Print

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

[MetalSlimeHunt]

If I throw rocks at children in a sufficiently complex pattern, the force of their spite will form an ethereal shitposter.

[chaotic skies]

Aren't we all ethereal shitposters to begin with?

[Max™]

It's a little different with computation devices.

A ridiculously simple CPU (far simpler even than what's in a cheap calculator) could model the most complex computer on the planet, given enough RAM. It wouldn't be fast, but it would in fact be a perfect digital copy of the supercomputer. And this isn't just conjecture: there are rigorous mathematical proofs that this is the case. Which is science².

If meat brains worked the way computation does, then a mouse could out-think Einstein, given enough writing paper to jot things down on.

I know all about Turing completeness and so forth, but a computer emulating more powerful hardware is going to suck at it, and be slooooooow, and a system of hardware capable of thinking isn't going to bear much resemblance to a modern computer.

[Starver]

Being slow isn't itself a problem.

https://xkcd.com/505/

And unless you're into some flavour or other of mind/brain dualism, 'thinking' isn't really that special.

[Max™]

Thinking is just not something which x86_64 chips running binary code are good at, has nothing to do with dualism nonsense, we designed computers to do tasks which meatbrains aren't good at, there's no reason to be surprised that they aren't good at simulating meatbrains.

[Starver]

A digital chip can, with enough knowledge behind the effort, be made to emulate a single neuron.  A chip (with enough side-memory, as already indicated, and as much additional research as required) can thus emulate many neurons, workng in concert. If 'thinking' is merely many (or many many, it makes no difference) neurons working in concert, then a suitably programmed chip with enough storage and enough 'time' (it need not be realtime) can 'think'.

It would actually be as wasteful to develop such a 'thinking engine' based upon precise emulations of biological mechanisms by electronic circuits by way of human-generated code, rather than shortcut the process and get the chip to achieve 'sentience' in its own right by ignoring the biological kludges behind the "take input, develop a worldview, consider the worldview, develop own ideas, imagine those ideas made real, take the best of these and convey it as output" sort of process (To Be Discussed!) and running it at an already low-level.

Like I could create a Tetris 'game' within DF, through pumps and pressure plates doing the emulation (and drawbridges as a 'display'), but it'd be more playable just to program the computer wirh straight Tetris code rather than have it run DF and through DF build the compurational engine.   But "Tetrising" is possible either way, albeit at vastly different speeds.

[forsaken1111]

The difference is we know how tetris works, we don't know how consciousness works. So emulating the only known example may be the best/only way to study it in action.

[Max™]

Yeah, I never said you couldn't do that, my point was that barring some magical souljuice required for sentience we will end up with computers powerful enough that they can support a fully sentient human-equivalent mind at realtime (or faster) clockspeeds.

After that point you get into the realm of more intelligent minds, minds which are qualitatively better at thinking than we are, which are about as easy to understand for us as it is for my cat to understand why I like to poke at my fondleslab or clackclackboard.

[Sergarr]

I'm not sure if such "qualitatively better" is actually possible, though. We, unlike cats, have access to powerful tools called "language" and "abstraction", which means that we can essentially externalize and combine our individual knowledge without any apparent limit. Not sure how exactly you can rise above that.

[Max™]

Exactly! You aren't sure how to make something which can think better than we do.

Does that mean we are the pinnacle of thought, or just that understanding a mind which can model your mind is not a trivial task?

[TempAcc]

And then we stumble upon one of the philosophical problems with superinteligent AI: can humans create things that are qualitatively better at thinking than themselves?

I say, not quite, but possibly. Humans can prob end up creating a human level AI, which can then improve itself due to not having the limitations of the human brain. Hell, superintelligent AI isn't needed, all we need is a group of human level AIs which can process information way faster than we can, and thus come up with solutions that a top level large group of scientists could, just, say, 1000 times faster. Imagine an AI collective copy of NASA, all running in a computer network, working 1000 times faster simply due to the fact the intelligences involved have no need to sleep, eat, or have families, and can process information 1000 times faster than humans.

Which, while not being a real superintelligent AI, is still pretty scary.

[Starver]

The difference is we know how tetris works, we don't know how consciousness works. So emulating the only known example may be the best/only way to study it in action.

Hence "(To Be Discussed!)"...

(I was addressing "going to suck at it" (depends on what we do about it), "be slooooooow" only a problem if you demand realtime results, or better) and "isn't going to bear much resemblance to a modern computer" (in my example, it's exactly a modern computer, physically, in fact I honestly wasn't even thinking 64-bit, and was seriously considering something i486ish or less, but obviously custom memory manager and FS to grant such a necessarily extended set of storage reservoirs to swap data in and out of).

(Ninjary "qualitatively better" bit has me half agreeing (we don't know what we don't know about 'higher minds than ours'), but the difference between cat brains and ours is 'firmware optimisations'. A cat neuron compatibly transplanted in the place of one of our own neurons won't force us to meow, but our own neurons being rearranged from human-layout to cat-layout in some vital subsection of the brain might well do. And cat can has cheezeburger abstract thought, as far as we're able to work out, this side of the communication divide.)

((Ninjas, ninjas, ninjas...  Hot topic... Or chocolate... ))

[Max™]

In addition, I don't know about you but *I* can't simulate a cat's brain. Primarily because I don't have the 'software' to do so. If you gave me the ability to, I probably could. I could probably do the same with a superior intellect if I had any idea how, and to do so I'd probably have to segment the superior intellect into parts. So say, I separate it into brain parts through 1-10, with a starting point of certain memories, processes and the like.

Well, I used the term model preferentially over simulate, it would be a lot slower to simulate than to simply produce a cognitive model of how a cat sees and responds to different things.

Similarly our fat little ninja Reggie has done a great job killing all the moles in our yard and the neighbors yards. He can't dig for crap, but he doesn't need to, he's able to see that moles occasionally come up and peek out of their holes in the dirt, and I can tell that Reggie knows that a mole knows if something doesn't move the first time it sees it, and doesn't move the second time it sees it, it probably won't move and is thus safe. So all Reggie has to do is sit still and wait the first couple times he sees a mole, rather than jumping instantly and losing his prey.

So they sniff around, don't see anything odd, climb out of the hole, and POW, Reggie'D mole.

It's easy for me to do this sort of modeling, theory of mind is something we primates have been making use of for megayears.

Modeling a mind which can model my mind (and thus the model of itself in my mind) is not easy.

[Sergarr]

Exactly! You aren't sure how to make something which can think better than we do.

Does that mean we are the pinnacle of thought, or just that understanding a mind which can model your mind is not a trivial task?

What do you mean by "a mind that can model your mind"? Actually, don't answer that, because I see you've already done so:

In addition, I don't know about you but *I* can't simulate a cat's brain. Primarily because I don't have the 'software' to do so. If you gave me the ability to, I probably could. I could probably do the same with a superior intellect if I had any idea how, and to do so I'd probably have to segment the superior intellect into parts. So say, I separate it into brain parts through 1-10, with a starting point of certain memories, processes and the like.

Well, I used the term model preferentially over simulate, it would be a lot slower to simulate than to simply produce a cognitive model of how a cat sees and responds to different things.

Similarly our fat little ninja Reggie has done a great job killing all the moles in our yard and the neighbors yards. He can't dig for crap, but he doesn't need to, he's able to see that moles occasionally come up and peek out of their holes in the dirt, and I can tell that Reggie knows that a mole knows if something doesn't move the first time it sees it, and doesn't move the second time it sees it, it probably won't move and is thus safe. So all Reggie has to do is sit still and wait the first couple times he sees a mole, rather than jumping instantly and losing his prey.

So they sniff around, don't see anything odd, climb out of the hole, and POW, Reggie'D mole.

It's easy for me to do this sort of modeling, theory of mind is something we primates have been making use of for megayears.

Modeling a mind which can model my mind (and thus the model of itself in my mind) is not easy.

Yeah, this is not "modelling". At all. You've simply "explained" its behavior in a qualitative manner and by using high-level vaguely defined concepts. There's a vast gap between that and actual modelling. By that standard, we all can model each other's behaviour, more or less (since that's how we function in a society - by predicting each other behaviors, or at least acting as if we do), and thus we're all "a mind that can model your mind", producing a paradox (since there cannot be minds that are vastly superior to themselves).

[Max™]

A model is simply something which can be used to explain and predict behaviors or outcomes. I can produce a model of how a cat thinks and responds to inputs, and I can fine-tune it through observation and comparison with actual cat behavior (I chose cats in my case because I've had close to 200 cats in my life) so I can "put myself in their paws" in ways which they could not reciprocate.

Reggie is not capable of modeling the sort of language and abstraction based thinking I do, so in that sense I have a more powerful brain/mind/whatever than he does.

We can comprehend the idea of a more powerful mind than our own, we can speculate about it, we can even work out various physical limits which might constrain such a mind, but doing the same trick of "if mind a sees this, it will expect mind b to respond in this fashion" that I can do easily with a cat/mole is not a simple feat for a more powerful mind than your own.

Those abstractions which are beyond cat-thought allow us to do things like postulate and discuss more powerful minds, but they don't actually allow us to be sure what those minds would think or do in a range of situations.