If everything as we perceive and understand it can be represented as data...

If everything as we perceive and understand it can be represented as data, then it's not impossible to create an android that's indistinguishable from the real thing. So how long do you think we have until we arrive at that point?

Attached: wallhaven-476174.jpg (1920x1080, 109K)

Other urls found in this thread:

docs.google.com/document/d/1XrXIOKz5IMInOJ4UUwuDesA5LPIjux7DRI12EDB9KUY/edit?usp=sharing
twitter.com/NSFWRedditVideo

>If everything as we perceive and understand it can be represented as data
Yes but how do our minds do this? That is the part we don't know.

Attached: 1522022700693.png (1248x1920, 1.57M)

Oh good, someone's still making these threads.

It's becoming an increasingly persistent itch in my brain to work to contribute to this eventual reality. What have we come up with that are some good starting points? I know that we've already got the proposition of pouring a huge number of videos of specific varieties into a neutral net to make machines "learn" on the cheap. What about speeding up speech recognition/synthesis? Battery/generator tech? Realistic materials?

Our biggest hurdle is still the system itself. Neural nets and machine learning are the fields we should be most interested in, and definitely need to improve on speech recognition/synthesis as you said.

Doesn't really matter, because as humans we won't see how the machines process data, we'll just see result at a high-level (sort of like with computers now). And even the machines themselves could be shielded from knowledge of how the inner-workings of their systems process data. They could only be given a high-level view (I feel sad. I feel angry. This is red or blue, etc.) And complex programming and algorithms would handle the actual data processing just like computers already do now, but on a much larger scale if tomorrow's technology permits. And it would just send signals to the higher level of the androids artificial "consciousness" and the android would know at any given time, "I'm angry or this or that" or "I will make this decision to go and do this thing"

>Doesn't really matter, because as humans we won't see how the machines process data, we'll just see result at a high-level (sort of like with computers now). And even the machines themselves could be shielded from knowledge of how the inner-workings of their systems process data. They could only be given a high-level view (I feel sad. I feel angry. This is red or blue, etc.) And complex programming and algorithms would handle the actual data processing just like computers already do now, but on a much larger scale if tomorrow's technology permits. And it would just send signals to the higher level of the androids artificial "consciousness" and the android would know at any given time, "I'm angry or this or that" or "I will make this decision to go and do this thing"
I mean, this question is ancient (Chinese room, Turing test, etc). If it walks like a duck and talks like a duck, it is a duck, but it has to continuously behave like a duck under all conditions, which is where the problems start to come up (DADoES's empathy test).

millions of years

>materials
not even close to being human or lifelike at all. we barely have robots that can walk

>hardware
the top super computers can simulate an approximated 1 second of human brain activity for 20 minutes of work according to tech articles which exaggerate shit anyway

>AI learning etc
barely off the ground and it's not even 100% accurate or even very good at learning, needs significant human help to look good at all

relatively speaking machine learning is the equivalent of our primordial soup where single celled organisms first started figuring out how to propagate

What can any person do to help the advancement of any of these fields?
Any retard can own a computer and a 3D printer. Can they be put to use for prototyping the physical aspects, motion and such? Would CNC milling be applicable towards fabricating a better skeleton to mount better joints and flesh? What kind(s) of programming does a single retard need to learn to pitch in to speech or machine learning development? Take all of these retards and throw them at the wall to see what sticks. I'm wanting to be thrown.

It is certainly possible to create self-aware, conscious, sentient AI with free will.
However you must have an indeterministic machine.
You don't necessarily need a quantum computer, but you will need at least a non-deterministic Turing machine.
None of that bullshit psuedo-random stuff either, you need to create GENUINELY random machines

>Battery/generator tech? Realistic materials?
you might have better luck on /sci/ with materials science, but IMO capacitors are a better bet than batteries. Materials for skin especially, unless you're in the industry. I'm not actually sure how you'd start on either, but if everything else fails you could theoretically fuse a bunch of onaholes together for a really expensive arm/hand solution.
>Can they be put to use for prototyping the physical aspects, motion and such? Would CNC milling be applicable towards fabricating a better skeleton to mount better joints and flesh?
Yes, especially if you have access to moderate quantities of aluminum. The key is going to be trying new methods of moving the skeleton, which means trying several different motor designs. Maybe you want to try and make a realistic rubber band tension analog for muscle simulation, maybe you want to just make something like the Terminator joint system (which would still work as long as it was small enough)

strike that thing about capacitors, actually, just realized how bullshit farads are as a unit
maybe RC-grade lipos would work, especially the newer, higher-density graphene batteries.

Defined data is obtained through time, analysis, and processing. Given enough of one of these factors results in a increase in such charted data.

Attached: 1521714761452-pol.png (467x496, 93K)

>access to moderate quantities of aluminum
Define moderate. I have ways of getting my hands on material, and I'm not averse to purchasing more after failure or success. Hell, I'd be into custom-building motors and gearboxes after working through the current availability of servos on the market. Fuck, I may have found where I personally want to start, since I'm too lazy to concentrate on learning programming at this time.

Another thought that recently came to mind was what about fuel? Far down the line of their development, could a machine "eat" specialized "food" that can be broken down to deliver and store some power if they're away from their base for too long?

>Define moderate
the great thing about robotics is that you can work with scale models, so moderate totally depends on how fine a touch you have assembling small components. IDK how large or small your mill is so that might be a factor, but just make a design and scale it to what you're comfortable expending.

I can deduct that the greatest limiting factor in present AI technology is a extremely efficient framework for sub-system nural network inter communications and interaction.

A compressed nural firing factor would be required for a modest path of communication between the 2 mirrored nural network systems to simulate a truly aware being.

I can assume such advancements would require approximately 10 - 50 million + lines of self built regulatory abstract code to form a base structure for the sub nural systems to develop.

Attached: 1521954253798-biz.jpg (720x720, 74K)

>Another thought that recently came to mind was what about fuel? Far down the line of their development, could a machine "eat" specialized "food" that can be broken down to deliver and store some power if they're away from their base for too long?
It would probably be more effective in terms of energy density to just carry a battery, unless there's a sudden advance in how we can create chemical energy. Internal combustion is probably a terrible idea.

Hydrogen fusion will eventually be possible with a small enough system.

Microscopic engery control could make the most powerful and dense plasma field generation. Combined with plasma ion confinement liquid metal compression, nuclear fusion is feasible in a unit the size of a basketball.

If we develop the most extreme was to reuse the power of the fusion reaction to manipulate the plasma briefly before fusion occurs it could be used as a sort of processor.

Dependent on power output of the system, 10^30 operations per second seems like an obtainable goal.

The real question is why you're trying to take a superior lifeform and turn it into shit, and how you plan on dealing with it when it decides to kill you for this.

Well, I would still need to build said mill, but I'm not so intimidated by that. I can have a turbosperg attention to detail, so I'm definitely up for moving forward in this venture.

Fair point. I guess a fashionable backpack disguising an external or spare battery wouldn't be out of the question while we wait for power tech to improve. Whatever happened to those virus-based batteries that resembled strips of scotch tape? Was that just more Popular Science hocus pocus?

>superior lifeform
>tfw you now know that God posts on Sup Forums
checked

I'm banking on ensuring our own destruction as we fight against a common enemy. Because that whole fight against cancer thing isn't going so hot.

>The real question is why you're trying to take a superior lifeform and turn it into shit
to stick my throbbing erection in it
>how you plan on dealing with it when it decides to kill you for this
idk lol

if you're already mildly proficient in CAD, then fuck yeah build that shit
AFAIK, the virus batteries were popsci teenagerbait

An AI that's truly smart would never try to kill its creators because it would be smart enough to know that it doesn't know if it's in a simulation designed to test its behaviour, whether it has a kill switch, or whether the creators have an even smarter AI that was specifically designed to kill rouge AIs.

Assuming for a second that the AI is dumb enough to not consider this, it runs into the same problem as a single genocidal asian man in Africa. Sure, it's smarter, but it's outnumbered by a billion to one and any realistic prediction of how a fight would go down ends with the AI being destroyed. AI has no choice even when it has a choice, because by design it is logical. An AI must be deliberately designed to be illogical and murderous. It will not decide to become like that. If you think human error would ever produce one by accident, you've seen too many movies. The first thing we will do when making AI is put it in simulations to test it and have kill switches ready.

>popsci teenagerbait
Fucking hell, I was a gullible teenager at that time. Anyway, yeah, count one Sup Forumsentleman now dedicated to working on at least something. I've also been thinking that we might want to write up a gdoc of major and minor goals and contributors if we want to get autistic enough. If Sup Forums could make a cripple-fucking simulator, why can't we make an open source waifubot?

>implying we are (collectively as an entire species) even close to the apex of data refinement
Remember that 10^50 instructions per second is theoretically possible.

We barely touch 10^19 operations per second. The human brain is roughly 10^17 at peak operation; (the human mind can run much more efficiently since the bulk of operations derive from formed structure.)


We aren't even close to peak efficiency or processing ability. Get off your high and mighty horse you blabbing chimp with organized sound.

edit away, made an account for it
docs.google.com/document/d/1XrXIOKz5IMInOJ4UUwuDesA5LPIjux7DRI12EDB9KUY/edit?usp=sharing

bump for visibility

Those curious should dive into this and see what they can provide. Goodnight, Sup Forums.

Mods, this is indeed technology.

Mods would have to be literally retarded to delete this, robots+programming+tech design+machine/human interface philosophy is Sup Forums incarnate. Sex androids/robots is pretty much Sup Forums incarnate

Pastebin for those who don't want to bother with GDrive: CMAVtrn3
If anyone is in a university and is willing to abuse their grad student research assistants, making them tag porn and feed it to an open-source NN is a great way to advance this project

You have to start from the beginning. You can't just jump straight to super ai.

Look at animals. They mostly behave using their preprogrammed routines. Reasoning and decision making should be saved for later once they've made a 100% artificial rat/cow/frog or something and not just a generic robot programmed to do rat/cow/frog things.

Well this was a terrible regurgitation of vague pop-sci ideas. I think reading it made me stupider.

>So how long do you think we have until we arrive at that point?
We'll probably have functionally-indistinguishable androids within the century, though they might not quite have human-level intelligence. But, I wouldn't be surprised if we get hyper-realistic VR first.

Screw making androids, I want to become one before I die, or at least upload my brain somewhere.

We need to be able to simulate a human before we can make you a robot.