Quote from Ash

"Will ever be able to create"? Why so certain?

Quote from Zed

For those who don't read big paragraphs:
-anything physical can be reproduced

Quote from Zed

I confess, it's only theoretically possible, but it seems irrational to say that we will never be able to do something if it can be done in theory.

Quote from Exilement

It's not. Replicating human emotions would require virtual perfection, because even tiny errors in a human brain create less-than-tiny defects. We don't even understand the human brain cohesively, so it's completely rational by now to say that we'll both acquire absolute perfection in both understanding the human brain and how to build a synthetic one.

Just for an example, look at Rubik's cubes.. it takes somewhere around 20-22 moves to solve any configuration of the cube, and the world's fastest speedcubers could probably do that many turns in 2-4 seconds. That would be considered the theoretical minimum solution time, but the best we've come to is somewhere around 7-8 seconds. Do you honestly think we'll ever reach that threshold? The answer's no, simply because of human error. Humans aren't perfect, and that translates into not achieving theoretical perfect situations in anything but the most simple of tasks, and synthetically replicating a human brain is not one of those things.

The only thing we can ever hope to do is come close, but as I said, human brains with defects create catastrophic differences from a normal brain, so how close would we really be getting?

Quote from Ash

Arguing that humans are imperfect sounds to me like a cop-out, because modeling the behavior of the human brain doesn't require so high a degree of accuracy that we should call it perfection.

After all, we aren't talking about replicating the exact atomic structure of one specific brain, that's just excessive. We're talking about recreating the behavior of a brain to the point where it is indistinguishable from a human.

I don't think the fact that the humans doing the poking and prodding are imperfect means that it's automatically destined to fail.

Quote from Exilement

Err, I'm pretty sure for a fake to be indistinguishable from the real thing, it has to be about as close to perfect as possible.



Are these two things really that different? For something to recreate those behaviors it needs to receive, interpret and process any kind of stimuli, new or familiar, in the same way a human brain does. As far as I'm concerned nothing is going to be able to mimic that in any indistinguishable way without trying to replicate the atomic structure of a brain.



I'm already convinced that it's never going to happen, nothing you've said has made me think otherwise and until it does, human error is more than enough of a reason to think it will. It's not a cop-out at all, especially since trends agree with me. I can't think of any absolutely flawless things humans have created, and if there are any, they're things whose functions are insignificantly as complex as a human brain.

For the record I do believe we'll come very close to mimicking the human brain artificially. But no matter how good a mimic is, it's still distinguishable from the real thing if you throw enough stimuli at it. What you're talking about is perfect replication of human behavior, and since we don't even understand all the processes that go into it on a natural level, I refuse to believe that we'll ever reach a point where we'll be able to replicate them perfectly on a synthetic level. Though as far as I'm concerned "indistinguishable" is a bar that you've set way too high.

Quote from Ash

Exilement, you've taken the usage of the word "indistinguishable" and conflated it to be so much more than we meant. You KNOW that no one was talking about 100% perfection ANYWHERE in this thread. You KNOW that. It's like if I were to ask someone to create a perfect replica of the White House and you were to point out that a particular amoeba is missing from a specific cockroach.

Indistinguishable means indistinguishable to someone not spending 40,000 man-hours and a couple hundred thousand dollars on tests, indistinguishable to any given individual.

Quote from Zed

The fact of human error doesn't mean this can't be done eventually. Let's take your Rubik's cube example.

Quote from Exilement

I used it in the context of its own definition, don't tell me that I KNEW that you meant it as something else. When you say "AI indistinguishable from human intelligence" I take that to mean "incapable of being perceived as different", which means it's an exact, perfect replica. Excuse me if that was some sort of profound assumption.

Okay, so what part of my argument made the task so much harder than this? Is there something about recreating all of the processes of a brain that deal with emotion that wouldn't go into creating AI that's indistinguishable to any given individual? I'm legitimately confused at where you're getting at here.

If you just think that we can program all the possible scenarios and hope a robot can mimic the human brain's functions without actually replicating them, LunarDeath covered why it wouldn't work very nicely.

Also, ducking behind hypothetical technological advances that are beyond anything we can even perceive now (atomic/molecular technology being autonomously rearranged according to medical scans?) really doesn't help to prove anything. Even I can acknowledge that yeah, technologically we see breakthroughs every day, but it really doesn't help your argument to say it might someday apply to this.

Quote from Ash

You kind of ignored my white house analogy.

You don't seem to understand the difference between replication of the physical PROCESSES of the brain and the physical STRUCTURE of the brain.

That's not how programming works. When making a program, you don't make a table of every possible circumstance and the appropriate reaction. You create algorithms that put data through logic.

Basically, it would have to want to have to **** the shit out of that blonde, or perhaps be gay, or maybe have the personality of a serial rapist. The same elements that go into human behavior would have to be included.

Quote from Exilement

Mostly because I don't recall ever trying to say that the only way to recreate a human brain artificially would be by creating an anatomically correct physical substitute along with all the other working processes. I'm focusing on synthetically recreating the processes of how thoughts and emotions are carried out in the brain, not on creating little machines that are physically identical to a neuron to make it look like a brain when all is said and done. Which is reinforced by this:



If I ever mentioned "structure" (which I don't recall doing but I'm not about to comb through my previous posts) I was talking about the structure of the processes the human mind goes through in order to create thoughts and emotion. It's not something that I think can be duplicated with a different type of programming in any convincing way.



Again, you misunderstood what I said, but I can see why, I worded that really badly.

However, no, that's not how programming works either. You don't create algorithms that put data through logic, you create algorithms that feed variables into data. That data analyzes the variable based on factors, and as long as the variables meet certain factors, it spits out predetermined actions, those factors/actions being as complex as the coding that goes into it. I still recall the debate we had about what logic is and I'm amazed that you were willing to throw that word around like you just did. Nothing about programming is logical, it just compares incoming data to what it already knows through what was programmed into it.

Which is what I meant by "situations", forgive me for the terminology, but you need to create enough core data so that incoming situations or variables that this AI would experience would create a realistic, human-like emotional response. And quite frankly I don't think that's possible, given the infinite number of variables and the complexity with which humans interact to them.

Thanks for the comparison to the digital watch but I was in no way saying that if the AI was to ever react to a piano being dropped off of a construction crane onto a Puerto Rican's fruit stand carrying papayas but no mangoes on the 13th of February at noon in San Francisco, the AI would have to have a pre-determined response programmed into it, and if it didn't it wouldn't meet your criteria of "indistinguishable". For all the years I've been here I would think you would know by now that I'm smarter enough to suggest that, that's completely idiotic.

Okay, so program ideal facial ratios and characteristics into an AI and it might be able to do those first two things, we already have facial-recognition software. Those are simple variables. The problem is the kinds of variables that are simple to humans but would be abstract to AI. Humans have years of context to support their reactions to stimuli, humans have physical responses, humans are conscious about how other people will react and how their actions will effect the future. These aren't things I ever see an AI doing, and these are elementary things that the average person would probably be able to see through.

I just don't see why you think human experiences are something that can be both represented in lines of code and plugged into an AI with the expectation that it'll accurately replicate human behavior, even on a superficial level.