JITOTM41: Simulated Consciousness
It's the year 2030, and Moore's Law still holds true: in fact, computers have become so sophisticated that it is now possible to simulate human consciousness perfectly. So perfectly, that it's indistinguishable from the real thing. So it's possible, by inputting really rich data, to simulate a past environment in complete detail, effectively endowing sim-like characters with their own consciousness.
That's the sort of scenario envisioned by philosopher Nick Bostrum. Cool stuff, eh?
Now... suppose someone simulates, say, the month of June, 2007? Suppose they create a couple of characters called "Jay" and "David." Suppose those two characters have a simulated discussion, using their carefully crafted artificial intelligence to optimum effect...
Just how, exactly, will you, the listener, know whether this is the real deal, or not? That's the question in this episode of JITOTM, Simulated Consciousness. Trust us if you dare!
posted by David Newland at 4:45 PM
7 Comments:
Hello Jay and David,
I wanted to post a comment letting you both know how interesting I find your podcast and how great it is to find that amongst all the crap that is predominantly being published in all these wonderfully 'advanced' new mediums of communication, something positive (and Canadian) is finding its way into the public space. I have listened to your show for two summers now but have heard little in terms of attempts to understand concepts of consumerism and the human brain. I think that this would be a facinating topic to explore, especially in light of the frequent talk of late regarding the consequences of Western consumer (and industrial) societies' impact on the planet as a whole. Discussion of whether or not each of you believe the human consciousness - and particularly the consciousness as it exists within a democratic, capitalistic framework - is even CAPABLE anymore of recognizing the fundamental disconnect that today exists between nature and civilization. This may seem like a fairly cultural, and not so psychological/social issue, but coming from an interdisciplinary background I find it difficult to disentangle these very different, but at once strikingly interdependent fields of study.
Hope you both have a great summer, and thanks again for continuing these podcasts,
Brennan McConnell.
11:18 PM
Here's the argument as I understand it:
1. In the future, computers will be capable of simulating whole worlds.
2. In the future, some people will be interested in simulating *our* world.
3. Therefore, our world is possibly (or perhaps probably) a simulation.
I think there is a fallacy here. The conclusion depends on the premises, right? But the premises depend on a very important *unspoken* premise. And if that hidden premise is false, then the other premises are false. Hence the conclusion is unsupported, right?
The unspoken premise of the first two statements is that this is the *real* world. That in the *real* world future computers will be capable of such simulations. That in the *real* world (this world), people in the future will etc.
I think the conclusion ("This is not the real world") disproves the hidden premise ("This is the real world").
-Dan Gibbons
6:48 AM
Hey Jay & David,
Im in a robotics and automation program. Ive always had in interest in robotics since seeing these machines on a show called Daily Planet when I was in gr.5
Many books and movies deal with the subject of robotics and consciousness. Blade Runner,A.I. so on and so forth.
Ray Kurzweil has a book called "The Age of Spirtual Machines: When Computers Exceed Human intlegence" Its a facinating read, I dont have it on me to reference. I hope you've come across it. It starts with the begining of the world and evoloution, Moores Law and much more.
Ive decided that robots would be a new form of life. Not life as we traditionally define it. These machines have a function, they need energy they act, react; they interact. They make some sort of sense of their surroundings. Essentially we are robots. "organic automatons"
Like the Raccoon in the tree. consciousness varys between people right? If we both look at an object and interpret it differently and are getting different meanings from the same thing, so are we looking at the same thing?
And one last paragraph: Cybernetics is a combination of sociology,psychology,Physiology and robotics.
So mapping the brain and copying neural-nets can be a kind of simulation the brain. Once a computer establishes the proper connections, learning, growing, evolving. Over time this machine would have a vast amount of knowledge if not intelegence. And like eveloution, there's a law called The Law of Accelerated Returns. Basically it talks about exponential growth. So after ten years this computer is a genious, we can make exact duplicates of it.
I enjoy your pod-casts. Look foreward to hearing more.
12:21 PM
Hey Jay & David:
Just wanted to let you know that I started listening to your TOTM podcasts a few weeks ago and have been enjoying them tremendously.
Thanks for producing a lively and interesting podcast. (Don't worry Jay, I'll still buy your books and continue to enjoy them, but these podcasts do add a whole new way to appreciate your work while walking, driving and doing other tasks that make holding a book in front of me a bit difficult)
3:25 PM
I listened to your very interesting show today. It seems to me that you've severely under-estimated the complexity of the situation.
Let's assume that in a few decades we'll have computing power equal to that of a human brain. For this scenario, let's further assume that we've managed to figure out how to make an approximation of an individual human intelligence with these computers.
What we've then got is a single individual. We've just gotten to the point where we're capable of doing this, so it's state of the art.
How much farther are we going to have to advance after that in order to build a whole artificial Earth? And then populate it with billions of people, as well as animals?
That kind of advancement seems like it might take hundreds or thousands of years, if it's even possible at all. There are theoretical limits to how powerful you can make computers, creating an entire populated artificial Earth may be well beyond human capability.
Perhaps it's possible that somebody created this simulation, but that not everybody in the simulation is real, and not all of the places in the simulation actually exist at the same time. It could be that most people are just NPC's with sophisticated programming but not actually sentient.
So, if you believe that you are, in-fact, sentient... then you've got to wonder if you're only one of a handful of people who are. How many would the user want to give sentience to? A single family? A community? A whole city?
If there's only a single user from the real world, you may be one of the handful of people he or she has chosen to create. If that's the case, you likely know the only real person in this virtual world and interact with that person on a regular basis. That's an interesting thought.
The movie "The Thirteenth Floor" is a very interesting look at a situation where an entire historical city is created and none of the inhabitants know that their city is all that exists. It's an interesting take on the idea that's quite a bit difference from the more popular "Matrix" movies.
Even an entire city would require an incredible amount of computing power though, and it's more than just being able to create human and animal intelligences. In order for the simulation to be realistic, the computer would have to be able to keep track of every blade of grass, every pebble, and every grain of sand in the environment.
Even if the computer only manifests those items when somebody is looking at them (if I look away, does that rock still exist?), that's still an incredible amount of computing power, and it will still require an enormous amount of memory to keep track of the things that aren't manifested.
And this is all assuming that we'll even be able to duplicate or approximate human sentience. It's more than just having the computing power to hold the information. Will we ever figure out enough about how the human brain works and what sentience actually is to really do something like this? I think that's an open question.
10:18 AM
Doesn't the advanced state of technology discussed in this show imply that the "teenage programmer" would have augmented intelligence? If so, then isn't it possible (call me pollyanna) that the teenager would realize that it can only hope to exist unmolested if as a minimum condition it left all others unmolested? And controlling sentient beings is the most profound form of molestation I can imagine.
What if evolution usually results in intelligences that go "off the grid" when they advance to the point at which this realization becomes apparent? Then a trillion times races of superintelligent beings would've either evolved past the illusion that existence is worthwhile or they would all be present right now, perhaps superimposed upon our environment and leaving us entirely alone.
6:48 PM
hey guys. I'm wondering where the discomfort around this thought experiment comes from.
My first thought was that it had something to do with the underlying assumption that you alluded to at the beginning of the podcast - the assumption that consciousness is epiphenomenal, as David put it. I think many people are uncomfortable with that assumption (perhaps even only at a subconscious level - haha), and that is the root of the discomfort with the end scenario.
when will computers be as powerful as the human brain?
is a different question from:
how does computing power relate to the subjective experience of consciousness?
My other thought was that when I compare my feelings about the 'real world' vs. the hypothetical virtual world presented in the thought experiment, I realize that I have an expectation of heightened determinism in the virtual world relative to that of the 'real world.' There is a certain amount of indeterminacy present in our experience of the 'real world' which is different from that of our designed computer systems (which incorporate data checking and redundancy to root out undetermined data). A simulated world, as we currently culturally perceive it, would be much more deterministic than we're ready to accept our 'real' experience is, and I think that this discrepancy could also be the root of discomfort around this thought experiment.
Just some further thoughts. Thanks for your podcast, it has a very nice conversational tone.
-aaron.
6:52 PM
Post a Comment
<< Home