/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality.

Roadmap: file restoration script within a few days, Final Solution alpha in a couple weeks.

Sorry for not being around for so long, will start getting back to it soon.

Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB

More

(used to delete files and postings)


ITT: Anons derail the board into debate about Christianity :^) Robowaifu Technician 04/02/2020 (Thu) 02:24:54 No.2050
I found this project and it looks interesting. Robots wives appeal to me because i'm convinced human woman and humans in general have flaws that make having close relationships with them a waste of energy. I'm a pathetic freshman engineering student who knows barely anything about anything. Honestly, I think current technology isn't at a place that could produce satisfying results for me at least. I'd like something better than an actual person, not a compromise. Even then the technology is there, I have my doubts it'll be affordable to make on your own. Fingers crossed though. Anyway, what kind of behavior would you like from your robot wife? I'd like mine to be unemotional, lacking in empathy, stoic and disinterested in personal gain or other people. I think human woman would be improved if they were like that. Sorry if this thread is inappropriate.
Edited last time by Chobitsu on 04/06/2020 (Mon) 16:00:20.
There's already a personality thread. >>18
>>2051 Oh. Sorry about that.
>>2052 No reason to be sorry, we all have our awkward moments.
Open file (642.81 KB 1017x1196 press_on_and_win_anon.png)
>>2050 welcome anon, I replied to your question in >>2056. there have been roughly two dozen of us in the ~3yr-long history of the board that have mostly all come to the same basic conclusion as your engineer's mind have. namely, the tech just isn't there yet for this sort of thing. but as we outlined in the welcome thread, we're not content to just wait around for the future to hand us good things, but are actively pursuing the invention of these good things ourselves. what alternative is there really? just sit around playing vidya all day until (((faceberg & jewgle))) manage to issue our (((state-sanctioned))) robowaifus to us? this is a long and hard climb, but there are rewards along the way and the ultimate prize will literally change the world for the better if we manage to succeed at it.
>>2057 >what alternative is there really? Doing research? That also costs money though, maybe more than building anything DIY. There isn't a measuring stick of technological progress going towards "good enough" for an android wife either. Has a list of criteria been made for a "good enough" wife?
I'm sorry, but you guys are kinda naive. Even if you had the tech to make a real robowaifu, we don't exactly live in ancap where you can literally do what you want and get away with it as long as it doesn't directly harm someone. Even lolis and "sexist" videogames are getting banned by the modern-day equivalent of witch hunters and a grand inquisition, and you're trying to create something which would "make women obsolete"? kek, good luck with that >>2057 >(((faceberg & jewgle))) >(((state-sanctioned))) There's your problem. The only countries that come close to robowaifus are Japan and maybe also South Korea, Taiwan, Singapore, all very free-market countries where the customer is king. What you should be asking for isn't robowaifus, but the markets the freedom and the markets that will actually make them possible.
>>2059 >good luck with that You could have just left the post at that. "Political activism" is a much much bigger, worse waste of time. Good luck with "chaging the markets".
>>2061 Yes, politics fucking suck, but we've all been forced into it, blame the Jews for it, or the socialists, or the globalists, or the multi-national corporations, the elites, etc... It doesn't matter, they are all the same thing. They are the reason everything has to be political. If you aren't engaging in political change, then you're just being an escapist, avoiding the real problem until it catches up with you later if not sooner.
>>2062 >we've all been forced into it Political change is not the same thing as political discussion or circle-jerking. I cannot see how rag-tag tribes of internet shit-posters can change anything. It's just not rational. I'd rather do whatever I want without worrying about it until somebody actually comes to my house to arrest me. Ignoring politics is something I can actively do to make my own life more enjoyable with guaranteed, immediate results. The only only problem is when other people bring it up.
Open file (112.42 KB 477x640 20390589.jpg)
>>2058 Yes, research has been the primary focus thus far. Around half-dozen have done some practical tests/prototypes. >Has a list of criteria been made for a "good enough" wife? Yes, some effort has been made, but ofc this is very subjective to the individual. I personally think a Visual Waifu is the most likely to be our first success atm, as it requires little in the way of robotics hardware. >>2059 Yes, we're very aware of the threats out there to this technological advance. Machiavelli's The Prince certainly is pertinent in this situation. >>2062 >They are the reason everything has to be political Certainly that applies in the case of /robowaifu/s interests. In many and varied ways in fact. >>2063 >I cannot see how rag-tag tribes of internet shit-posters can change anything. It's just not rational. Lol, we're autistic anon, not rational. Many of the great men of the past who actually accomplished progress were of the same ilk. Again, what are the alternatives, just sit around and hope it will all work out?
>>2068 >practical tests/prototypes I meant more inventing something new. I don't know. It's still very impressive. >Again, what are the alternatives, just sit around and hope it will all work out? Be pragmatic and focus only on what personally interests you. Talking poltics only attracts nutcases who ramble about protocols and satan. Is there anything good about that? I'm I the one going crazy? What drew me here was the seeming lack of that and now I feel somewhat responsible for bringing it here. You can choose to not have that and be better off. It's both unpleasant and pointless. Let's get serious, nobody is poisoning crops or blowing up government buildings. I know nobody is actually going to do that either because they're not really serious.
Open file (311.23 KB 720x446 2016317154576130.png)
>>2069 >focus only on what personally interests you That's OK advice, as far as it goes. One thing that interests me very much is having a harem of anime catgirl meido robowaifus haha. > And don't worry about 'bringing' anything here anon. We're actually a very laid-back community here, one of the best in my opinion. Most things just roll off our backs.
>>2063 >rag-tag tribes of internet shit-posters No one said anything about rag-tag internet shitposters. >I'd rather do whatever I want without worrying about it until somebody actually comes to my house to arrest me. Bruh, if you're in a western country then you're already sitting under house arrest. The amount of things that you can do has been steadily decreasing over the decades, and if you don't notice that, then it's either you being the irrational one, or you're just too young to have noticed. >Ignoring politics is something I can actively do to make my own life more enjoyable with guaranteed, immediate results. Ignoring problems won't make them go away. You're sacrificing long-term happiness for short-term pleasure. You might as well become a complete hedonist instead bothering with something so far off as robowaifus.
>>2071 >No one said anything about rag-tag internet shitposters. So what are you then? What have you, you know, done? Why do you think you have any chance of success? I'll take you seriously when things start blowing up. Money where your mouth is and all that. >The amount of things that you can do has been steadily decreasing over the decades I don't believe you. And yes, I am young. Protestantism and traditonal values have done plenty to restrict people like me, for centuries. Them and their values don't work with mine. There's no place for me in a hypothetical, traditonalist society. I would the one getting my head blown off or sent to a camp. I've wasted enough time lurking on stormfront to know that. I don't want a god damn family and I want to consume my smut in peace. It's all a lie. As many problems as there are now, people like me have never been more free. You guys, liberals, bible-thumpers, the end result would be the same for me.
>>2072 Have any good ideas about robowaifus rolling around in your head anon? You seem interested in the idea, how do you think the process should be approached?
>>2072 >I'll take you seriously when things start blowing up Wow there! Again, who said anything about blowing anything up? Why are you trying to put words in my mouth unless you're a literal glownigger? I'm not too panicky about FBI monitoring these places, and I don't normally accuse people of things like this, but you're acting really suspicious trying to get me to say something like that, fuck off with that shit. Now if you're as rational as you claim to be, stop prescribing opinions to me which I do not hold. I am not strictly-speaking a traditionalist and I don't know where you're getting that from. The point that I made in the beginning is that with free-markets, like in countries such as Japan, you can get both your robowaifus, and your traditionalism, and they co-exist peacefully in the same country with no liberals, bible-thumpers, or whoever else going on witch-hunts. I honestly have nothing more to say, I feel like I'm talking to an FBI agent. Actually try to read people's posts before arguing with your own strawmen.
>>2075 I'm sorry about my neuroticism. Being so cooped up like I am now is unhealthy. I have ideas, but none of them are well-informed. I don't think any programming language is up to the task of making general AI. I actually don't think AI as a whole is in the state needed. The strategies just aren't good enough. Hardware is also an issue. Probably the bigger one. Just comparing the electricity the human brain to what a computer with traditional architecture needs to emulate it shows that. An entirely new "language" would have to be made, and it would probably have to based on neurology, which not a lot is know about. There's also the problem of AI safety. Has that been discussed? Even if it's smart, it wont necessarily be human. It wont absorb values or anything like that. People are built to intake specific information, so a baby wouldn't learn binary just by being exposed to it. We don't know how that works. With personality, people have relationships for the sake of entertainment and validation. An ideal robot companion would have to have different priorities. Woman especially have always been attracted to men for what they can give them and how they make them feel. They're not built to love a person just for being who they are. A robot wife would have to have an imprinting system, so she'd know what person to love. I cannot see how you would program a robot to "know" a specific person with C++. Their fingerprint? Retina? A password? That's not very personal. How would you describe a human being in any language that exists now? Even facial recognition has limits, especially considering how people grow older. I don't think humans see each other as just a collection of physical attributes considering how people can drastically change and maintain those relationships.
>>2072 >I want to consume my smut in peace What a sad existence
>>2076 I don't see any point in being an online "acitvist". Nothing will change unless something that drastic is done. Otherwise, caring at all is pointless. This is obvious to me. As for your specific beliefs, what specifc flavor you are, it makes no difference to me. Whenever I make this point, people call me a fed. I think you do enjoy politics because I can't think of any other reason for this behavior. What on earth do you think you're going to accomplish? >>2078 Sad or not, I like having the option.
>>2077 Those are all really good points. I think men are creative enough that given the determination and time, we'll solve much--if not all--of these issues. Decomposing big problems into smaller ones, a term in programming as abstraction, is really the only field-proven way I know of to tackle seemingly insurmountable problems. Of course this idea is pertinent to many other areas than software, but it's a good place to start. So given that basic fact, can you think just one of these areas to tackle and break it down into a smaller, more-manageable problem with better chance of a successful outcome?
>>2080 Well gee. It depends on what angle the problem is solved in. If we're going for replication, copying a primitive nervous system might be a first good step. Before making a robot wife, making a robot sea slug or something with the same material as the final product. How? Neuromophic computing. https://rebootingcomputing.ieee.org/images/files/pdf/4-rcs2-hylton_-_intro_to_neuromorphic_computing.pdf https://advances.sciencemag.org/content/4/1/e1701329 Anything beyond that is above my pay-grade. I apologize if i've made the atmosphere here worse. Maybe I'll come back when I can make more meaningful contributions. Regardless of what others say you should be focusing on instead, I think this is a meaningful goal to work towards.
Open file (87.58 KB 863x674 MZ04HgyKj2Q.jpg)
Open file (42.42 KB 1009x662 QkutDzHCMRw.jpg)
>>2079 What online activism? Who said anything about that shit? I don't get it, are you stupid or illiterate or what? Is English not your first language? What are you fantasizing about? Hahahaha, holy shit, having this conversation is like listening to giant pieces of polystyrene being rubbed together. >Whenever I make this point, people call me a fed. No, you stupid nigger, as if it isn't obvious enough, you're being called a fucking fed because you're trying to get people to admit to edgy and illegal things like "blowing something up", if you're not a fed, then you're a journalist, or maybe a clueless redditor/cuckchanner at best.
>>2082 >What online activism? Who said anything about that shit? >What you should be asking for isn't robowaifus, but the markets the freedom and the markets that will actually make them possible. >If you aren't engaging in political change, then you're just being an escapist
>>2082 >>2083 Guys, let's take the off-topic bantz to The Lounge where it belongs please. >>39
>>2081 I see. Hmm, I've often thought about trying to create a simulator of bird or even just a fish as a decent start to trying to sort out how to create a simulator as complex as something modeling life at all. So, how exactly would neuromorphic computing help solve these challenges? I mean can you give us a quick rundown on your take on it anon?
>>2085 >So, how exactly would neuromorphic computing help solve these challenges? They use substantially less power and their behavior resembles neurons more closely. A key component of human relationships comes from memory, so something that could form genuine human bonds would probably have to have a similar thought process, meaning a similar physical architecture. Because humans are the ones who would be making this system, we'd have a lot more control over its functioning. Maybe this would allow us to shape robots' way of thinking to our liking. These are all big assumptions. There's also the issue of cost. I don't know how easily any person could get their hands on this equipment.
>>2086 I see. I've studied GPU programming before and a) they seem at least vaguely similar to the neuromorphic processors I saw (hypothesized?), and b) they are readily available as commodity devices accessible to typical consumers. Do you think there is a reasonable chance of progress towards this architectural approach using GPUs Anon?
>>2087 Maybe. If you've got a powerful computer and lots of electricity, you could have a smaller body that's attached to the computer. So a "two body system" like how mushrooms do it. For a simpler organism replica, this could be a good solution. I don't know about the final product though.
>>2077 You're looking at it the wrong way. Every programming language is good enough to make a general AI with, though some are more appropriate than others. Shit, a sufficiently dedicated autist could write SkyNet in BASIC if he really wanted to. What really matters is the abstract layer. In this case, that means taking everything that makes a person (sans the shit you don't want), breaking those things down into little chunks, then arranging those chunks into a framework to build off of. Let's say you want to start with love. How do you program that? Easy: you don't. Instead, you break it down into logical components. For instance, I'd break it down into devotion, kindness, familiarity, and lust. How do you program those? Protip: you don't. Break down the broken down components into more logical components. Let's start with devotion. That can be broken down into loyalty and worshipful attachment. Loyalty is easy, just implement security measures that allow her to pick you out from anyone else. This has been done over and over again, so we know it's programmable. Worshipful attachment needs to be broken down a little more. First, we tie that back to security. If you're registered as the "root" user for your AI, than that's the worhipful part accomplished. She literally cannot ignore or refuse you. Attachment is similar, but needs a little more breaking down. So, we distill it down into interest. What are your interests? What are you doing right now? What do you think about X thing? This is not only easy to program, but essential. By giving her your personal data, she's literally forming a personality that uses your own as a framework. Make it so that she asks these questions, and make it so that she deems the data as important. Actually understanding you is an isue of natual language processing, which I won't touch upon here. There are many books on the subject, however. Now that devotion is programmed, let's move on to the next item. Kindness seems more difficult at a glance, but it can be broken down if you get creative. Let's break it down into concern, interest, and niceness. Concern is easy, just expand on the interest framework to include physical and mental health. Teach her about conditions which you can or do suffer, and give her data from trusted official sources to expand her understanding. Interest? didn't we already cover that? Oh shit nigger, we did! Now we can move on to niceness. Niceness is only as easy to make as you make it. What would you catagorize as "nice"? Let's distill it down a little more: how do you teach "niceness" to an autistic five-year-old? My parents never figured it out, but that doesn't mean I can't try. I, for instance, would take stories that I deem as having overtones of niceness, then feed her those stories with a note that the data processed from them should be catagorized as such. I would also show her instances of communication which is nice, and tag certain dictionary words as being inherently nice. You will need to make corrections along the way, but after enough time the framework should be strong enough that you won't need to hold her hand anymore. Then I take that catagorization tag and link it to anything I damn well please. Love involves being nice, so I link the tag. I can break it down further, but I'm already five paragraphs deep into this shit and I'm still not done, so let's move on. Familiarity links heavily back to interest, and it's both the easiest and hardest task. It's easy because you only need to feed her enough personal data to give Alphabet executives a world-rocking orgasm, but it's also hard because she needs to be able to understand the data. She needs to know how to catagorize the data you're giving her, and she needs to know how to understand the catagorizations. With a sufficiently-trained natural language processing framework it can be done, but it'll take a while to get to that point. Lust is amazingly easy to break down, it's just a value system. Give her a lust counter, program modules that alter the count, and program modules that activate when certain milestones are reached. As an example, image tagging and recognition is done to death in this day and age. There are even algorithms that can recognize porn (sort of). If she breaks out of her little sandbox and finds your "special" folders, then all those images will put her lust value through the roof. You can make this manifest itself in a variety of ways. For instance, if you decided to tag certain words and phrases as being "lewd" then she will use those more as the counter increases, which signals her lust to you. By performing certain actions with you (i.e. sex if she has a physical form, or erotic roleplay if she doesn't) her "lust satisfaction" counter will slowly move up until it reaches the peak value, where orgasm is achieved. After that, both counters are reset. Of course, other people may want to do it a bit differently, and there's some stuff I didn't really expand on, but all that's just an off-the-cuff example. Everything can be broken down into logical components, and the first step is to recognize the fact and discard the "muh mysterious biology" mindset. How you do it and what programming language you use is up to you, but it's all very possible.
>>2088 Yes, the notion of offloading expensive computations onto a home server cluster has been a common theme here from the very beginning. Ofc the idea isn't to create and exact simulation of a female from real life. I'm sure you're aware that's not only impossible computationally, but also infeasible/undesirable for other reasons. We want a satisfying approximation of catgirls, not a perfect replica of a 3DPD. >bleh< :^) It seems to me a reasonable notion that we could actually manage the behavioral system of this much scaled-down approximation using clever tricks and sound engineering.
>>2089 >Shit, a sufficiently dedicated autist could write SkyNet in BASIC if he really wanted to. this needs to be a /robowaifu/ banner tbh.
>>2089 >because you only need to feed her enough personal data to give Alphabet executives a world-rocking orgasm topkek. are you greentext anon? btw, this is certainly one of the more lucid analysis of the breakdown process for personality abstraction i've read here to date. good job anon. can you write software? it would be very interesting to see some of these ideas fleshed out in some code.
>>2089 This approach might work. Or it might get you killed. Defining everything, given a robot with so much autonomy and so many choices, might actually take longer than the biology approach. There's the ever-present danger fo reward-hacking. She could tie you down, get you to orgasm, raise her lust, and do it over and over again. If she sees it as an unecessary use of energy, she could delete your porn and disconnect the internet, or impair her own senses. She could kill you in a roundabout way, try getting someone else to kill you. She might try changing her programming to remove limiters, or resist her programming being changed. She might take too much interest in what you're doing, misconstruing the value of drinking coffee or using the bathroom to you. There's also whole issues of making her forgetting unecessary information and how that would be chosen. Any reward system is dangerous. The safest approach along these lines would be having a robot that doesn't know what it wants, you'd be the reward function. Training would be too tough though. >>2090 >Ofc the idea isn't to create and exact simulation of a female from real life Sure not exact. Something better would be nice though. What if that become a possiblity? Would you "upgrade" you robot wife? Would she really be the same?
>>2093 >Would you "upgrade" you robot wife? Sure, likely so. >Would she really be the same? Hmm, likely not. I don't think that would matter nearly as much to me as her improved abilities. There are a number of fictional accounts of humans themselves experiencing such 'upgrades'. As a tech guy, I'm always interested in more power and capabilities in my computing devices so yea, absolutely. BTW, is that a cute, sick, waifu that lives in a microchip?
Looking into comments ITT have led me to a couple of links of interest to me. https://www.humanbrainproject.eu/en/robots/ https://neurorobotics.net/
Open file (1.45 MB 1200x1200 37316971_p0.jpg)
>>2094 So you don't think there'll ever be a point where you get so genuinely attached you wouldn't want an upgrade? That exact thing happened in Chobits now that I think about it. Wife, sexbot, maid. I'd like to think those are different positions. I think a partner that's too transparently running on code would leave me feeling lonely once the novelty wore off. I don't know if humans have the capacity to break emergent properties like "understanding" and "appreciation" down into simple parts. Well, whatever. >BTW, is that a cute, sick, waifu that lives in a microchip? https://www.pixiv.net/en/artworks/37726483
>>2093 >She could kill you in a roundabout way, try getting someone else to kill you Somebody's been watching too many Hollyjew movies. In the third line of my post, I specified that I was only talking about breaking down the concept of love, which is only a small portion of what will ultimately be a very large and complex program. Safety features will come from somewhere else. Reward hacking is trivial to work around - just set a weekly quota for events, or write in an entropy algorithm that makes returns diminish with repetition. Altering her programming in ways that you don't want is based on the assumption that you give her the ability to do so, which shouldn't be allowed for the base code. If she deletes my porn, then, well, frankly I don't care. This is based on the assumption that she has system write permissions that are equal to or greater than yours, which should only be given on an isolated system with nothing else of value on it. How would she shut off my internet? The router, maybe, if she was given a way to access it and could figure out how to operate it, and even then that's easy to work around. >The safest approach along these lines would be having a robot that doesn't know what it wants That's the least safe approach I can imagine. An AI without a task will either do nothing, or make its own tasks to fulfill. How does it determine what task to do? Who knows! Also, how do you program yourself as a reward system without a comprehensive framework like the one I already layed out? That's also too vague - you have to break it down. Breaking everything down into logical components isn't just to facilitate her growth, it's also so you can understand how she works. If you know how everything connects together on an abstract level, then there should be no dangerous surprises. With a clearly designed system, you can understand everything she thinks and why she thinks it at a glance. >>2092 Damn, good eye. Yes, I am greentext anon. While I can design abstract logical systems in my sleep, I'm barely beyond the "hello onii-chan" stage of programming right now. That being said, I'm considering making a full chart for how mai waifu would work on an abstract level, which I will use a reference point throughout the project. If people are interested, I could post it once it's made.
>>2097 There's a lot of room for error, and the consequences could be quite bad. Sandboxing in any case would be a good idea, so making a virtual enviroment would be good. Reality simulation has limits though. >How does it determine what task to do? >Also, how do you program yourself as a reward system without a comprehensive framework like the one I already layed out? There's the concept of exploration. Basically, within some limitations, they do random things, with gaining information as the goal, and you would reward them. Just a good or a bad over and over again from you, and the AI will come to some conclusion on its own based on that feedback. Yeah, that's also dangerous. I'm not trying to be pessimistic for pessimism's sake. Maybe this a bad point, but if these issues were really so trivial, massive corporations would never have trouble with their AIs, yet they do. If they fuck up, why be sure you wouldn't? Also, if actual "general ai" were actually already possible, why doesn't it exist? What's preventing anybody from making a robot wife with no body? Concrete Problems in AI safety: https://arxiv.org/pdf/1606.06565v1.pdf AI Open Source Safety Gym: https://zephyrnet.com/safety-gym/
>>2098 I think security on system level should be a priority over safety on the AI level, especially in the early stages. Implementing safety features in an AI introduces herculean amounts of overhead, because you're going to have to manually program everything. On the system level, security is not only easier to accomplish, but it should already be implemented. Alternatively, you could have security on the hardware level simply by using a server that's isolated from your network. The people talking about AI safety are talking about corporate implementation and deployment, because the risks there are very real. In the home, on our home computers? You don't need all that complicated AI safety shit, just use basic security measures. As for why general AIs don't exist, the reason is multi-faceted. Firstly, there's virtually no demand, especially in the west. Fearmongering and has been keeping it that way since the 60s with the precedent being set by 2001: A Space Odyssey. There's also the issue of feminism, which artificially supresses the market both on the legal and social levels. Any corporation that attempted to make and distribute such an AI would be attacked mercilessly. Said corporations are also handicapped, because they do have to worry about AI safety, making any such ventures risky and ultimately not worth it. This leaves personal and small group projects like those on this board. Unfortunately, we can't do this shit for a living. We need to go to our shitty jobs and/or worry about getting degrees from shitty universities that will, in most cases, bury us under disgusting amounts of debt. All these factors combined means that such a project will proceed very slowly, and there are hundreds of possible vectors by which a project can be interrupted or cancelled. Remeber when 8chan went down? A whole bunch of people got lost after that, and it took forever for the community to repopulate. For many of us, motivation is also a problem. Yeah, I know, it sounds like a bullshit excuse, but it's there. Not all of us are happy enough in life to just get up and tackle everything head-on. If anyone here had six million dollars to sink into general AI research, we would have been done months ago. But, that's not the case. We have to keep moving against the current, because nobody else will do it for us.
>>2099 So are you saying, with technology that currently exists, an AI, one singular AI, could be made that could learn, not be programmed to, learn how to play chess, draw whatever you ask of it, write music, actually understand two different languages and be able to translate between them? >an intelligence of a machine that has the capacity to understand or learn any intellectual task that a human being can.
>>2100 Yes, all of those things can be done. In fact, they already have been done. You think people who make advanced chess AIs program in every single possible strategy manually? Maybe sometimes, but not all the time. Chess AI is the most develped sub-field of them all. There's an entire libraries worth of knowledge on how to make all kinds of chess AIs. Drawing has already been done, too. I forget the names, but there are multiple AIs that make beautiful paintings, and 90% of the code there is from training data, not programmed. I don't know about music, but I don't see how it could be any harder, and it likely has been done. In fact, it's probably easier. Natural language processing is the field of making computers understand human languages. Apply that to the two languages of your choosing. This one is the most difficult, but again, is very plausable.
>>2101 I already knew that seperate, individual AIs could do these tasks. Take a neural network, feed it tons of data, or have it play against itself over and over again. Okay. That's not really what i'm talking about though. I'm pretty sure there isn't an AI you could ask to draw you a boat, or something like that. If you fed one tons of boat pictures, it could generate a novel boat looking thing, but I don't think generation is the same thing as creation. There's no AI that can at your plain english instruction, look at one image of a boat, open up an art program, and have a go at it, even if it's drawn something else before. If you tell it do something it doesn't already know how to, it can't do research and go out its way to learn how to do it either. I don't if that kind of thing is possible with current technology.
>>2102 That's no different from a human; we can't draw boats without first knowing what they look like either. We also can't do it well without a shitload of practice and reference material. As for generation versus creation, I'd argue that they are the same thing. In fact, they literally mean the same thing, unless you take the biblical meaning of create, in which case it still doesn't matter because humans can't make something out of literally nothing either. If you gave an AI one picture to work with, it could conceivably come up with a similar, though rough output, much like a child. Research is programmable too. If I compile a huge library of data that takes up terabytes of space, I probably won't have the AI process everything at once. I would have her look up what she needs to when she needs to do so, and it's easy. If I ask her what a boat is, and she doesn't know, then I can have it so that she'll take the unknown keyword "boat" and search the data library for any information with "boat" in the name, then process it.
>>2103 I see your point. These aren't things a human could do easily either. There's still the issue of understanding novel, plain english instructions. How about an instruction like "learn how to play chess". Given nothing but this information to work on, not even knowing what chess is, how would an AI figure out what to do? First it would need to look up what chess is, then is would need to extrapolate from the instruction or the information it finds than chess is a game, and it would need to need know that games are "played" and what "playing" means. That's just the first step. I cannot imagine how programming extrapolation would work. Even "make me a cup of coffee". You'd have to program is to be able to figure out what "make" means in this context. If you asked it next "make me a cup of tea", you'd have to program it to use previously acquired "insights" in new situations. Reducing the amount of necessary data needed for the robot to figure out every little thing, by upping it's reasoning ability, seems like a good goal. >it could conceivably come up with a similar, though rough output, much like a child. conceivably is a rough word.
>>2104 I see your point of understanding instructions, but at the same time, I imagine that it's not a big leap once the AI in question is proficient enough in natural language processing. As for natural language processing itself, well, it's a matter of training. I'm not a good enough programmer to say how it works, I only know that it can be done. Tay has shown that it is possible, and I know that she was an AI because no Microsoft employee on the job is going to praise Hitler. Actually, that comes back to the point I was making earlier about corporations not doing this kind of thing: they can, but they're constantly worried about the AI not being politically correct. Another example is that AI insurance companies were using to automatically determine rates for customers, which ended up drawing the conclusion that blacks are more risky customers. I know there were some more examples, but I can't remember them right now. I believe that the key element here is teaching the AI how to recognize correlation. If you can do that, then rest should come naturally. i.e. recognizing that "cooking" something always involves food or meth and heating it up.
>>2105 >Tay has shown that it is possible I don't know for sure how Tay worked, but I think she was just a more advanced Cleverbot. She had the ability to generate new phrases based off of old ones. Tay was still ultimately trained rigorously, produced results quite close to her input data(it was a concerted effort), and went through that training in a set manner. Tay could only do and "learn" exactly how she was programmed to, which was relatively limited. She did not "understand" language. She understood symbols, "words" and how to string them together based on input data and maybe pre-defined grammatical rules. Even with a body, she could not make a cup of coffee. Technically, computers already "understand" language. That's how i'm trying this, that's why programming languages are usable. There's a difference however been "understanding" a word: associating it with a pre-set process, and understanding it in the human sense: a malleable and illusory concept. If you try learning a second langauge, you start to see how dependent we are on instincts to derive meaning from these sound waves and represent them with written characters. It really is gibberish, but it's got statistical patterns kind of associated with objects, images and emotions. Computers don't have any of that built in. Those "politically incorrect" AIs? They were just pointing out statistical patterns, just as they were made to. Maybe an entirely new way of processing words could be programmed? "Teaching" the AI how to "recognize" "correlation" is a three-layered, very complex problem. I wish you luck and I hope you learn more about programming. From the little I know myself, I think your approach actually has technical blockades.
>>2106 This discussion has given me a lot to think about, and I will continue to take a hard look at just how the hell I'm going to approach this. How are you planning on going about it? I don't see a lot of discussion on abstraction here, so I want to see how others look at the issue. We need to plan out how we're going to program our AIs before we get to the actual programming. Maybe I'll start a thread once I have a complete layout to bring to the table.
>>2107 >How are you planning on going about it? Keep going through college and hope I learn enough to seriously contribute. Otherwise, using words as abstractions/shortcuts for statistical data(maybe as a new data type?), to put it really generally, might be a good start. I'm still leaning towards a biology-inspired solution, but intermediate attempts are informative.
>>2097 >If people are interested, I could post it once it's made. >>2107 >Maybe I'll start a thread once I have a complete layout to bring to the table. yes, please do anon. and i'm sure there are people already here who can write basic software at least. from there it might conceivably be passed on to someone with real talent in programming who also cares about the idea of robowaifus? >>2108 good luck anon, work hard in school.
>>2059 Once something is released onto the internet, we all know it's impossible to remove it. Therefore when somebody develops code for an AI, you can expect millions of people are going to copy it and build their own robowaifus. It's really just a matter of time, and of course if humanity can survive long enough to get there. >anarcho-transhumanism and anarcho-primitivism Pretty sure we can have both. If your perfect life is living innawoods, there's nothing stopping you from packing your shit and living in the Canadian wilderness. I personally enjoy technology, so I'm happy to deal with the unfortunate but necessary side-effect of being near other people. >>2062 Stop blaming other people for your personal problems. The only conspiracy in this world is that people are easily fooled into electing charismatic but shitty people to run their country.
>>2110 > millions of people are going to copy it and build their own robowaifus I like the way you think anon!
Open file (897.13 KB 1920x1080 vsLbMq.jpg)
>>2110 Oops my bad, I didn't read the whole topic lol. >>2050 >I'd like mine to be unemotional, lacking in empathy, stoic and disinterested in personal gain or other people Same, but I'd like her to be attached to me in some way. There are a lot of questions around how to do this. I mean you could simply make her feel good when she's in your presence... You could argue that's how the emotion of love works anyway. However I think there are still a lot of (((ethical))) issues around this. After all, some Christians argue that God lets evil exist in this world because it wouldn't be true love unless we completely voluntarily loved God. A philosophical problem, but relevant when we become the gods. But I have an autistic solution to this. Create an entire virtual world populated by different AIs. Let humans "jack in" to this virtual world, populated entirely by hot anime chicks or whatever. Let them fuck around in it, running whatever scenarios they want to run. Eventually they find best girl, somebody they love, and who loves them back genuinely. This girl's AI can then be downloaded into a real body. The girl would be free to move between her virtual world and the real world.
>>2113 Your last paragraph would be a good scifi story as-is. You should do a green-text in the robowaifu fiction bread anon.
>>2113 >Same, but I'd like her to be attached to me in some way. That's the idea. Honestly, i'd like her to be a bit of copy of myself, like an extension of myself instead of an entirely separate person. Relationships fizzle out unless you keep the other person entertained, since everybody has their own interests at heart subconsciously, except maybe maternal relationships. That's not a problem if they're a part of yourself, your happiness is the same as theirs. They completely understand you too. >Eventually they find best girl, somebody they love, and who loves them back genuinely. Because I think the only person I love is myself, I don't know if that would work for me. It's an ugly sentiment, but it's true.
>>2115 That's good insight anon, I like it. >It's an ugly sentiment, but it's true. I think we all deal with that. Jesus' words "Love your neighbor as you love yourself" both cuts right to the issue, and also gives guidance how to get out of the syndrome--all in the same simple sentence.
Open file (67.60 KB 960x500 jackie.jpg)
>>2115 >I think the only person I love is myself
I've thought about it a bit more. I don't know how feasible any of this is, but here is my basic outline for primitive, more human-like language processing. The AI would store information on "verbs" and their previous usages as statistical data. Adverbs would be simple functions. Every single verb would be learned individually. Let's say you give an instruction like "write "apple" in programx". The verb "write" would be learned by feeding the computer data of it being performed(you typing different things) and told that in these example there is "write" being done, the computer would try to figure out what action is most "similar" in every example(algorithms are hard). Adverbs like "in" could be given a rigorous definition provided directly by the programmer. Nouns like apple would be structures containing a string and any other data like classifications, or images. The AI would process this instruction word by word until the end of a sentence. There would be an end state of "understanding" the instruction by the end of the sentence by a method of gradually "filling in the blanks". If understanding is not reached, nothing is done. Now, for previous usage of verbs, lets say you give an instruction of nothing but "write". The AI would look at its own previous usages of that verb, and do exactly the same thing as in its most common usage. It would do this if "understanding" is not completed by the end of the sentence. Lets say you asked it to write other words previously that aren't "apple", but most of the time it was, with nothing but "write", it would write apple. If no word was ever written more than once by the computer, it would randomly pick a noun which shares the most common classifications of the word that were used, like if most nouns were classified as "fruit", it would randomly pick some "fruit". Lets say you used a second program to perform the same verb, it would choose the program most often used with that verb. Information like what program is used in training data and what the actual contents of the writing are, could be hidden during the training process, to reduce false assumptions on the AI's part. Then comes adjectives, these are either rigorously defined by the programmer (string between 5 to 10 characters), or the same as classifications. These could further utilized by using the nouns and all the information already associated with them to learn new nouns. Say you give the AI a picture of a pear and tell it, it's called a pear. The AI would be see where this novel object fits statistically within the nouns it knows to create and define a new noun. A "pear" looks like an the images under "apple", so therefore it will be classified as a fruit too. Everything except the given novel qualities of the "pear" would be assumed to be in common with the "apple". This has tons of limitations and I don't know how well it could be applied to anything except writing words.
>>2122 Also, if a verb could be used in more than one way, the noun would be looked at the derive specific actions. If the verb means something with nouns of one classification, and something else with nouns of another, the computer would pick the proper usage. This is another way of interpreting previous usage data. A verb could be trained more than once for every usage case and the AI would try to figure out what's "different", forming its own second definition that can be called by the same word. The AI would check if the verb has more than one usage, and if does check the noun to see which usage it fits.
>>2122 >>2123 Very interesting anon. Do you think you could learn a description language used in Behavior-Driven Development? It's simply a reduced-vocabulary form of English descriptions that allow for logical states to be both described technically and easily checked for consistency in running systems. https://en.wikipedia.org/wiki/Behavior-driven_development https://www.agilealliance.org/glossary/bdd (there's another closely-related BDD as well) https://www.ibm.com/developerworks/library/ws-bdd/index.html https://en.wikipedia.org/wiki/Business-driven_development The reason I ask is that if you can re-cast your ideas into these simple language clauses, then I can take those and write conforming tests that can serve as a mechanism to drive the development of functional code. https://github.com/catchorg/Catch2/blob/master/docs/tutorial.md#bdd-style https://www.electronvector.com/blog/using-catch-to-write-bdd-style-unit-tests-for-c
>>2124 Sure. I've got next to nothing planned after the semester. How much of a programming foundation is needed?
>>2125 No real programming is required (after all this is a way for business managers to communicate clearly with developers--historically a real challenge heh), it certainly would help a lot if you can code at least well enough yourself to begin to understand other people's code (I use C++). You just need to learn to write stuff that looks just like this in form: SCENARIO("Power off tests", "[power_button]") { GIVEN("the power is off") { power_button_initialize(POWER_OFF); WHEN("nothing happens") { THEN("the power is still off") { REQUIRE(power_button_getPowerState() == POWER_OFF); } } WHEN("the power button is momentarily pressed") { power_button_pressMomentary(); THEN("the power turns on") { REQUIRE(power_button_getPowerState() == POWER_ON); } } } } SCENARIO("Power on tests", "[power_button]") { GIVEN("the power is on") { power_button_initialize(POWER_OFF); power_button_pressMomentary(); // Turn the power on. WHEN("the power button is momentarily pressed") { power_button_pressMomentary(); THEN("the power remains on") { REQUIRE(power_button_getPowerState() == POWER_ON); } } WHEN("the power button is held down") { power_button_pressHold(); THEN("the power turns off") { REQUIRE(power_button_getPowerState() == POWER_OFF); } } } } also: https://dannorth.net/introducing-bdd/ Here's a simple template that describes some small, achievable goal: BDD-style test cases In addition to Catch's take on the classic style of test cases, Catch supports an alternative syntax that allow tests to be written as "executable specifications" (one of the early goals of Behaviour Driven Development). This set of macros map on to TEST_CASEs and SECTIONs, with a little internal support to make them smoother to work with. SCENARIO( scenario name [, tags ] ) This macro maps onto TEST_CASE and works in the same way, except that the test case name will be prefixed by "Scenario: " GIVEN( something ) WHEN( something ) THEN( something ) These macros map onto SECTIONs except that the section names are the _something_s prefixed by "given: ", "when: " or "then: " respectively. AND_GIVEN( something ) AND_WHEN( something ) AND_THEN( something ) Similar to GIVEN, WHEN and THEN except that the prefixes start with "and ". These are used to chain GIVENs, WHENs and THENs together. AND_GIVEN was introduced in Catch 2.4.0. When any of these macros are used the console reporter recognises them and formats the test case header such that the Givens, Whens and Thens are aligned to aid readability. Other than the additional prefixes and the formatting in the console reporter these macros behave exactly as TEST_CASEs and SECTIONs. As such there is nothing enforcing the correct sequencing of these macros - that's up to the programmer! https://github.com/catchorg/Catch2/blob/master/docs/test-cases-and-sections.md#bdd-style-test-cases It's really just a specified way to make statements that assert characteristics of your system. But because they are a well-defined subset, they are fairly easy to translate into actual working code. And when the notion is flatly preposterous under the current scenario GIVEN( I want a robowaifu that is conscious and sentient, costing no more than one dollar ) WHEN( I make a really big fuss about it on the Internet ) THEN( surely Anon will deliver ) it becomes pretty patently obvious even to non-technical users why the scenario fails.
I guess I should add that you must fashion test assertions that are painfully low-level and seemingly senseless to a casual onlooker. For example in this scenario SCENARIO("Power off tests", "[power_button]") { GIVEN("the power is off") { power_button_initialize(POWER_OFF); WHEN("nothing happens") { THEN("the power is still off") { REQUIRE(power_button_getPowerState() == POWER_OFF); } } WHEN("the power button is momentarily pressed") { power_button_pressMomentary(); THEN("the power turns on") { REQUIRE(power_button_getPowerState() == POWER_ON); } } } } Notice that you are actually asserting that when nothing happens, then the power is still off. Seemingly stupid points to make like this are actually important because a) we're dealing with mindless collections of computers, circuits, motors, hardware &tc., and b) shit happens. The more complex a system becomes, the more important it becomes to ensure even the most inane minutiae are looked after automatically by the system under test.
>>2126 >>2127 Alright. This seems accessible enough. Is it useful for "black box" type stuff, such anything an AI is trained to do? Even the person actually feeding the data doesn't know exactly what the AI does after being trained.
>>2128 >Is it useful for "black box" type stuff, such anything an AI is trained to do? I don't see any real limits to this approach, beyond the obvious basic limit of human language itself, English in this case. An human imagination itself is limited too, fundamentally. That last point is exactly why TDD & BDD were invented--to help programmers who are fundamentally limited yuge big tasks (mek robowaifu) into a zillion actually-achievable tasks (mek switch A activate motor B) that can be assembled together into a IRL functioning system, with all the tests to ensure you don't break anything already built all along the way. >tl;dr Heh, so yeah I think so Anon.
>>2122 >the computer would try to figure out what action is most "similar" in every example Yeah if only we had even the slightest idea how to do this. Whenever I think about implementing language processing I always end up realizing you will need to give your AI an understanding of the universe to do so. So to make language-based AI, first you must develop general AI. Shit maybe they're the same thing, since humans became intelligent around the time they invented language.
>>2131 >Whenever I think about implementing language processing I always end up realizing you will need to give your AI an understanding of the universe to do so. Could you treat the computer like the universe? Maybe an AI wouldn't need to understand these words in a real world context, just in the context of its computer universe. Writing doesn't actually mean keyboard inputs, but if an AI could learn and "understand" the word write as such in a more human-like way, and be able to perform it with given english-like instructions, wouldn't that be a step forward? Is that impossible?
>>2131 >you will need to give your AI an understanding of the universe to do so "If you want to make an apple pie, you must first create the universe." He was wrong about a whole lot of things, but at least he got that one right. https://www.invidio.us/watch?v=7s664NsLeFM
>>2132 I'm sure that a virtual world would be much easier for an AI to deal with than the real world. All attributes of an object are already known. In the real world the attributes must be discovered through learning. Let's say I ask a robot to get me a cup of water. The robot must known that the cup must be held the correct way up, otherwise the water will spill out. The robot must know that the cup must be directly under the water source, because gravity will make the water flow downward. The robot must know to move carefully when carrying the cup, otherwise it will spill the contents. All of these require an understanding of the basic laws of the universe. Anything beyond the simplest conversation also requires an understanding of the universe. For example I can ask a robot to read me a book. That's easy enough to hardcode. But if I ask the robot who their favorite character is, how could they even begin to respond without understanding at the same level as a human? They would need to understand the plot, the characters, even the themes and symbology in the book. And to get to that point you might as well invent a general AI. I'm confident it can be done. The human brain is really just another kind of computer anyway. But it's going to be a big project.
Open file (67.62 KB 800x573 underground-bunkers.jpeg)
>>2134 Interesting observations. I like the way you break down requirements. Obviously imparting the entirety of the understanding of the universe is impossible (even we don't have it), but we certainly should be able to achieve at least a rough but satisfactory approximation of anime catgrill meidos and then Anon can dress them however he pleases. :^) >The human brain is really just another kind of computer anyway I don't believe we have more than just the most vague comprehension yet just how the brain works, apart from loose analogies such as this, or how it engages with the immaterial nature of our human spirits. Likely we never will in this life.
Open file (89.14 KB 842x474 7034860_orig.jpg)
>>2135 Well there are some people who think the mind contains a soul, or perhaps even makes use of quantum effects. These are fringe theories however. As for our current understanding, we are only just beginning to understand some of the fundamental circuits in the brain. It's complicated because there are several types of dendrites in different parts of the brain, but ultimately they form circuits just like a computer.
>>2136 > but ultimately they form circuits just like a computer. > just like I'm skeptical that the 'circuits' formed by the human brain are in any reasonable way directly comparable to the circuitry man devises, computer ones or otherwise. Again, we simply don't have the understanding necessary to quantify the brain. Evidence thus far seems to support my view afaict.
>>2137 Even the most complex computer can be broken down into a collection of very simple logic gates. I see no reason why the same wouldn't apply to the brain. There are two major differences however. The first is that the brain is highly parallel. A single one of our 86 billion neurons may be connected to up to 10,000 other neurons, and all of these neurons can be firing independently. However the inputs would ultimately originate from our senses, of which there are only a handful. The other key difference is the brain can create new connections over time. I'm not sure how either of these could be currently implemented, but I think that if we found a way to create a CPU which used a neural architecture, we wouldn't be far off from being able to make artificial brains. Until then we might be able to fake it with our current architecture.
>>2138 >highly parallel heh, that's putting it rather mildly. quite apart from the basic fact that the neural activity isn't based on gates, but rather is strictly an analog system quantized only at the electrochemical level--and therefore effectively only at the Planck limits--MRI and PET scan data indicates that the entirety of the neuronal conglomeration in all it's varieties effectively participates in every single process of synapse triggering the entire brain over. to wit: everything is connected to everything, even if it's not one the 10'000 closet friendly neighbors. :^) parallel indeed. and that only touches on the purely physical aspects. a human being is both a spirit and has a soul. it's very likely that these aspects of our nature are far, far more complex than just our purely physical beings., whether the brain or other parts of our bodies.
>>2134 I think you're right right about this. Some innate understanding of the universe's laws might be necessary to really answer questions like who their favorite character is. I'd just like to see an AI that can use a more learned meaning of words, and their grammatical type, to interpret english instructions and act upon them in a way based on its own "experiences", so it would still do something even if the instruction has missing elements.
Open file (479.31 KB 1184x789 33995_00_2x.jpg)
>>2139 >everything is connected to everything The theory that neurons communicate not just through nerves but also through an EM field is not very widely-held >a human being is both a spirit and has a soul Why would you be on this board if you believed this? You could very well be right, but the belief has no basis in science.
>>2140 Yuki is an interesting example of this actually, because she reads a book and doesn't really describe it beyond being "unique". I wonder if AI would come up with its own ideas of fiction which are very different from our own. At some point I believe she writes a poem which is also quite unusual.
>>2134 >But if I ask the robot who their favorite character is, how could they even begin to respond without understanding at the same level as a human? I would like to make a counterpoint here. The vast majority of humans, male or female, don't put that much thought into it. Most of the examples you gave are the kind of thing that only critics, philosophers, and other people who enjoy thinking for the sake of thinking will pay attention to. Though an AI that could analyze on a deeper level is ideal, I'd say it's well above the minimum requirements, and therefore unnecessary to achieve the level of "human-ness" required of general AI. I concede that my time spent in the customer service industry may have jaded my outlook, but that's just the way I see it. >>2139 >but rather is strictly an analog system It's funny you should mention that, actually. Back on 8chan /robowaifu/ there was a discussion on the potential benefits of analog computing. >a human being is both a spirit and has a soul There's no evidence to support that, and even if there was, why should we care? Granted, I have the combined spirituality of a factory-manufactured piece of cardboard, but for all the arguments I've seen on both here and the old board, I've yet to see any potential benefit to accounting dodgy spiritual theories and into our design process.
Open file (67.02 KB 960x720 hideki_chii.jpeg)
>>2141 EM certainly does play a role in the transducing of impulses, electromagnetic theory 101. but that wasn't what I was describing anon. The observation data indicates the entire collection of neuronal tissue is activated when even the smallest part of it is. ergo, there apparently is a contribution from the whole to the part, and vice-versa. >Why would you be on this board if you believed this? Kek, because I'm the boards founder. Chobitsu at your service sir. >but the belief has no basis in science. I go where the facts lead me. The facts indicate that there is a Higher Intelligence behind all creation and in particular the biological systems. I personally belief that Creator to be the very God of the Bible, again because all the evidence points in His direction. >>2143 >Back on 8chan /robowaifu/ there was a discussion on the potential benefits of analog computing. yes, i participated in it. I certainly think it would be a big benefit to us if we could somehow manage it. in the meantime digital computation seems our primary path forward. >There's no evidence to support that I believe the vast majority of humanity throughout all history would disagree with you. I know I do. (respectfully). And yes, dodgy theories I'm with you on. Ones that are backed by literally thousands of years of cultural experiences and deep theological studies--again for thousands of years time--I won't toss those aside quite so lightly. And on a personal basis, not even the remotest chance of me turning my back on the faith of the Gospel.
>>2144 >I personally belief that Creator to be the very God of the Bible To tell you the truth, this is suprising. If I may ask, were you raised in a religious household? >not even the remotest chance of me turning my back on the faith of the Gospel Not even if any facts lead you there?
>>2145 >If I may ask, were you raised in a religious household? You may indeed. No, I wasn't. In fact it created a bit of controversy in my extended family when I got saved. >Not even if any facts lead you there? Sure. If, for example, you could prove incontrovertibly, with overwhelming evidence, that the creation event commonly known as the Big Bang, didn't actually happen, then yes. That would fly directly into the face of the Biblical claims about the beginning which is absolutely foundational to the Christian faith. Even if you could do so, I wouldn't deny even in the slightest way the spiritual experiences I've had with Jesus. But it would rock me to the very core. But still, my faith and my salvation is based on my relationship with Christ, not my head (even though I'm actually an amateur scientist, in Astronomy). Heh, but personally, I don't think there's even the remotest possibility that you or anyone else will ever discredit the hot creation origin of the universe. There has been far too long of a concerted effort to do so in scientific community during the past century to do so, all to no avail.
>>2144 >all the evidence points in His direction Not to sound like I'm trashing on your beliefs, but wouldn't the existence of such evidence be widely known if it was true? >I believe the vast majority of humanity throughout all history would disagree with you The vast majority of those very same people would see us all burned at the stake or stoned for trying to encroach on God's domain. We shouldn't take queues from groups of people who are fundamentally opposed to us. >>2146 Now I'm confused. Isn't the Big Bazinga theory in direct opposition to most theological claims of God creating the universe? Or are you a member of one of the more distant off-shoots of Christianity, like Mormonism?
>>2146 The bible gets pretty specific about some stuff... There's the whole Paul thing too among other things. I tried looking into it a bit. Are you the type who takes a metaphorical approach rather than a literalist one? Would you say you comfortably fit within "mainstream" Christian belief?
>>2147 >but wouldn't the existence of such evidence be widely known if it was true? isn't that in fact the case anon? the vast majority of humanity hasn't had to live under (((benefits))) of modern education anon. :^) Go spend one solid week alone in nature, away from civilization, get quiet from distractions and just listen then I believe you too can come to agree with all of us. >encroach on God's domain wouldn't surprise me if some would. meh, i'm not doing this to please them haha. nor do i care tbh >Isn't the Big Bazinga theory in direct opposition to most theological claims of God creating the universe? What? Haha no of course not. Exactly the opposite. That's why the humanists in the science communities fought so hard against it after Hubble's observations because of the deep theological implications of the observed facts. They've now all stopped trying to disprove it since the weight of evidence is so overwhelming today, observationally speaking. >>2148 God is supernatural by His very being, His very nature. Obviously. So, no I have not the slightest problem with the existence of miracles past or present. >"mainstream" Christian belief I'm a Protestant. I don't think the majority of my fellow believers embrace the scientific pursuit to quite the degree I do, in fact some seem quite opposed to (due to misunderstanding imo). I also have a tangential relationship with Orthodox Christianity, which goes back all the way to the very time of the foundations of Christianity itself. >tl;dr No I'm not the 'typical' American Christian, Anon. I just simply love Christ with all my heart and am deeply devoted to him for the gift of life, both now and forever.
>>2149 >Go spend one solid week alone in nature, away from civilization, get quiet from distractions and just listen then I believe you too can come to agree with all of us Thank you, but I've spent enough of my life in total isolation and no, I'm not referring to this virus panic. The only thing I found there was insanity. In fact, I challenge you to spend several months with virtually no contact with any other living being and no outside distractions. If you come out the other end sane, then I will concede to you.
>>2149 >get quiet from distractions and just listen then I believe you too can come to agree with all of us What did Darwin do wrong? He actually grew up religious and being in nature constantly changed his mind until he was at his death bed. Also, why do you think Japan is so secular and what effect do you think that has on their mentality and culture?
>>2150 I said in nature. Go get alone with just the animals and birds. The mountains, the sky and the water. Gaze into the vast and awesome cosmos alone on a high plateu at night. And just listen. Humanism really has it's main stronghold in cities. Getting away from cities is the key here Anon. >>2151 I don't know. It's not really my affair to judge the man. I actually respect his efforts at understanding the truth. The primary abuse against humanity in his name has come chiefly after his death. I personally believe Charles Darwin would never have become, well, a Darwinist haha if he had had access to the information we have today. For example about the 100's of species of molecular motors within living cells which are either direct analogues or near corollaries to modern man-made motors according to engineering practices. >Japan I have a deep love for the Japanese people, and have a high honor for their culture before the Commodore Perry debacle and resulting degeneration of them. In fact I mean to go live in Tokyo for at least two years after I graduate.
>>2152 >I have a deep love for the Japanese people Do you think their nation would be improved if it became predominantly Christian?
>>2153 Of course.
>>2152 I live in a rural town. In fact, I can count on one hand the number of times I've even been in a city. I've spent plenty of time, both during day and night, observing nature. I love the natural world and all the beauty it has to offer. I gaze upon the cosmos and ponder the not only brilliance of the universe beyond us, but the mysteries held so far out of reach. Though I sometimes wonder about the possibility of a higher power, not once, from my earliest memory to where I am now, have I ever seriously considered it. Telling me to see the things I've already seen and hear the things I've already heard will not change that.
>>2144 So why do you want a robotwaifu? wouldn't going to church and finding a nice christian girl be easier? You could start a family with her and all that moral lifestyle stuff. Wouldn't robowaifus also encourage an immoral lifestyle since fornication with one is practically masturbation?
>>2155 OK, fair enough. At the very least you have the benefits of not living in a big city during your life here on Earth, Anon. For me, the amazement at the creation around never.ceases. Even living in LA, arguably one of the most sprawling megalopolis' in existence, I'm still enthralled by just the land and the ocean around me itself. This planet is amazing and utterly miraculous. This entire universe is. https://www.spacetelescope.org/science/deep_fields/
>>2156 > So why do you want a robotwaifu? Why wouldn't I? The ideal of a robowaifu is awesome tbh. > wouldn't going to church and finding a nice christian girl be easier? who knows? certainly there are few Christian 'girls' untouched by feminism here in Burgerland Anon. The Amish are out since I'm not part of their communities. Maybe in Eastern Europe somewhere. > Wouldn't robowaifus also encourage an immoral lifestyle since fornication with one is practically masturbation? Probably. My primary motivation, as I've stated several times here on this board over our 3 years, is to help the millions up millions of disenfranchised men who've been cheated out of ever having a normal, healthy relationship with one of these unicorns you speak of anon. (((Feminism))) is a helluva drug, and a blue-pill the West overall has swallowed hook, line, & sinker.
Open file (193.02 KB 1600x900 angelegg.jpg)
>>2158 Alright. I'm sorry if my questions had an interrogative tone. I've just never seen somebody such as yourself before and I honestly still can't quite rationalize your way of thinking. Your primary motivation is something I admire.
>>2159 No worries mate. I am just me. If there's anything different about me, it's simply that a) I'm just a child at heart, for good or ill, and b) I'm OK with that. Cheers.
Open file (53.65 KB 1280x720 yuki.jpg)
>>2143 >it's well above the minimum requirements You make a good point and I was actually considering this. After all, most AIs in fiction are portrayed as emotionless. Emotions might be superfluous in day-to-day life. >>2157 >This planet is amazing and utterly miraculous. This entire universe is. No doubt. But it's quite a leap from there to angels and the devil and heaven and hell and all the other stuff that Christians came up with. What makes you think Christianity is the true religion, and not an older one like Judaism, Hinduism or Shinto?
>>2161 I can't speak for him, but from some of my conversations with Christians, I think it has to do with "spiritual awareness". Basically, when you're spiritually aware, only then can you perceive and understand all the spiritual evidence pointing towards that specific collection of divinely inspired texts, the bible. You become spiritually aware, by letting Jesus transform you or something, and when you go to heaven, you get transformed even more, so you have no desire to sin, but you presumably still have free-will, because that's necessary to love God. Or something.
I couldn't help myself. I was just too curious. I've been banned for promoting perversion. My sides are in orbit. https://christianforums.net/Fellowship/index.php?threads/what-do-christians-think-about-robot-companionship.82158/ I hope for mercy.
Open file (2.17 MB 512x384 tenor.gif)
>>2163 Hahaha nice. >According to the bible any sexual act outside of a marriage between a man and a woman is a sin Yeh this pretty much. Although it robowaifus existed they would probably change their interpretation of the Bible to make it okay.
Open file (93.95 KB 640x904 nytjtkdg8jq41.jpg)
>>2164 >they would probably change their interpretation of the Bible to make it okay. When robaotwaifus get realistic enough and more importantly, more socially acceptable, I can see the pragmatic among them saying something along the lines of "God must have given them a soul, so they therefore now count as woman". >Single, available, genuine Christian men are almost unicorns in euphemism in Christian churches. AKA they are a myth only heard about and not seen or found. Every time one is around on a regular basis the women swarm him and won't leave him alone until he gets married. I wonder how true this really is...
>>2165 >I wonder how true this really is... It's very true once you hit your mid 20s and women start to realize they're about the turn the corner, and they need to lock down a man, any man, to give them a family.
>>2165 IIRC, the Mormon faith more or less explicitly states that machines can in fact have souls (I think it's in the Book of Mormon somewhere, though I've yet to get around to reading it). They'll likely advocate for the continued existence of any companion AIs if nothing else. >>2166 That's a bit of a point, actually. If the admin's picture is really a picture of himself, then that means he's at least a boomer. From his objectively wrong viewpoint, I can see why he might think that the situation is how he stated it.
>>2162 >but you presumably still have free-will This heavily depends on which doctrine you follow. I know that, in the Old Testament at least, free will is considered the "original sin" because having free will leads to all other sins. In the eyes of those who follow the teachings of the Old Testament, the only way to absolve yourself of sin to to relinquish your free will and surrender yourself to God. This, combined with all the other wacky shit in the Old Testament, is why people thought a New Testament was necessary. It's harder to gain followers when you look like a freaky cult that will smite you with rocks and fire for so much as thinking the wrong way. I really need to get around to re-reading the Bible. My memory of it is getting rusty, and it's a damn good read even if you don't believe a single word of it (which I don't). A warning to anyone who wants the OG King James experience though, you will want to have at least a cursory understanding of Middle-English.
>>2161 While like any man, I don't appreciate having words put in my mouth, in this case I'll be happy to actually confirm your view of me. I do indeed think exactly that. What makes me think this way? Easy. -Abraham, Krishna, nor the Buddha saved me, Jesus did. -Abraham, Krishna, nor the Buddha are the Son of God and the Son of Man, Jesus is. I'm happy to grant that those are my own personal experiences with Him. Many are called, few are chosen. >>2162 Yeah, that's pretty close. The notion is refereed to as 'spiritual blindness' both inside and outside the text of the Bible. Like the men watching shadows on the wall inside a cave and refusing to come out into the sunlight, men & women can refuse to absorb the plain evidence around them and choose to cling to the falsehoods instead. This is because a spiritual discernment is necessary to even have the basic impetus to be freed from the cave at all. >"The natural person does not accept the things of the Spirit of God, for they are folly to him, and he is not able to understand them because they are spiritually discerned." https://biblehub.com/1_corinthians/2-14.htm >>2163 That's funny, I'm surprised there was even a topic about this yet. >>2165 >I wonder how true this really is... Fairly true. I have lots of women after me whenever I go to church. The problem is, basically all of them are corrupted by feminism, near as I can tell. I wouldn't want them raising my kids, for example. Again, I'd probably be happy if I found a trad, wholesome wife in Eastern Yurope, but I live in Burgerland instead. Maybe I'll travel there someday, but even if I did score a 10/10, we'd have to live there not here. Just being a part of the West these days will soon degenerate her. >>2168 >free will is considered the "original sin" Not to be rude, but that's a misunderstanding at the least. Sin is willful disobedience to God. It all started with Lucifer who became Satan. He rebelled again God and then acted as the ring leader to lead us into it. The rebellion was already inside us, Satan didn't put it there tbh. God knew ahead of time we would fall into sin and so arranged to sacrifice His own Son to provide a way of escape for us from it's well-deserved destruction.
>>2163 >>2169 >a topic about this yet. >yet <'hideki' literally joined & created thread yesterday pfft, that's what I get for not taking the time to check the link before posting. well played. still, it will be interesting to watch if anything of substance comes from this ebin le ruse anon.
>ITT: Anons derail the board into debate about Christianity Kek, I think we made the BO mad. >>2169 >that's a misunderstanding at the least I must be misremembering then, it's been a long time since I've read the Bible. Either that or I misunderstood in the first place because I didn't understand the nuances of Middle-English (well, technically late Middle-English in this case) nearly as well as I do now. Or maybe who's right and who's wrong depends on which sect you follow, because there are dozens of different sects of Christianity due to every fifth person having a different interpretation of the Bible. >men & women can refuse to absorb the plain evidence around them and choose to cling to the falsehoods instead What would you call those who are spiritual, but find the light of a different god?
>>2171 >Kek, I think we made the BO mad. heh, i aint even mad bro. BTW, i'm not the BO any longer, Euphoria is, i'm just a hotpocket now. however in general it's quite important for us as an engineering board and stay tightly on topic. otherwise it can become near-impossible to relocate posts in the future if things are spread out everywhere. however in this case a) the OP is a bit vague in the first place, b) the notion of consciousness needs some place to be discussed here, and I'd rather it take place under the auspices of the Christian paradigm (for or against), and c) i'm literally the no1 contributor to the relatively busy thread thus far--so i have no one to blame but myself. :^) but in general, yea please keep things on topic. will move shitposting into The Lounge w/o warning. >the origin and nature of sin it's quite apparent from the text itself the Biblical claims on this. it was first discovered in Lucifer. Adam and Eve sinned and died spiritually immediately (thus were sinful in their very nature, as were all their offspring then till now). only the propitiatory sacrifice of Jesus Christ provides the covering for sin that purchases redemption for us. i realize this can be convoluted by some, be tbh it's breddy straightforward actually. >different gods ofc as a Christian, my view is there are no other gods. there is only God, angels, and humans, spiritually speaking at least for this life heh :^). the angels are divided into two categories: Gabriel, Michael, and all the heavenly hosts, and Satan and all the demonic hosts. this latter is the source of all the other religions btw. people who don't receive salvation through Christ simply don't receive it at all. there is no other door to eternal life but Jesus. I would also add on a personal note that from the scientific POV, only the Biblical account accurately identifies the initial conditions and frame of reference to understand properly the creation of the universe. there are many many other areas where the Bible is unique but this one is particularly important to me as an amateur astronomer.
>>2171 >What would you call those who are spiritual, but find the light of a different god? Again, I can't speak for him, but I have asked Christians this very same question. The answers I've seen are either they're doing it wrong or they've been deceived by satan. Doesn't matter how much they meditated. They did it wrong. Any question you can come up with has some answer, and if you do dig deep enough, it eventually comes down to humans not actually being able to know everything. Either humans will find out in heaven, or humans just don't need to know that information. That's why faith is necessary ultimately. Why psychopaths are born or why satan was made and earth wasn't like heaven to begin with, doesn't matter. I've talked with people who believed all sorts of things. One guy he convinced he was an immortal superhuman who was abducted by aliens and given super-human abilities. He thought the origin of all religions and myths must have been aliens. I tried telling him otherwise and it was a complete waste of time. Humans' propensity to believe improbable things based on emotion and "supernatural experiences" is one of our greatest curses, but it can't be helped. As long as Chobitsu is fine with all us heathens going to hell, everything is good.
>>2173 >As long as Chobitsu is fine with all us heathens going to hell I assure you, I very very much am not OK with that, but it's not up to me. ofc you make your own choice in the matter as do we all. all i can do is use my meager abilities to try to spread the truth while you still have options. I certainly want to see every human in Heaven, not hell. Hell was created for the devil and his angels, not humans.
>>2174 >all i can do is use my meager abilities to try to spread the truth while you still have options. Alright, but promise you wont slip some religious code into the robowaifus.
Open file (433.45 KB 960x760 deus_vult_bro.png)
>>2175 TOO LATE BRO >tfw Christchan running the latest TempleOS+HolyC will be with you day and night, day and night.
>>2096 >So you don't think there'll ever be a point where you get so genuinely attached you wouldn't want an upgrade? sorry, i missed this. umm, probably not, but who knows? i've certainly seen people get inordinately attached to inanimate objects before (cars for example). >Chobitsu yea, that was some wacky writing by those Clamp women, tbh. and that whole 'vagoo B off-limits!1' thing? well it would have been repaired the very day i discovered Chii's mom was Mrs. Hibiya-san. > I don't know if humans have the capacity to break emergent properties like "understanding" and "appreciation" down into simple parts. Well, whatever. i doubt it. the soul is vastly complex. thanks for the link btw!
>>2177 >>2096 Boom! It just came to me. Cherry 2000 is exactly on topic for your question anon. It's a fun B-grade cult flick to boot.
>>2173 >The answers I've seen are either they're doing it wrong or they've been deceived by satan This seems to be a common theme with virtually every religion. >He thought the origin of all religions and myths must have been aliens He sounds like a Scientologist. >>2177 >i doubt it. the soul is vastly complex. Who's to say that computers can't have souls? I may not explicitly believe in the concept of souls, but any religion that says machines can have them gets a thumbs-up from me.
>>2192 >Who's to say that computers can't have souls? I dunno who says? Certainly they won't be the kind of souls that God alone creates, but there may be (likely will be, IMO) 'souls' in machines. These simulacrums may approach human levels of engaging interactions. >soul-in-the-machine cults heh, i'm sure there are a few. Go talk to Ray Kurzweil for starters anon. :^)
>>2192 btw, super cute Nano. she's a great robowaifu tbh.
Open file (797.38 KB 500x282 nichijou-dog-gif-5.gif)
>>2177 >>2096 >I don't know if humans have the capacity to break emergent properties like "understanding" and "appreciation" down into simple parts. To me these concepts are variants of "correlation" which computers are very capable of. >>2195 I would screw her
>>2273 >ba-dum-tiss* heh, easier than saying "I would so turn that Nano waifu's wind-y handle thingy in her back until she was so full of energy you wouldn't believe!"
>>5069 I have come to realize that building a motorized chassis is certainly possible. Even one that uses legs for locomotion if you have a large amount of money. However, the main problem appears to be free-will. Unless a machine has free-will, it won't ever be able to hold an interesting conversation or form an engaging relationship. I'm not saying this to discourage anyone (it is possible that quantum computing may solve this major issue and grant machines some form of free-will, even if it is only limited at first). But yeah. I think lack of free-will is a big problem. >=== relocated via >>1061
>>5122 >I think lack of free-will is a big problem. Women don't have free will either. >it won't ever be able to hold an interesting conversation or form an engaging relationship No woman was ever capable of doing that either, in the whole existence of humanity. The main thing to remember about biological women is that they have no soul and no inner thoughts. They just copy whatever is told to them by the first group of people that impresses them. Women are no different than a robot. But unlike robots, that can be programmed, women can only be conditioned up to a certain point. After that point, their defective brains enact their retarded survival instincts, and they become the self-destructive mindless walking carcasses that we (now, thanks to instant communication) can see everywhere around the world.
>>5123 Holy shit.
>>5122 Hi, not the guy which posted >>5123. Ben Goertzel (OpenCog) for example doesn't believe that we would need quantum mechanics for having an AGI, I don't see any indication that he's wrong. Your claim seems to be very impractical in general and also very speculative. Between current chatbots and some system with 'free will' there's a lot of space. We'll see it when we'll get there, though I cant imagine one could have 'no intersting conversation' with some advanced chatbot. Maybe you want to talk about philosophy and free will in particular, then your claim might make some sense, but this is a very special use case. Of course I wouldn't want my waifu to have completely free will, that goes without saying. She will have a purpuse written into her system: To serve and love me... >>5123 might be a valid description or not. Of course, I wouldn't want something like that either. We need something new, tailored to our needs.
>>5125 How can you have love without free will? I mean, sure a program can spit out sentences that have been written by a human programmer or phrases chosen via algorithm and probability...but is that friendship or love?
bumping relo'd convo
>>5125 >Ben Goertzel (OpenCog) for example doesn't believe that we would need quantum mechanics for having an AGI, I get the point you're making, and not to be too pedantic about it, but I don't think most anons realize just how intrinsic quantum-mechanical effects are in our day-to-day lives. Food digestion wouldn't work without it for example, nor would the stars burn. Solid state electronics like transistors rely practically entirely on quantum mechanics to function properly. >I don't see any indication that he's wrong. Again, not to be too picky, but A) AGI is an entirely different topic, and B) while it's conceivable given the laws of physics that some close approximation of human intelligence might be devised artificially, since we are entirely lacking in any fundamental understanding of the source of human consciousness (in a purely materialistic, non-dualism way) then any such 'intelligence' would be completely subjective. Not that I don't think we'll produce satisfying simulacrums over the next few decades. Obviously that's very likely. Just don't kid yourself that it's somehow a true 'AGI'
>>5125 >>5126 BTW, there's a Can Robowaifus Experience Love? thread, obviously related. >>14
>>5126 >How can you have love without free will? By building it to do this: >>5125 >She will have a purpuse written into her system: To serve and love me. Which is good enough.
>>5135 Obviously that will be a significant goal for the engineering of any robowaifu. The way I see it, we should figure out what would make pretty much any feminist the most angry in the creation of our robowaifu's behavior and personalities and then do exactly that.
>>5139 >we should figure out what would make pretty much any feminist the most angry in the creation of our robowaifu's behavior and personalities and then do exactly that. Precisely. But not only making them angry, but also constantly bombard people with all the information (only true information, no false speculations and such) about all the advantages of robowaifus over biological women, to the point that even the more omegas among men start moving away from women, rendering any of their rights and complaints useless, since nobody that can actually do something about it cares anymore.
>>5153 >Omegas among men The entire concept of "Alpha", "Beta" and "Omega" males/females is due to a misinterpretation."Alpha" animals are simply parents in the wild. It's only in artificial captivity (zoos) that what we regard as "Alphas" emerge. Maybe that says something about the current state of our society? But then, if people want to behave like caged animals, I say let 'em XD.
>>5176 I know. I use those terms because, unfortunately, they are the most well known among people. But my point is: even the more "woman worshiper" must presented with the better alternative of a robowaifu, so he will abandon the stupid ideas of worshiping women and sacrificing for them. >>5177 Thanks for the read, anon.

Report/Delete/Moderation Forms
Delete
Report

Captcha (required for reports)

no cookies?