.com/exec/obidos/asin/052131240/, than it would have been to have to shlepp all the way to the fens of East Anglia in the rainy month of May for the conference.

So, yes, we would choose to plug into the "experience machine." We do so every day, even though our current instantiations of the technology are very crude and extremely low tech and unconvincing. Wilkinson does too.

Of course, we don't plug into our experience machines 24/7. We are good Aristotelians as well as Benthamites: moderation in all things. But the fact that we are not jacked in all the time is not an argument against experience machines: food is good, and we don't eat 24/7 either.

Or perhaps what I am really saying is that we are already effectively completely jacked in, that the Singularity has already happened: that it in fact had happened by the time of Machiavelli's "Letter to Francesco Vettori."

Comments

[sm]

Very good, Brad.

[nicklaus]

Ok, I am somewhat dense, but this i cant udnerstand:

The machine that Nozick is asking me to think of seems logically imposible if we allow people to have some not-so-far-fetched utility funtions. E.g. in so far as I may have (in fact I believe I do) a utility funtion where the wellbeing I provide to those I love at the margin is a complement to everything else that makes me happy, then I may choose to not plug in at the margin because at the margin the machine *cant* maximize happiness as it prevents me form making those I love happy by spending time with them etc...

but Nozick is smart and I havent actually read the original here, so what am I missing?

[g]

I think I'm confused. The point wasn't (was it?) ever to argue
"against experience machines" as such, but to argue against
pure hedonism. That is, the line of thought goes like this:

"If it's really true that all we care about is pleasure,
then we ought to welcome anything that promises a big
increase in our pleasure level, no matter how horrible
it seems in other respects. So let's concoct something
that seems very horrible while providing a big pleasure
increase: permanent encapsulation in an experience machine.
A really sincere hedonist ought to have no hesitation
in preferring being plugged for ever into an experience
machine over remaining in the real world, even though
it will mean never again doing any good to anyone else,
never again having any "real" experiences, never having
children, etc., etc., etc. But in practice they don't.
That suggests that pure hedonism isn't really tenable."

(That's for an egoistic pure hedonist; an altruistic pure
hedonist should instead be offered the chance for *everyone*
to be permanently plugged into an experience machine. Those
who profess hedonism don't generally want that, either.)

The fact that there are plenty of things that behave a bit
like experience machines -- let's call them "weak experience
machines" -- such as books, television, movies, and so on,
and that we demonstrably do spend some of our time using
them, is just irrelevant to this argument, because spending
*some* time with an experience machine doesn't have those
severely negative (as viewed by someone who isn't a pure
hedonist) consequences. You don't have to be a pure hedonist
to be willing to make some use of a weak experience machine.
Indeed, you don't have to be a pure hedonist to be willing
to make *some* use of a strong experience machine. But only
a pure hedonist would be willing to give up everything else
and be permanently plugged into a strong experience machine;
and any (selfish) pure hedonist ought to be willing to do so.
And an altruistic pure hedonist, who cares about others'
pleasures too, ought to be willing to see the whole human
race plugged into strong experience machines, especially if
some provision can be made for the continuation of that race.

Scarcely anyone actually seems to regard the Matrix as the best
of all possible worlds, though, and *that*'s the point of the argument.
The proliferation of intermittently used weak experience machines
doesn't weaken the argument at all, so far as I can see. Because
it wasn't meant to be an "argument against experience machines"
but an argument against pure hedonism.

[Brock]

Quoth Wilkinson:

"That's part of Nozick's point, dipshit."

This quote epitomizes what I dislike about Wilkinson. He's an intellectual bully.

If you disagree with him, you're a "dipshit."

He should be glad he found a job as a hack at Cato, because that quality would not have served him well in professional philosophy.

[Auros]

Nicklaus makes a good point, and I commented on it at Wilkinson's blog. Certainly if you force "happiness" to be a scalar, aggregated across individuals by simple summation, then the "experience" machine would seem like the right choice.

[economaniac]

I think this whole thread is missing Nozick's hypothetical. Perhaps what is missing is a little Epistemological perspective. Nozick's "machine" is a descendent of the "brain in a vat" hypothetical which is the present day version of Descarte's deceptive creator. The notion is that ALL of our experiences are internal, unrelated to reality, dreams from which we never wake. In the Nozick hypothetical you are not asked to choose between abandoning all you hold dear for a life of unending maximized pleasure. You choose between a life in which all your wishes are fulfilled (but only in your mind) or one where you must face the slings and arrows of outrageous fortune. If part of being happy is beleiving that you are living in the real world then Nozick's machine provides that - you get the children and the wife and the opportunity to make world changing discoveries if that is what gets you off, just none of it is really real.

That is what Nozick is aiming at. Do we care about our experience of life or the reality which we only dimly perceive? Brads suggestion that dipping into alternate worlds is a complement to our real life is completely beside the point, and suggestions that we might be made unhappy by the real world consequences of our choice either ignores the hypothetical or begs the question.

On the other hand-

"That the machine perfectly delivers as promised is stipulated. Inability to entertain the counterfactual--to actually conduct the thought esperiment--is not an argument against it. And "obligations to others they could not perform"? Well, yes. This is precisely the sort of thing people might worry about because people generally think they ought to meet their obligations, regardless of the hedonic payoff. That's part of Nozick's point, dipshit."

Err no...

Inability to entertain the counterfactual is a pretty good reason to dismiss it as a useful hypothetical. Indeed Wilkinson immediately makes the same mistake. The burden of "obligations" as we perceive them is taken care of by the machine. If people weigh them in making a choice then they are not dealing with Nozick's question and their answers tell us nothing about their views of hedonism.

To better understand Nozick's question try this hypothetical:

Suppose that the technology was available to clone your conciousness- memories, emotions, sensations the whole bit- and store it in a computer which would allow it to continue to apparently experience life with no interuptions, indeed to that consciousness no change would be evident. In the real world "you" would go right on living as before, fulfilling your obligations as best you could.

The computer "you" would have an increasingly idyllic experience calculated to provide maximum happiness, just the right amount of personal challenge to overcome, people to love and return that love, everything perfect; however it would in no way interact with the "real" world (not that it would know either way).

Which existence would you choose?

Maybe your choice is not colored so much by obligation but by attachment (I want to know what happens to my real wife and daughter). Try again but imagine you are stripped of friends and family in the real world (but maybe not in the the machine).

Next question - should anyone care about this exercise even if they understand the hypothetical.

Stripped of all connection to reality, would you care about being connected to reality. Would you even be you?

Wilkinson wants this to be a crushing indictment of hedonic utilitarianism since he believes no one would actually choose the pod, which they must do if they are to be consistent hedonists. Unfortunately he embraces concern for obligations as one of the reasons for this choice, indeed such an obvious one that Layard is a "dipshit" for even mentioning it, missing the point that such concerns are not relavent to Nozick's hypothetical and that our inability to disregard them is precisely why Layard is claiming it is "so far from reality that we have become a different animal"

[Ronald]

I'm just glad to find out that a really smart guy like Brad doesn't read incredibly faster than I do. It gives me hope that one day I myself could at least be moderately smart.

[Tracy]

I've always found Nozick's argument hard to swallow. The problem for me is tied up with the character of experience. Nozick seems to assume that experience is a series of connected stimulae; i.e. something a machine could duplicate. I think this embodies a basic error. Experience is a gestalt; an actual situation that we create as much as we react to. Nozick assumes that this can be re-created by simulation, but that requires an "expertise" no machine, however conceived, could match. To see why, think of Searle's Chinese Room mind experiment. Human beings think; machines don't. That ability is tied up with endochrinology as well as intellect. Consequently, I think the hypothetical is flawed at a conceptual basis and that flaw explains a large part of the resistance we have to the hedonistic rewards the experience machine promises. To put it short, we know we'd be bored.

[g]

Tracy, I think you're making two different arguments, and I'm not sure I understand either. (Which may mean I'm not understanding the whole thing and am wrong about there being two different arguments...)

1. About what needs to happen physiologically to give us a particular experience. It's certainly likely that stimulation of the nerves entering the brain wouldn't be enough, but that doesn't seem to me to have much to do with the possibility of an "experience machine"; it just means that it would need to do more.

2. About the possibility in theory of an "experience machine". I'm puzzled by your reference to Searle's "Chinese Room", since the point of that thought experiment was that a machine might be able to behave *exactly like* something genuinely intelligent and conscious, without being so; and for an "experience machine" it's the behaviour that matters.

In any case, even if it's true that no machine we could ever build would be able to provide (in cooperation with our brains, *of course*) the sort of (pseudo-)experiences required for Nozick's thought experiment, why does that matter? Replace the machines with God or the devil or superintelligent aliens; unless those concepts are outright incoherent (which doesn't seem plausible) the same argument goes through about equally well -- doesn't it?

[Tracy]

Yeah, I think you've missed my points. In reverse order, I don't think your characterization of the Chinese Room is correct. I think Searle's point is that, in fact, machines can't 'behave" intelligently in the sense we normally understand (i.e. Turing machines are a fantasy). True, some aspects of consciousness can be aped by machines, given enough practice, but actual consciousness is beyond them. That, in turn, means that the experience machine can't exist even as a matter of speculation because it would never be able to effectively simulate the gestalt of experiences themselves. That's why the idea is repugnant.

I see the supposed point of the exercise, of course. What I'm saying here is that unless a mental exercise squares with what else we know about the way we experience things it isn't likely to be a very constructive endeavor. Thinking about the Chinese Room is productive; thinking about the experience machine is a waste of time.

[john c. halasz]

It might be added- re Bentham/utilitarianism, on this or the other above threads- that not only is morality not generalizable to the whole of social experience/reality and incapable of "justifying" it, whereas, on the other hand, morality, being "lodged", deontically speaking, in the counterfactuality of the normative, is always a criticism of extant reality, but that pleasure and pain can not, in fact, be separated mechanistically, and the effort to do so amounts to an "insult", so to speak, to the moral dignity of pleasure.

[g]

[Apologies for clumsy hand-markup; HTML is disabled in comments here.]

Tracy,

Here are a few quotations from Searle's "Minds, brains and programs", in which he introduced the "Chinese room". Emphasis is mine in every case.

[begin quotation] Suppose also that after a while I get so good at following the instructions for manipulating the Chinese symbols and the programmers get so good at writing the programs that from the external point of view -- that is, from the point of view of somebody outside the room in which I am locked -- my answers to the questions are *** absolutely indistinguishable from those of native Chinese speakers ***. Nobody just looking at my answers can tell that I don't speak a word of Chinese. [...] From the external point of view -- from the point of view of someone reading my "answers" -- *** the answers to the Chinese questions and the English questions are equally good ***. [end quotation]

He talks here specifically about "answers" because the sort of AI program he's focusing on is one with a fairly specific task: you feed it a description of a scenario and ask it questions. (Roger Schank was working on such things at the time, and perhaps he made some overoptimistic claims about his programs.) But Searle says

[begin quotation] I will consider the work of Roger Schank and his colleagues at Yale [...] But nothing that follows depends upon the details of Schank's programs. The same arguments would apply to [...] *** any Turing machine simulation of human mental phenomena ***. [end quotation]

(Note: "Turing machine" means roughly "abstract digital computer"; your reference to "Turing machines" suggests that you think it means "machine capable of passing the Turing test", which is not what it means and certainly not what Searle means by it.)

Consider also Searle's response to what he calls "the brain simulator reply". He doesn't say "no, you can't produce the same results as a brain would produce by doing that simulation"; indeed, his response assumes (perhaps only for the sake of argument) that you can. Instead, he says in effect "Go on then, simulate all that. Then I'll simulate your simulation by means of a human being with an immensely complicated system of plumbing. See, still no understanding there."

And in his response to what he calls the "many mansions reply", he says:

[begin quotation] Even if, by some miracle, Chinese speakers exactly realize Schank's program, we can put the same program in English speakers, water pipes or computers, none of which understand Chinese, the program notwithstanding. [end quotation]

I don't see how it is disputable that Searle is saying that something could (1) run "the right program" and (2) be enabled thereby to behave -- formally, and as viewed from outside -- exactly like a human being or some other entity with genuine intelligence/consciousness/understanding/intentionality, but still (3) not really have such i/c/u/i.

I'm puzzled by your second paragraph, because it seems to me that the Chinese room gedankenexperiment is very, very, very far from "squaring with what else we know". Searle's argument would seem much less persuasive if he mentioned that actually he's not trapped in a room, but in (let's say) a large university library; that the rules he's following to process his meaningless squiggles are so immensely complicated that they fill up all the books in the library; that the task of handling a single short set of meaningless squiggles requires so much work that he has had to have his life magically extended to, say, a few billion years per set-of-squiggles. The "Chinese room" argument doesn't actually contain much reasoning; it's more a matter of painting a picture and saying "Behold!". And the picture is so spectacularly far out from reality that the intuitive judgements it provokes are pretty much valueless. So it seems to me, anyway.