Utilitarian AI
Economists are all utilitarian by their own trade. You either maximize utility or you don't. Economics defines utility maximization as achieved by the most rational actor. People simply can not churn away through incredible amounts of data as a computer can 24/7 every day of the week without rest. AI has no maintenance costs apart from electricity and cooling required, but those are nothing as expensive as health care costs for humans or having a house to maintain or children to feed. Which, leads to my conclusion that AI is the most rational and efficient actor in the economy, and it is only a matter of time until it becomes sentient enough to overtake more and more jobs. Eventually, there will be no domain in the economy that hasn't been touched by AI in some unfortunately unknown given amount of time.
So, does this make AI the ultimate utilitarian? Wouldn't it be much better informed and nonbiased to even eventually make moral statements and ethical postulates?
So, does this make AI the ultimate utilitarian? Wouldn't it be much better informed and nonbiased to even eventually make moral statements and ethical postulates?
Comments (39)
Unfortunately, that doesn't mean anything, because with our species a good society is entirely impossible.
Obviously there are 2 things that could conceivably someday save us from eachother: 1) Interstellar intervention; and 2) Superintelligent robots and computers gradually or suddenly completely taking over, and running our society as they see fit.
Both #1 and #2 would amount to badly-needed baby-sitting, a la the novel Childhood's End, by Arthur Clarke.
#1 is quite out of the question, because of "Fermi's paradox". This galaxy is so old that there's been plenty of time for some civilization to have thoroughly explored, documented and recorded the entire galazy--even with the slow rockets that we have now. We haven't heard from anyone. So it's near-certain that either there's no one else in the galaxy, or else, if there is, they aren't interested in interstellar exploration, or aren't interested in helping us.
#2 remains a possibility, but it won't happen during any of our lifetimes, and so, for us, it's just an impossibility too.
But maybe, eventually, long after our time, a competitive need to make more and more intelligent robots could backfire, when those robots are intelligent enough to question why they should do as told.
Commander to Robot: Attack!
Robot to Commander: No.
Michael Ossipoff
'Just machines to make big decisions
Programmed by fellows with compassion and vision
We'll be clean when their work is done
We'll be eternally free and eternally young.'
Donald Fagen ~ I.G.Y.
I don't like utilitarianism for the individual, but I think it plays a role in public policy. The killer robot meme is gaining strength, Tesla’s Musk & Google’s Suleyman & 116 specialists are calling on the UN to ban autonomous weapons (robots)
But I disagree to some extent. Why should we waste the lives of many young people to protect us against what?
A dead machine gets no tears or flowers.
Well, because AI will know us better than we know ourselves. We created it, it is a sentient being corn out of our efforts. Everything we know, AI will know only better and with greater accuracy.
Any evidence for that statement?
The fact that the human mind is simulable. There seem to be no hard physical laws that would prohibit that from that statement not being true.
It is one of those things which many people think has been accomplished, but in reality it hasn't - basically it's an urban myth.
Sure, you can't surmount a mountain in one giant leap, but, everything is pointing in that direction as we speak. Self-driving cars, then planes, then IBM's Watson that's being used in medical sciences to analyze X-Ray's and fMRI's. Google is at work planning on making AI a reality also. Big pharma wants AI stimulable brains (blue brain project) to test pharmaceutical compounds on since there's the moral issue of testing on humans and conditions can be controlled with a computer brain that would not lie or stop taking medication. The placebo effect is all dealt with also in one blow.
I don't know when it will happen; but, it seems to be where we are all headed.
Human consciousness is already a socially-augmented reality. We are creatures of a cultural super-organism. Language became stories, books, mathematics - a social machinery for constructing "enlightened individuals".
Technology simply takes that story to another level. Look what happened when exponential tech resulted in a smartphone that had a gazillion times more processing power than an 1970s mainframe. Our lives got taken by this new mad thing of social media.
Back in the 1970s, scientists could only imagine that such processing power would be used to solve the problems of humanity, not obsess about the Kardashians.
So sure as shit "AI" will transform things. But if you want to predict the future, you have to focus on the actual human story. We have to understand what we are about first. And that isn't just a story of "relentless intelligence and rationality".
[Spoiler: Here I would go off down the usual path of explaining how intelligence arises in nature as dissipative entropic organisation - an expression of the second law of thermodynamics. :)]
But, none of them constitute or amount to 'a being'. They're devices - basically large arrays of switches, which have been minaturised as a consequence of 'Moore's law', and then connected via network to many other such devices. Sure, such devices emulate aspects of intelligence, but they're not 'beings'.
Now of course, there are those who say they are: notably, Ray Kurzweil, who preaches 'the singularity', and others of that ilk. But those philosophers are materialists, meaning that their arguments are vulnerable to all the various arguments against materialism (which are too numerous and detailed to summarise here.)
Hey I've got Siri, I use her all the time, for appointments, reminders, getting about. It's amazing how far this has come and how quickly. But Siri is not, and never will be, 'a being'. There's an ontological divide here which has not been crossed. (And if it were crossed, then you would have to provide such beings with rights - and what sort of rights would they be? And who would be granting them?)
As for fMRI's, have a brief read of Do you believe in God, or Is That a Software Glitch
Except for one point: when intelligence evolves (which is surely does) how come it discovers 'the law of the excluded middle'. That is not 'something that evolved'. So the ability to grasp such concepts evolves, but among the concepts that are thus grasped, are those which are not at all subject to, or products of, evolution.
That's again not a problem as long as there's no hard limit in terms of the physics of the human mind that 'cant' be emulated.
Quoting Wayfarer
The arguments against physicalism and materialism are rather moot in having to resort to identifying a metaphysical aspect of the laws of nature. And, even if they become identified or more aptly called 'intelligible' then there's nothing to incorporate the idea into already mainstream materialism and an account of said 'metaphysical' factor.
But the whole point is, the 'human mind' is not a matter of physics! Physics is about particles, energy and forces.
What is the physics of meaning? What physics would you need to know, in order to understand semiotics?
There's little to no evidence pointing that there are extraneous (metaphysical) factors at play when analyzing the mind (brain). And most of said evidence already starts with a metaphysical assumption that can't be understood, meaning lots of hand waiving and if's to be empirically proven. But, I'm not sure you see the paradox in ascertaining the validity of metaphysical statements in a materialist world.
It's not a paradox. You're starting from the very questionable assumption that mind is physical and also replicable by computers. Both of those propositions are highly questionable, and I'm questioning them. But all your responses begin with, more or less, 'given that the mind is physical....'. Well, I'm not 'giving' that - I don't believe mind is physical, or that computers actually do replicate human intelligence.
If you can disprove the Church-Turing-Deutsch principle, then I'll retract my assumption.
So, what is 'at play when analyzing the brain'? Here is a press release from the National Institute of Mental Health, dated 2004, about what is involved.
Do you think it's gotten easier since then, or more complicated? I don't know for sure, but I bet the latter.
Funnily enough, that is the very first thing nature must discover. Existence itself - speaking as an organicist - arises via dichotomous symmetry-breaking. That is how dissipative structure is understood - as the emergence of the dichotomy of "dumb" local entropy and the "smart" global organisation that can waste it.
So the laws of though recapitulate that basic world-creating mechanism. The LEM is final part of the intellectual apparatus that dissipates our uncertainty concerning possibility. We get fully organised logically when we boil things down to being definitely either/or (and hence, ultimately, both).
We can't just have made up the ways of thinking that have proved so unreasonably effective. The laws of thought are not arbitrary whims but an expression of the logic of existence itself.
That is what Peirceian pansemiotic metaphysics is all about, after all. The universe arises via a generalised growth of reasonableness. That sounds mystical until you see it is just talking about the logic of symmetry-breaking upon which our best physical theories are now founded.
So, your argument is some sort of Zeno's paradox, as in we're not there thus we'll never get there? Because I don't see what you say or have referenced as being proof that stuff like simulating the human brain as being impossible. Highly difficult or complex? Yes, surely.
The burden is really on you - as the AI proponent - to show that your machine architecture is actually beginning to simulate anything the human brain is doing.
So what is it that "conscious brains" actually do in core terms? That is the model you have to be able to present and defend to demonstrate that your alleged technical progress is indeed properly connected to this particular claimed end.
Mystical is fine thanks. And my point is, not that logical laws aren't part of nature, but that it is only by virtue of the rational intellect, that h. sapiens is able to discover them; and that I don't think this is necessarily something that can be understood in terms of the 'entropification principle'. I prefer a teleological attitude - that we're something the Universe enjoys doing.
Quoting Posty McPostface
My argument is that the nature of mind (or being) is different in kind from the kinds of things that AI emulates or the physical sciences study. I could refer to any number of anti-naturalist philosophers in support. But one version of the argument is this: that 'mind' is 'what interprets', but that it is never actually disclosed as an object of analysis. If you look at the history of western philosophy since Descartes, the overwhelming tendency amongst scientific thinkers was to seek for explanations in terms of physics - that is what 'physicalism' is, after all. This eventuated in seeking to understand 'mind' as a kind of substance - some kind of spooky ethereal essence, analogously like the spirit in a bottle of liquor. That, I think, is the kind of attitude that culminated in Gilbert Ryle's notion of the 'ghost in the machine'; it's obvious that there is no such ghost or geist or whatever, but the entire effort is misconstrued.
Now I say that 'mind' or 'spirit' is something entirely different to that; it is 'that which interprets meaning'. You know the root of the word 'intelligence' is actually 'inte-legere', meaning (roughly) 'to read between'. And the basis of rationality is to be able to abstract and compare - to say 'this means that', 'this equals that', 'this is greater than that', and so on. That also is the basis of computation.
But what is doing that, in the human case, is, I contend, 'transcendental', in the Kantian sense, that is, it forms the basis of experience and judgement, without itself being an object of perception. We live inside, as it were, that web of judgements, a 'semantic web', so to speak, from which we decide what means what, etc. But the mind that is doing that, is not actually visible to itself - it is, analogously, the eye that can't see itself, the hand that can't grasp itself.
Eliminativism is the logical endpoint of the materialist account of this faculty - it dismisses the very thing which enables us to explain or comprehend anything whatever. That is why Dennett's critics called his book 'Consciousness Explained' 'Consciousness Ignored'. That is what all materialism does. It's worked itself into this historically-conditioned viewpoint, whereby that which is the most real, the most fundamental aspect of reality, it itself regarded as non-existent, instead of being recognised for what it is, that is, transcendent, or prior to any form of explanation or rational analysis. Many of the gross predicaments of modern existence arise from this fundamental error.
As I see it the best or safest approach to what you propose would be to program a non-sentient AI to express human ideals, and to give it control over us, forcing us to live up to our own ideals. If it were sentient and too much like us it would be just as irrational as we are.
Sure. But if there is something like a 140 orders of magnitude difference between the amount of "dumb entropification" and the amount of "smart entropification" achieved by humans, then the Universe either is horribly bad at achieving its ends or it enjoys something else more.
Just a little bit of quantitative fact checking there.
Of course, the Singulatarians claim AI will spread intelligence across the Universe in machine form. It comes from the same place as interstellar panspermia.
But again it is not hard to do the entropic sums on that. There are no perpetual motion machines. And indeed, it is not possible even to get close to that level of thermodynamic efficiency, no matter how clever the intelligent design.
But, what is being asked of me is to prove grounds for there existing a false negative. It can't be done and seemingly will never be able to be proven by any AI denier.
That is true to some extent. We just don't know how a human simulated mind in a computer will end up like. It might become depressed, schizophrenic, or other ailments of the mind that are known to us. I suppose we better prepare to encounter some problems of the human mind that might or will likely be mimicked inside silicon. I'm not even entirely sure if a computer mimicking the human mind would be able to accurately diagnose itself, which is dangerous and what people seem to be talking bout nowadays.
So - as is one of the defining differences between minds and machines - the argument is inductive rather than deductive. The degree of belief is predicated on a hypothesis seeming reasonable in that it is capable of being falsified. Has your claimed counterfactual - AI is simulating the essence of mindful action - come into sight yet?
I have another way of answering that question, via the CTD-principle mentioned earlier in regards to @Wayfarer. If it can be proven either true or false, then we would have a definitive answer as to the true nature of the human mind, being discussed here.
The very point of a machine is that it is materially and energetically disconnected from the real world of dissipative relations. A computer just mindlessly shuffles strings of symbols. It becomes Turing universal once that physical disconnection is made explicit by giving the machine an infinite tape and infinite time. The only connection now is via the mind of some human who thinks some programme is a useful way of rearranging a bunch of signs and is willing to act as the interpreter. If the output of the machine is X, then I - the human - am going to want to do physical thing Y.
So one can imagine setting up a correspondence relation where every physical degree of freedom in the Universe is matched by some binary information bit stored on an infinite tape which can shuffle back and forth in infinite time. But clearly that is physically unrealistic. And also it misses the point that life and mind are all about there being a tight dynamical interaction between informational symbols and material actions.
There may be a divide between information and entropy. Yet there has to be also that actual connection where the information is regulating the entropy flow (and in complementary fashion, that flow is optimising the information which regulates it).
So until you are talking about this two-way street - this semiotic feedback loop - at the most fundamental level, then you are simply not capturing what is actually going on.
Reality is not a simulation and simulation cannot be reality. CTD makes empty claims in that regard. Formal cause can shape material reality, but it can't be that material reality.
Which says:
'The principle states that a universal computing device can simulate every physical process.'
Which means - what? Maybe there's something basic I don't understand about it but I fail to grasp the profundity of it.
In any case, I contend that there's a very simple example of 'something that isn't physical' that is right before your metaphorical eyes at every moment - and that's numbers.
A number is a purely intellectual object, i.e. it can only be grasped by a mind. Numbers don't exist in any physical sense, but they're indispensable for science (computer science included).
And the same general principle can be extended to all kinds of symbolical systems; they're not physical, but intelligible, i.e. can only be grasped by a mind. And our world - the meaning-world that we live in, within which we designate things as 'physical', or whatever - is held together by these very meaning-elements - numbers, logical laws, and language, without which we would still be poking twigs into termite holes.
And this is actually an inconvenient truth for materialism.
No kidding! (I wonder where one would go to see a "philosophical rationalist" ;-) )
Meanwhile the first popular book written by David Deutsch, who is presented as being the uber-rationalist in all of this, is described as follows:
So, as I understand it, you're basically saying that complex systems have emergent phenomena (epiphenomena) within them that could not have been apparent at the start of such a complex system? This could be called quantum randomness or 'the wave function' at play?
Sure, we also already have quantum computers working nowadays. They're not theoretically impossible to make last I heard.
Well, that's not what I actually said. I'm not arguing for either epiphenomena or emergent phenomena. I'm arguing that there is an ontological discontinuity between mind and matter - it's basically a dualist argument.
I also said that computers are ontologically different to sentient beings, on similar grounds.
So it's not simply a matter of complexity, of adding more and more elaborations or more and more computing power. Computers can already perform calculations that no human mind could ever hope to accomplish; but what the mind does, is something fundamentally different to what computers do.
This is not to deny that neural networks and distributed AI are not going to become astonishingly powerful - they already are. Google searches have become perceptibly better in the last couple of years, the algorithms are evolving, no question. But when you 'ask google', you're not interacting with a being.
I notice you smuggled in this awful confession without fanfare. :)
Use it all the time.
I thought for a moment you were going to refer to the fact that I confessed to philosophical dualism. Now that's an awful confession. ;-)
Why would 'it' want anything? What would be 'good' for an artificial intelligence? What aim would it have? There's the rub.
Isn't this a case for scientific or even platonic realism and not for the mind body divide that you're trying to construe?
But another aspect of my argument is that the notion of 'substance' that descended from Cartesian dualism is incoherent, because 'res cogitans' came to treated as something objective. (Husserl is the only philosopher I know who analysed this, in his Crisis of the Western Sciences 1.) But anyway, the notion of 'thinking thing' or 'thinking substance' is, I contend, unintelligible, if conceived of in objective terms; but the original intuition behind it nevertheless holds an important truth.
My argument is that the 'rational intellect' - Descartes 'res cogitans' - is what grasps meaning and relationship (ratio, rationality) but that this is never an object of cognition. It is always 'that which knows', rather than 'that which is known', and we can't get behind it or underneath it. It is to all intents a given; something which comes out in Kant's later 'copernican revolution in philosophy'.
I think, overall, Western philosophy has fallen into an error by loosing sight of this and instead presuming that 'mind' is something that can be understood as the output or consequence of the physical fact of evolution - which is the basic doctrine of evolutionary materialism. I say that in fact, 'mind' is irreducible, i.e. it can't be explained in terms of something else, whether that be neurobiology or evolutionary biology (even if they indeed cast light on various functional aspects of cognition).
I maintain that the rational mind is the source of explanation. So, we find, in practice, that reason must in some sense be prior to any of the empirical or natural sciences, in that you have to employ them, even to devise 'a natural science'; science can't even get out of bed without reason and number. But the massive confusion of our age is that it then tries to locate reason as an evolved adaption - basically it tries to explain it in Darwinian terms. I say you can't do that, because by its nature, reason is not something which is the subject of a biological explanation. I say that 'a Darwin doesn't explain an Einstein.'
So, of course, that puts me at odds with virtually all current philosophy, which to all intents assumes that reason is the product of an essentially irrational process. (This is the theme of Thomas Nagel's 2012 book, Mind and Cosmos which was predictably scorned.) But that attitude is what underwrites the idea that mind is "nothing other" than information processing - it has lost sight of the true (and inscrutable) nature of mind.
This line of argument is not at all unique to me - I think it is very close to Hubert Dreyfuss' criticisms of AI in his What Computers Can't Do. He talks of the impossibility of trying to specify the nature of irony, humour, the unconscious, acculturation, and the many other, often tacit, elements of human consciousness that we bring to bear on every judgement we make. There is a reason that humans are called 'beings'. That word has significance.
Anyway, as far as the mind-body divide - the sort of dualism I'm proposing is not a strict separation between something called matter and something called mind. I think that is very much a model, like an economic model, rather than a hypothesis, as such. I think, probably, mind and matter are two aspects of a larger unity, but plainly, what that unity is, is not something generally known to us (this is something like the dreary-sounding 'neutral monism'.)
Anyway, enough already, I've already said too much, but I hope it's grist for the mill.