Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[fishboyliam]

Actually, we don't have moderators, we have exactly one admin who does all the moderating himself. And personally, I think that's a good thing. When you have some members of the community who have power over everyone else, you can end up with an unhealthy power dynamic.

Doesn't Three-Toe have some admin privileges as well? I don't seem him around a ton, but I'm sure he lurks a bit; that, or I haven't been around long enough to see him.

Also I'd disagree to the power dynamic thing; while it might cause some problems, a mod team would definitely be able too deal with the forum much easier, though A) Toady doesn't seem to want that (at least, not right now), which I respect completely, and B) Toady seems to be fine handling the moderation on his own at the moment. The community is simply to small for a power dynamic to be a net gain.

Anyways, not looking to get into a huge debate on the logistics of running a forum, I've got 0 experience to work with. Just thought I'd offer my opinion.

[Dorsidwarf]

Isn’t Mephansteras a mod in the mafia section?

Anyway, I saw someone saying here that the “hardware is irrelevant to the debate of sentience” as a rebuttal to the statement that dwarves “have” no actual autonomy and are all centrally controlled by one process.

That seems silly, like saying that the fact that a ventriloquist doll is being controlled by the ventriloquist is irrelevant to a debate over whether ventriloquist dolls are sentient (as they exhibit any mental trait they are controlled to)

[Reelya]

Yeah. e.g. there's a huge different between the conceptual state of "happiness" that's an emergent property of real brains and the type of "happiness" that exists in things like DF:

Code: [Select]

var happy = 1
var sad = 0
if sadness < happiness_threshold then emotion_state = happy
Here, "happy" is just an arbitrary label given to an arbitrary state, which is arbitrarily defined as occurring when one arbitrary number falls below another arbitrary number. It's no more "happy" or sentient than writing some code that re-orders eggs when the number of eggs in your fridge falls below some threshold.

Similarly, the process that assigns dwarfs to jobs is just a really basic version of the type of computer code that assigns tasks to different work queues in a factory. e.g. all that happens is each time a unit finishes a job, the CPU loops through a list of unassigned tasks and finds one that fits the allowed labors of the unit. It's not some ultra-complex task that could become sentient. It's just reading through a list of things and comparing values in the list against some number it's been given, a glorified database table query, basically.

[Enemy post]

...theres no particular logic to regard oneself as quite such a walking apocalypse...

You don't know me!

[strainer]

heh, well yes perhaps I overreached there

all that happens is each time a unit finishes a job, the CPU loops through a list of unassigned tasks and finds one that fits the allowed labors of the unit.

It is a bit more complex than that.

For those who believe sentience results with appearance of sentience (deep AI etc) the hardware is relevant only to its ability to create/host an appearance of sentience. Dwarves have non-zero score of appearing sentient - they each maintain hundreds of data points designed to be analogous to organic conditions and characters and to produce relate-able behaviours.

On the other hand if we believe conscious experience requires something so far unexplained to make it important enough to respect - give rights to etc, then there is no certainty that unexplained magic is applicable to non-biological hardware or abstract informational systems. If we even imagine a convincing enough entity within our own minds could it experience its own plights as we are able? Can a hallucination be sentient? I dont expect so.

[Dozebôm Lolumzalěs]

I don't have much to say about the majority of your post, but on the topics of sentient hallucinations, hallucinations meet some of the criteria for independent sentience. They're entirely held within the brain, of course - I'm not proposing dualism - but they can still be somewhat separate, on a higher level of abstraction, from the self; they seem to have a degree of agency (perhaps only as much as an animal, though); they often have motivations, etc. that differ from that of the self; and so on. I suppose it would depend on the individual hallucination, with some hallucinations (of objects, ferex) being entirely non-agents, others being semi-agents, and a few (of people?) being nearly full agents.

[Eschar]

I don't have much to say about the majority of your post, but on the topics of sentient hallucinations, hallucinations meet some of the criteria for independent sentience. They're entirely held within the brain, of course - I'm not proposing dualism - but they can still be somewhat separate, on a higher level of abstraction, from the self; they seem to have a degree of agency (perhaps only as much as an animal, though); they often have motivations, etc. that differ from that of the self; and so on. I suppose it would depend on the individual hallucination, with some hallucinations (of objects, ferex) being entirely non-agents, others being semi-agents, and a few (of people?) being nearly full agents.

This is very interesting.

[strainer]

I found that essay to be very fanciful and terrible taste at the end to use the picture of the deformed cat, a baby with the extra arm and conjoined twin women. The baby is presented as a "this:", and the women as "these two:"

[Reelya]

For those who believe sentience results with appearance of sentience (deep AI etc) the hardware is relevant only to its ability to create/host an appearance of sentience. Dwarves have non-zero score of appearing sentient - they each maintain hundreds of data points designed to be analogous to organic conditions and characters and to produce relate-able behaviours.

I don't agree. e.g. if you have a meter called a "happiness meter" that goes up when a pre-defined happy thing happens and goes down when a pre-defined sad thing happens. That's not actual happiness, it's just a single floating-point number. The meaning comes in when you the player reads that number on a screen next to a string that says "happiness". The issue here is that it's you the viewer assigning the meaning, externally, but the essence of meaning itself isn't internal to the simulation. e.g. "g is a goblin" is a semantic determination made by the player. The goblin is in your mind. What exists in the simulation is just a letter g.

The semantic meaning is what we ascribe to it purely because there's a text string "happiness" somewhere and when the number is shown on the screen it's displayed as "happiness: 70%". But, there's no underlying quality of happiness there, it's just literally the word "happiness" stuck against an arbitrary number. e.g. if we changed nothing but re-labeled the string "sadness" then the viewer would come to a completely different anthropomorphic idea of what 'going on' inside the dwarfs 'head'. However, nothing in the code is actually different. e.g. if we just change the word "happiness" to "sadness" so that the read-out reads "sadness: 70%"  instead of "happiness: 70%" then players will ask "why is eating good food making my dwarfs sad?". It's not "making" them feel anything. The simulation was literally running the same as ever, and all that changed was a single word on the read-out available to the player.

e.g. you talk about "relatable behaviours". but you can use the same argument to say that e.g. the Man In The Moon has true emotions because the appearance of the moon changes and sometimes the moon hides itself. The problem is that humans project sentience onto things that are categorically not sentient. e.g. by that definition we could say that volcano gods exist and are sentient. After all the eruptions and rumbling are "relatable behaviours".

The problem is that a 32-bit integer is not analogous to a real-world quality, by any means. They're basically like rolling stats on 3d6 in a d20-based system. Every stat is effectively interchangeable with every other stat, and all tests are rolled on a d20 against one of the chosen stats. if you have a "jumping" stat and have to roll against some score to succeed in your jump, then the concept of jumping exists in the players mind, but the innate "essence" of what it means to jump isn't actually encoded into the rules.

It's similar with mental qualities in DF dwarf minds, e.g. every dwarf can level up "leadership" but that's just an arbitrary labeled skill rather than measuring the dwarf's innate ability to lead. The problem is that this is going in the opposite direction as a human: for humans, we have intrinsic abilities, and we then measure them, coming up with a metric. But the metric is not the ability itself, it's just a very rough estimate of capability. Dwarfs go the other way: they start with the linear metric, then it's inferred as to what they should be able to do by that. e.g. if they have leadership: 5, perhaps 5 dwarfs will follow them, and at leadership: 10, perhaps 10 dwarfs will follow them. but this doesn't actually follow from them having good "leadership abilities". In fact, they completely lack leadership abilities no matter how high the skill goes. It's just decided that people follow them blindly if the score is high. So the trait "leadership" isn't even actually modeled at all.

e.g. where you want to get to, at least, if you want to e.g. claim that a dwarf is really a "good leader" and not just arbitrarily labeled with that is to have a dwarf who uses real-world tactics that we recognize as being the things a good leader does. e.g. if we get to the point where one dwarf gives a "rousing speech" and that speech is broken down into actual concrete things the dwarf said, that each have an actual effect on the other dwarfs listening, and this dwarf knows who his followers are, which ones are important and why, and talks to them for actual reasons related to plans that the dwarf has, then, that's a starting point for saying that you're properly modelling a quality like leadership. Just labeling a value "leadership" then rolling a % chance of swaying the crowd with a "speech" that's not actually defined, that's not "leadership", it's just dice-rolling against an arbitrary savings-throw.

[Bumber]

For the sake of argument let’s define “intelligence” as complexity of behavior.

Also bumbler, bacteria aren’t capeable of learning. Some can exchange plasmids though but this is out of the context of the conversation. Their adaptation is simply darwinian adaptation which is a natural law. Not a product of the bacteria itself “learning” behavior. Dwarves don’t learn either they are governed by a pre-written set of behaviors but those behaviors are objectively from a computing standpoint more complex then run and tumble.(which is simply a “biased random walk” )
[...]

I don't think that definition of intelligence is particularly meaningful in a discussion of ethics. Furthermore, bacteria do more than just walk. They breed, react to stimuli, release chemicals, etc.

I don't make the claim that bacteria can learn. My stance on the matter is that dwarves cannot be more intelligent than bacteria because neither of them exhibit what I would describe as intelligence. I would say that the ability to learn and adapt is prerequisite to a relevant form of intelligence. A basic Google definition of intelligence is "the ability to acquire and apply knowledge and skills", so that fits nicely.

The problem I have with your comparison between dwarves and bacteria is that the dwarves are not actually responsible for their actions. This is what I meant when I brought up hardware, and Reelya's puppet analogy puts it more simply. The DF main process runs all the routines, like a puppet-master controlling the dwarves. Bacteria are at least carrying out their instructions independently.

A colony of bacteria can at least approximate something like learning through natural selection. At least this puts the population on par with something like a neural network. DF dwarves don't even do that. Their behaviors are defined rigidly.

[GoblinCookie]

I don't have much to say about the majority of your post, but on the topics of sentient hallucinations, hallucinations meet some of the criteria for independent sentience. They're entirely held within the brain, of course - I'm not proposing dualism - but they can still be somewhat separate, on a higher level of abstraction, from the self; they seem to have a degree of agency (perhaps only as much as an animal, though); they often have motivations, etc. that differ from that of the self; and so on. I suppose it would depend on the individual hallucination, with some hallucinations (of objects, ferex) being entirely non-agents, others being semi-agents, and a few (of people?) being nearly full agents.

Yes it seems that devils are allowed to be called quite real as long as we call them sentient hallucinations instead, but why is everybody so scared of Dualism?  It is like they sit around the place telling eachother horror stories about how the Dualists will come and eat you for breakfast.   

Consciousness is not intelligence.  The ability of something to respond intelligently does not imply that the thing actually is conscious, hence the ability of computer AIs to make seemingly intelligent and rational decisions.  Exactly the same thing could be going on within the brain, various automated thinking systems slip out of the control of the consciousness of the person because they 'broke' the 'control interface', the only question then is whether these autonomous 'spirits' are actually consciousnesses themselves (devils are real) or whether Occam's Razor can simply have them work mindlessly like an AI without the same logic eliminating all of us. 

[strainer]

The problem is that a 32-bit integer is not analogous to a real-world quality, by any means.

That is the purpose of any analogy - to be analogous to something. Qualities of matter, roughness, hairyness, weight, etc - such things are quantised to some degree of numeric accuracy and used in formulations to produce a likeness of a system. They never amount to the things they aim to represent but they are by intent analogies of the things they aim to represent.

In Dwarf fortress if we render the stress counter as happiness, accounts of the dwarves experiences will be unusual in that regard. The purpose of the stress counter is to enable the units to act like they are susceptible to stress - it is to be a simple analog of stress as it is commonly understood.

The DF main process runs all the routines, like a puppet-master controlling the dwarves. Bacteria are at least carrying out their instructions independently.

If you believe that what runs the routines is important to the question of sentience (I do) then you have something as yet unconceptualised involved in generating sentience. If you had a convincing AI which ran on dedicated neuron like cores - inspired by familiar biology - that AIs hardware can be virtualised and simulated itself to run on a single simple core. All known human designed hardware, except perhaps qubits, can be converted to software (into to data) which is processable to create the same output as its hardware incarnation, by any Turing complete computer. Thats very solidly known. What is not known, is whether a sentient nature of the kind we value in our selves can be generated or induced within any hardware or software. If it might be, then I think virtual DF entities are perfectly suitable receptacles for a tiny little bit of it.

[Bumber]

Even if we were to accept the possibility of single-core sentience, DF is light years away from something like that.

To what extent are dwarves even distinct entities from one another, that their deaths would have meaning? The distinction only exists in the user presentation of the data.

[Reelya]

The problem is that a 32-bit integer is not analogous to a real-world quality, by any means.

That is the purpose of any analogy - to be analogous to something.

But my point was that it's not an analogy.

There are entire sections of websites on logic dedicated to "false analogies". the point is anyone can claim that anything is analogous to anything else. That doesn't actually convey any actual meaning or infer the transference of properties from one actual thing to the other thing.

e.g. the moon is like a ball of cheese. Because it is round. But if you draw the analogy any further out from that, you end up with a false analogy. It showed one property in common, but you can't just extrapolate that the things are similar-in-general because the analogy exists.

Similarly, a number can be said to be "analogous" to a quality (such as intelligence), because you can measure the number, and you can also (try to) measure intelligence. But just "able to be measured" doesn't mean the number and intelligence share any other properties. e.g. you can measure intelligence and you can also measure dick-length. Can you then infer that intelligence and dick-length are analogous quantities? Sure, people might like you more if you have more intelligence and they might like you more if you have longer dick-length, but that's already pushing the analogy too far: e.g. it's struggling to find something those things have in common, and even then, using very hand-wavey logic.

e.g. the act of measuring intelligence in a single number doesn't actually capture the essence of the thing itself. e.g. to say "a number is like intelligence because intelligence is measure by a number" is a faulty analogy. The logical flaw is clear.

[strainer]

 analogous (ə-nălˈə-gəs)►

    adj.
    Similar or alike in such a way as to permit the drawing of an analogy.
    adj.
    Biology Similar in function but not in structure and evolutionary origin.