Friday, October 9, 2015

When machines lie to themselves

The ingredients of this scenario are fairly prosaic, given the era we are in – though they might have generated cries of denial a mere hundred years ago.

The primary component is simply the optical reader of a smart phone that can read QR codes and an imagined environment where all objects have some sort of QR code (in some case, at different scales, many of them) imprinted on the object.

The QR codes often contain equations. Solving these equations or finding the equations amongst many other QR codes (which may or may not have information as to where an equation QR code is) leads to energy resupply and parts replacement. The optical reader and it’s processor are mounted on the armature of something essentially the same as a bomb squad robot. This allows control lines to run from the processor on board the robot to the various actuators in its arm – aiming the armature at different QR codes or running a systematic search for QR codes by moving the arm, or activating the treads of the robot in what is essentially a larger scale search for QR codes (by not just searching from one location, but changing location entirely before searching from one new location)

Apart from an environment with QR codes and electrical energy supplies scattered around it (hardly a natural surrounding), this is an entirely conventional arrangement with no controversy to be found at all. Maze running robots have been around for decades and this is simply a more sophisticated model.

What we will add here is not controversial either – adding complexity to the environment (harder to find QR codes that indicate energy, harder to find QR codes which point towards the general location of energy QR codes) leading to more pressing environmental demands – as well as random processing system/program change in various models (an analogy of mutation). The robots that keep finding energy sources (and where needed, replacement parts) will then be taken as the model for a next generation with a random change to their processing/programming. Those which don’t find energy or replacement parts cease to function and are not used as models for a next generation. To be clear, several generations of non mutation would occur to properly road test a ‘design’ (‘design’ being simply the old model with it’s new, random processor changes) to reduce the chance it’s survivability is purely dumb luck (just happened to be in the right place at the right time to get energy), before new mutations are added.

Further, there will be more than one machine and the environment will be difficult enough that it takes multiple machines operating to locate energy QR codes.

Arguably, given the random processor changes above, a machine might eventually change to where it does not need other machines in order to locate QR codes.

But for now we will assume that multiple areas need to be scanned at once (moving QR codes) and not only that, but in order to gain QR energy sources, the machines must be able to read each others QR codes, of which they can display a small range. In a particular set up this might allow the machine that spots a clue to QR energy, set a QR signal upon itself that the other machines, if they scan it, provides a signal to the processor that in a very particular set of arrangements between machines, guides the other machine(s) to a source of energy or parts.

Now, while before we ignored the potential for a machine to have a sequence of changes in it’s processing that latter generations can survive solo existances in the environment, here it could be argued that changes could easily occur cause a completely different occurrence than that scenario where one machine finds a QR clue and activates a QR display upon itself AND that QR display is triggers another machine(s) processing which eventually leads to the energy source for that machine.

For instance, what about a processor change where the finder robot just doesn’t activate the QR code?

Perhaps ‘selfishness’ leaps to mind? For the purposes of this scenario, this text does not inform you ‘do not think that’, but at the same time this text does not affirm you thinking that. Think it if you want to and draw associations if you want to – but you are not being asked to do so. If you do so it is your volunteering to do it off your own bat – this scenario, I would argue, does not require you to do so. And that is part of the point of the scenario. The optional nature of such an observation.

Continuing with random processor changes (in regards to which I am just going to break down and call ‘mutations’ from now on), there could be a number of break downs. The QR display signal is lit up, but the other robot has mutated to the extent their processor does not use it as an input at all.

A more complicated to explain breakdown might be where the QR code had markings, which triggered the processor of another robot to aim its camera in a certain direction (the movement of the arms actuators heavily influenced by the first robots display QR code markings) to a certain area. Where before this might have had the second machine looking at the QR energy, the mutation might break this delicate relationship as there are many ways the robot arm could move and just one different firing could have it aiming in an entirely wrong direction, making it miss out on energy and possibly it’s own extinction in regards to latter generations.

That was actually quite difficult to describe without resorting to saying ‘the information in the QR code points to the right location to scan’. However, what is important is to keep track of how a certain marking on a QR code can become an input to a processor, the processor – in regards to how the program triggers new on/off states inside the processor and how its arrays of semiconductor gates are configured in hardwired form or programmed form to those new states with further new states formed, until eventually we get the states sending on/off signals to actuators in the robot arm, which determines where the optical reader is aimed. And where the optical reader is aimed determines what QR codes are detected, the marks of those QR codes determines input to the processor, which generates new states, etc, etc.

When you keep this in mind you can see something quite active, but not really different from a plant – just far more active. Or if you must use the Z word, not much different from a zombie. The markings and semiconductor responses to markings and the actuator changes which then make the optical reader find new markings is a continual flow, like a river that triggers the release and closing of dams into it, with those closings and openings either releasing water that triggers new openings or closings, or especially important, the absence of water still triggering openings or closings. How could such a river keep going on and on without eventually running out of water? Well all the ‘rivers’ and the configuration of openings and closings that did run out (leaving a machine just sitting in a corner or driving endlessly against a wall) went extinct and the ones that flowed longer, long enough to lead to a following generation, are of course the ones that keep getting repeated from generation to generation.

Possibly one of the greater mutations in such a system is internally initiated Darwinism – one might say where ideas face adversity and are potentially allowed to go extinct. This adaptive model, instead of waiting for the animal to both enter into and die in a particular scenario it cannot cope with (and that death leading to an absence in following generations genetics, thus being the information in genes that is there by being absent), allows the idea/behaviour that drives entering that particular scenario to instead face some kind of adversity and potentially die itself, before that idea kills the organism by driving it into that scenario it cannot cope with.

But that, while it’s important to the subject enough to give a brief outline (enough to keep it vaguely in mind), it is a little off topic and I wont continue on it here.

One of the pivotal issues of this scenario is the breakdown of the machines ‘social’ system of hunting for energy QR. Here, like the reference to ‘selfishness’ from before the term ‘social’ is optional – though I grant I brought it in myself here and so you can blame me about doing that! Don’t think of a white bear! It’s terrible, but in raising so have I muddied the waters – and yet for a number of reasons, though I’m raising it as an optional consideration rather than ‘how it is’, I think it’s important to at least raise it as an optional consideration.

Having gotten those caveats over and done with, we can see the QR marking reactions that mean multiple machines (each with a set of scanning behaviours that scan the other machines and those other machines potentially lighting up their own QR codes on the surface of their machinery end up enacting the actions that eventually end in tapping into QR energy) could break down with a mutation here or there.

(And yes, long text in brackets will be our curse here as they are needed to avoid summarisations which, in their reduction of the events, give misleading conclusions. That is why I’ve tiptoed around words like ‘selfish’ and ‘social’ (even though it was me who brought up the latter! I know, I’m bad!))

What is important is to outline not ‘functionality’ (yes, more scare quotes to indicate optionals!) in regards to getting QR energy, but to instead outline how a breakdown can occur. With breakdown defined as one set of inputs and behavioural outputs and survival, with just a mutation here or there, being a set of behaviours that does not lead to energy obtainment/survival.

With an establishment of such breakdowns (how a careful sequence of input/out put can break, like removing a domino or two from a series of standing dominoes breaks the chain reaction), we can begin to see how machines can lie to themselves.

Machines that lie to the logic processes they are comprised of

As we can see, breakdown can lead to extinction. Thus mutations that somehow reduce the effects of breakdowns are more prone to survive in latter generations. Of course breakdowns occur from mutations – so it’s mutation vs mutation here.

As machines of a particular pattern who’s  following forebears existence hinges on the capacity (in the hardened environment we introduced) to not just scan their environment for QR codes but also scan other robots (of the same pattern as themselves) for QR codes as well, processing the QR codes displayed on other machines of the same pattern (in a way that leads to energy obtainment) is pivotal to survival.

At first glance this may seem far away from lying – we’re talking some kind of ‘understanding’ of other machines QR codes here, after all. Aren’t we?
Strictly speaking, no. We are talking the QR code on one machine, of which the markings are an input to another machine, who’s processors semiconductor gates go through a number of reactions to that input, creating on/off states, which more of the processors semiconductors react to, creating more inputs – this goes on for X generations until it hits an output to the robots arms actuators or treads.

Complex, but nothing about this requires anything we might call an accurate understanding of the other machine, at all. As long as the robot gets its energy in the end, it doesn’t matter what sort of process goes on in regards to receiving of input, the reactions and then outputs to actuators. As long as that energy is obtained.

Granted, a process that is a poor understanding of the other machine might lead to less energy gathering than could otherwise be optimised. But if the machine is getting enough energy even with a suboptimal process response to the other machines QR codes, then it’s going onto the next generation so in regard to Darwinism there is no issue there.

Now it’s so easy as to perhaps sound a little trite to simply state here that if that’s how the machine understands other machines QR codes…that just ‘if it gets the energy, then that’s good enough’…then that applies just as much to the machine understanding it’s own QR codes! Actually, even more so – ‘understanding’ other machines states helps it take advantage of their optical scanning, thus giving it much more capacity to survive than scanning alone would give it.

Where as understanding it’s own QR codes doesn’t grant it any further scanning capacity. If an understanding of its self is ‘good enough’ to get energy, then that’s as far as it will go.

Further, in it’s default state it can only scan the surface QR codes upon the outer hull of it’s robot body.  We’ll give it a break to some degree –we’ll say the components inside it’s processor have QR codes on each of them. Even the states have their own codes. But these codes are so tiny that the default optical reader simply couldn’t read them, even if it removed the cover of the processor (it would require a prosthetic and that the robot ‘trusts’ that prosthetic (a microscope/electromagnetic sensor) and doesn’t process it as the devils work or something). Even worse, this hits an Ouroboros point – it’s clear eventually the optical camera cannot look at the components it itself is made of, for being those components! Some amount of tracking is literally impossible, for it being impossible for a tracking device to track itself in detail (and sometimes, at all!). I believe the blind brain theory document refers to this tracking issue, originally.

Close enough is good enough when it comes to the other machines. A fairly promiscuous position already. When it comes to the machine understanding itself, it gets outright slutty! It’ll get down and dirty with the first understanding it lays its hands on!

This is the point where for those who argue against materialism, their own notions of ‘it’s just a machine’ turns against them.

Why would the machine do any better than that, in regard to itself? It’s just a machine, just as you say! Why wouldn’t it’s understanding of itself just be quite appalling compared to the actual state of things – the most convenient way of getting the energy it could come up with the least energy spent figuring that out? Whatever dross it comes up with (if any!) to gain an understanding of that thing it is comprised of. It’s just a machine, after all! Why would it do any better?

So, how does that tie in?

Well, if you take it that it does indeed show how a machine can lie to itself/to the processes it is comprised of, then we have a clear cut example of a machine doing something humans are well known to do. Lie.

“So…so what? It’s a parallel to human behaviour – in regards to us, that doesn’t mean…”

I’m not going to answer that. Instead I’m just going to ask you to put yourself in the shoes/treads of the machine and imagine it from their perspective as best you can.

From their perspective, instead of asking how humans differ from machines, let’s ask how the machine could differ from humans in regard to the lie it delivers to itself?

What is the machine going to do about that lie, what extra thing, to stop it from ending up in a lie about itself and recognising itself as a machine?

As you say, it’s just a machine – what else could it do given this limitation?

But it’s a machine lying to itself – it’s not just that that parallels human lying – it’s that the machine itself could be reporting that it has consciousness, it has experience, that it has…qualia.

“You’re just a machine! You don’t have any of that!”

But from the machines perspective, what extra thing is it ever going to do to stop thinking these plainly false conclusions? It’s clearly a lie – we’ve established the robot can lie to itself – and this is one kind of lie that could be taken up just as much - so therefore it could claim it and even feed such a claim, in information format, to it’s own processors.

And why would it ‘want’ to, given these reports it gives and it’s processor commitments to such end up getting it energy and it lacks the capacity to scan itself in fine detail, as well as the Ouroboros problem? Look at it from it’s perspective and there are so many hurdles in the way of disproving it’s notion it has some sort of ‘qualia’. It would have to develop some kind of prosthetic detection tools to really start to analyse it’s internal components and find no such qualia exist in there. Even then why would the machine accept that (what’s the energy profit in it for them?). The machine might report ‘There’s more to me than just the processor’. Claim there’s more to them than just the brain.

So you’re stuck with a bunch of robots reporting consciousness, experience, qualia. You might even say they are claiming such things.

And if you look at it from their perspective, you can see there’s nothing special about the machine that would suddenly snap them out of these lies and show them the truth of the matter. Instead they would indulge the notion, printing out massive reports about their consciousness and qualia – especially as such ‘social’ communications, given in the past generations communicating robots had some energy finds ‘shared’ with them, this communication with it’s flattering conclusions, gets more energy shared with them. (‘flattering’ being a derivative of robots which use a breeding process to determine new generations, with ‘flattery’ being something similar to the prime breeding stock signifiers that processors started to detect (after much mutation over time and some shorter term processing state ‘mutations’, with the processors getting so complex that the patterns in them can mutate, thus accelerating the evolutionary process))

So you have all these robots claiming consciousness, experience, qualia – I know, it’s appalling!

But what else would they do? Can you see it from their perspective – it would seem perfectly natural to them, just as much as you can see in mechanical terms there is nothing else they could do. They are just machines – there is no ‘out’ that would let them see otherwise. Indeed, if such an ‘out’ existed, it might mean they were indeed more than machines! A lack of divinity is what makes their sense of personal divinity exist!

Imagine trying to convince them otherwise – you can already feel it, because you know mechanically there is no way they can by default detect the lies they deliver to their own processors – you’d be arguing until you are blue in the face and they would keep reporting consciousness, experience and qualia.

Except maybe a few – maybe some, in a hunt for more energy collection, develop a kind of robot science – and at first incidental findings in regard to their own mechanical nature start to build up (as a new connection to energy finding is found to be enabled through it) and those robots, who ‘trust’ the measure prosthetics they have developed and then applied to themselves – they might actually listen to you and cease their claims of consciousness (or at worst, redefine the term radically). They might actually stop claiming consciousness, experience, qualia – at least in terms of how the other robots define them.

So what is robot science? How does it differ from our science?

Well, it doesn’t. It’s just more science. We all acknowledge the materialistic nature of scientific investigation – which means the robots would use the same thing (given they are in the same material plane as us, of course).

So some of the robots would use the same science as us. And for those robots, it would mean they would stop making claims of consciousness, experience and qualia like the other robots (the muggle robots!) do.

So now you have two robot perspectives – you’ve always know the robots claims of consciousness were naive – and now you have the post scientific conclusion perspective of some of the robots to consider as well.

How would the scientific robots explain to the naive robots that they are just machines? This is a particularly relevant question to those who argue against materialism – why do these naive robots, when we are used to calculators giving the right result every time, give such an egregious wrong result every time? We can’t say it’s purely the nature of the machines physicalism, if we take the scientist robots and their dismissal of the naïve robots consciousness claims to be the case.

Looking for an explanatory route, the scientist robots begin to refer to something that probably should have been checked some time ago. The mutating robots actually began to see QR codes where there were no codes – the veins of a leaf, for the mutated robot, began to resemble a QR code enough that the processor used its somewhat (emphasis on it being merely ‘somewhat’) equivalence to an actual QR code. Even though it wasn’t the same thing and so the identification is at first a false one, it proves slightly more beneficial to see these previously invisible things in the hunt for QR energy codes.

Us, not wearing QR codes previously, were invisible. But now the robots had started to see the spaces between QR codes.

At first we seemed like columns of mud, then columns of mud that moved to no particular immediately discernable pattern. Eventually the robots processors started an analysis that likened these mud column things to even, perhaps, being like the robots. This becomes a common knowledge amongst the machines, though it’s finer details are argued.

And eventually this is the scientist robots ‘in’. To show the naive robots that they are…robots, the scientist robots turned to us…and to the naïve, explained our evolutionary history, the Darwinistic pressures on us, our hunt for caloric energy. How we had to work in packs. How this required displays from one individual of us to be able to successfully trigger a response in another individual of us. How this sequence is breakable.

They used the way the robots treat us as ‘others’, as mud things and nothing to do with the robots, as a way to turn that upon the naïve robots and their superstitious claims. To explain how the mud men could continue after generations, but would put little into actually understanding each other beyond what simply worked in terms of energy gathering. They asked the naïve ‘think of it from the mud mens shoes – what else would they do? You say ‘they’re just mud’. Exactly, so what else could they do but lie to themselves? Without the prosthetics of science what else could they do?’

And the left it to their fellow naïve robots, the ones who just couldn’t stop saying they had consciousness, experience and qualia – which as you would say, is ridiculous - to explain how the mud men could ever do any better than, by default, lie to themselves about their own nature? Why would they do any better than that, the scientist robots would ask, when the other robots would say that the mud men are just mud. Carbon. Why would they have any extra capacity that allows them to tell they are just wet mud?

Though the scientist robots would point out the mud mens scientists as well, and how they don’t claim consciousness, experience, qualia, or do but with radically different definitions from the naïve mud people.

The scientist robots put this to the naïve robots, the ones we know are infuriating for insisting they have experience and qualia, and the naïve robots could not think of a reason the mud men would just, by default, understand they are mud.

But then the naïve robots stated and asked : ““So…so what? It’s a parallel to robot behaviour – in regards to us, that doesn’t mean…”

The robot scientists refuse to answer it, saying instead ‘Put yourself in the mud man’s treads/shoes…they lie to themselves. The mud men claim they have qualia, just like you claim”

“They’re just mud! They don’t have any of that!”

But what else could they mud men say other than that, asks the robot scientist to the naïve robots? The robot scientist says “I know, it’s appalling how they claim these things!”

And the robot scientist added “And they act just the same way as you – they think you are the pretender to qualia, not them!”

“Absurd! BLATHER!”, cried the most miserly of naïve robots, who’s mutant heritage budgeted the least processing power to speculation thinking, in true Dunning-Kruger style.

And a few, a scant few of the naïve robots begin to see how this other, these mere columns of mud…they begin to see how mud can end up lying to itself. Lying to the processes they consist of. And looking at it from the muds perspective, how the mud could do no different – and how the scientist mud men could at least acknowledge the truth of the matter that the naïve robot knew clearly already, but only through their scientific prosthetics.

And these scant few robots began to wonder ‘What…what if this applies to me? What if my claim of consciousness, experience and qualia are just more of the same thing…’

Just as you needed them to finally admit.

‘…the same thing that the mud men engage in?’

No comments:

Post a Comment