/robowaifu/ - DIY Robot Wives

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

LynxChan updated to 2.5.7, let me know whether there are any issues (admin at j dot w).

Reports of my death have been greatly overestimiste.

Still trying to get done with some IRL work, but should be able to update some stuff soon.


Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB


(used to delete files and postings)

Welcome to /robowaifu/, the exotic AI tavern where intrepid adventurers gather to swap loot & old war stories...

/robowaifu/meta-3: Spring Blossom Tree Chobitsu Board owner 02/11/2021 (Thu) 12:06:37 No.8492
/meta & QTDDTOT Note: Latest version of /robowaifu/ JSON archives available is 210711 Jul 2021 https://files.catbox.moe/sqb9ps.7z If you use Waifusearch, just extract this into your 'all_jsons' directory for the program, then quit (q) and restart. -The TITS thread #1 is now unlocked and open for business (>>9709). -previous /meta: >>38 >>3108 -Library thread (good for locating topics/terms): >>7143 >note: there's also a searching tool coded right here for /robowaifu/ that provides crosslinks straight to posts on the board. it's named waifusearch, and the link to the software is provided inside the Library thread's OP. -Latest version of BUMP v0.2e >>8769 >=== -expand /meta OP content -add BUMP xlink -rm file issues msg -add 1wk extension msg -add 'final comments' notice -rm council notices -add council decision notice -add TITS thread notification -add JSON archives link -add thread continuation notice -rm OP notice
Edited last time by Chobitsu on 07/16/2021 (Fri) 13:10:34.
Open file (83.33 KB 988x1198 IMG_20210607_035410.jpg)
The chatbot thread, which also aas the general AI thread, seems to have hit it's bump limit. I wrote this >>11477, but it didn't come up. I hope it doesn't delete the old postings there, it probably needs to be archived.. I guess we then need to have a new thread for something around talking, processing input, "thinking" and responding, like NLP and graph databases?
>>11481 OK, thanks for the headsup Anon. /robowaifu/ autosages threads at 350, and we start another thread. I'll give the OP of the thread the first opportunity to create his next thread, or I'll make another one instead if needed. > I hope it doesn't delete the old postings there No, we aren't at the number of pages (25) where that can become an issue.
>>11483 If you make a new thread, please name it differently, idk e.g. "Conversational AI, NLP and Knowledge Management" or so.
>>11486 It's probably a good opportunity for us to merge some threads. For example there's explicitly and NLP thread that obviously has a lot of overlap.
Open file (119.02 KB 550x222 09-00750-g001-550.png)
>>11486 How about "Conversational AI, Natural Language Understanding and Knowledge Representation"? >>11483 It would be great if we could summarize progress in the field and discussion from the previous thread. I think the most apparent thing is that hybrid approaches and ensembles of networks have been having the most success. Researchers are actually starting to mix transformers, BERT, LSTMs, graph networks and other things together but it's getting to the point they're just hitting a wall and not really doing anything new. The only paper I've been really excited about is Learning to Summarize with Human Feedback (2020) https://openai.com/blog/learning-to-summarize-with-human-feedback/ Some papers and books covering what has been done recently: Neural Approaches to Conversational AI (2018) https://arxiv.org/abs/1809.08267 A survey on knowledge graphs: Representation, acquisition, and applications (2020) https://arxiv.org/abs/2002.00388 Conversational AI: Dialogue Systems, Conversational Agents, and Chatbots (2020) https://sci-hub.se/https://doi.org/10.2200/S01060ED1V01Y202010HLT048 Towards a Human-like Open-Domain Chatbot (2020) https://arxiv.org/abs/2001.09977 Conversational Question Answering: A Survey (2021) https://arxiv.org/abs/2106.00874 Survey on reinforcement learning for language processing (2021) https://arxiv.org/abs/2104.05565 Improving Language Model Behavior by Training on a Curated Dataset (2021) https://openai.com/blog/improving-language-model-behavior/ Some current SotA models of interest: Commonsense reasoning with semantic similarity: https://paperswithcode.com/paper/a-hybrid-neural-network-model-for-commonsense Question answering: https://paperswithcode.com/paper/tanda-transfer-and-adapt-pre-trained Emotion recognition in conversation with graph networks: https://paperswithcode.com/paper/directed-acyclic-graph-network-for Emotion recognition in conversation: https://paperswithcode.com/paper/contextualized-emotion-recognition-in Entity linking: https://paperswithcode.com/paper/autoregressive-entity-retrieval Neural belief tracking: https://paperswithcode.com/paper/neural-belief-tracker-data-driven-dialogue General purpose: https://paperswithcode.com/paper/albert-a-lite-bert-for-self-supervised General purpose: https://paperswithcode.com/paper/exploring-the-limits-of-transfer-learning Translation: https://paperswithcode.com/paper/understanding-back-translation-at-scale Improvement to lightweight Transformers: https://paperswithcode.com/paper/lessons-on-parameter-sharing-across-layers-in
>>11493 >Conversational AI, Natural Language Understanding and Knowledge Representation Fine with me. >Some papers and books covering what has been done recently Lol. A lot to process, even only looking through. I'm trying to make one or several diagrams on what skills are covered with existing programs or researched ideas and what not. Also, based on that I'm thinking about how we might be able to design some good AI out of it, by combining these elements. One big problem I see is responsiveness. Even if some program can figure out some stuff, it may just take one or two seconds per sentence in one step. Needing to do more than one thing will accumulate these delays.
>>11486 I would suggest the word 'Chatbot' be used in the title, but otherwise I'd say that sounds reasonable Anon. >>11493 Sounds like you have a good handle on this Anon. Might I recommend that you take on the task of creating the follow-on thread? I daresay you already have a solid idea where you'd like to see it go. I may use one of the other threads for the more low-key talk on basic waifu apps/chatbots, that is, from the user's perspective. But again, we certainly need material from the implementer's perspective here, since that's our primary focus as a group.
>>11499 >suggest the word 'Chatbot' be used in the title If so, then maybe after Conversational AI in parentheses. Chatbot sounds to much like it's one (simple) program, but it's the user interface and much of the internal processing of information is related to it.
>>11508 Sure no debate. I withdraw my suggestion. As I mentioned in the second part, another thread more suited to the layman should also be addressed. That one can use the word in question.
>>11515 >another thread more suited to the layman should also be addressed Sure, maybe something straight forward like "chatbots: basics and user interfaces"
>>11499 I got my hands full at the moment but I should have some time in 2 weeks. Work being done on contextual AI is also of interest, where chat history is parsed into data structures and user intents and the AI adapts to that data and other environment variables. The research community doesn't touch on this much since it's more of an engineering problem utilizing existing tech but big business is all over it.
>>11520 OK, sounds good. Since it's been a couple days with no response by the original thread's OP, we'll just proceed on the presumption you'll create the follow-on thread Anon. I'll unpin this thread. >>11518 Heh, sounds like a whole other thread Anon. My goal was to consolidate, not multiply threads! :^) >>11481 We should have a new thread in a couple weeks or so Anon. Meanwhile you can make posts on the topics ITT, and I'll migrate them when the time comes.
Open file (125.10 KB 900x998 media_EyP59E3UYAIbjfC.jpg)
>>11520 > contextual AI These are all different terms for the same thing or parts of it. Of course some of us here are long aware of needing some complex system for conversations consisting of different parts which cares about context and parses current and past conversations, and sorts the data into graph databases, for example. There are all kind of terms for such advanced systems and it's parts, which can be used to find more information. Cognitive AI would be another such term to search for. Just claiming GPT is a chatbot is easier, though. Not that I have done enough to come up with a alternative, but more discussions related to it might help. It's a lot of things to look through, understand, and learn, which can be discouraging. >research community doesn't touch on this much That's why we're getting distracted over and over again. Businesses don't publish their developments in detail, but DL researchers advertise their findings. >but big business is all over it. Which could be reason for concern, since there's no general AI chatbot. However, that might be because they limit their efforts to customer support and such, I guess. >>11529 >Heh, sounds like a whole other thread Anon. My goal was to consolidate, not multiply threads Then I don't know what you mean by a "thread from the users perspective" and which one you want to use for it. The chatbot thread is full.
Open file (306.36 KB 728x588 K2XmKPw.png)
Question: Is anyone else missing the background mascot image of Chii in her nightshirt? I just want to confirm it. (the source image in the CSS is still up).
>>11537 Seems to be due to the content security policy of the server. Replacing it with a base64 image will bring it back: https://files.catbox.moe/w9gva5.txt
>>11541 Thanks! I've just discovered some kind of bug that's preventing me from updating it. I'll try to connect with Robi to let him know sometime soon. Hopefully the BG image will be back soon.
Help me out, what time zone is this server supposed to be on, the post times make no sense to me I guess I could look at my own post time and extrapolate the difference but I am really curious how or why we are offset by so many hours, in comparison to any North American time zone
Open file (96.83 KB 729x379 Selection_015.jpg)
If you enable JavaScript, then you can click the 'Settings' link along the page header bar, and then adjust the time display under the 'Other' tab.
Open file (78.60 KB 960x720 maxresdefault.jpg)
>>11545 Outstanding!
Will embedded YouTube videos work again, one day?
>>11620 I've just (re)emailed Robi again concerning this Anon, thanks. In the meantime, I'd suggest you rely on youtube-dl . >--- Date: Tue, 20 Jul 2021 04:18:13 +0000 From: robowaifu atddd cock dotddd li :-DDDD To: admin atddd julay dotddd world :-DDDD Subject: embedded youtube videos broken AUGGHHHHH! Will we be getting embedded youtube videos back, Robi? -anon's question regarding: https://alogs.theguntretort.com/robowaifu/res/8492.html#11620 -example URL containing YT links: https://alogs.theguntretort.com/robowaifu/res/2991.html#4797 -embedded video error message result: "This content is blocked. Contact the site owner to fix the issue." -regards Chobitsu >=== -minor email obfuscations -hand-patch gunt links
Edited last time by Chobitsu on 07/20/2021 (Tue) 04:30:56.
>>11620 >>11622 Sorry for not noticing this, my bad. I recently added CSP rules to prevent any sort of potential XSS, and forgot to consider that embeds need to be enabled too. Working on it now, they should work again soon.
>>11626 Alright, they seem to work again for me. Make sure to clear your cache if they don't work for you.
>>11626 >>11627 Thanks Robi everything's working fine now, much appreciated. We even have our beloved mascot back now too. :^) Cheers mate, hope things are going well for you Robi.
>>11470 Wow. I want whatever it is thats going to be put out. I can imagine a single artist being able to create an animated waifu... in film or in videogame... using nothing but a source set of keyframe poses.
- Risk of disruption in global logistics bc shipping labor shortages: https://youtu.be/GABLJbNvTIc - maybe don't order parts at end of the year, do it now. Especially if it's not something expensive but just some cheap electronics or mechanical parts. - GPU prices went down a lot on the other hand, but might go down further. Still not at MRSP (the recommended price when they came out) last week. >>11627 Yeah, thanks for all.
>>11697 I think an economic and potential societal collapse in the USA is going to happen soon. Else it will be a very slow and painful depression where we won't even be able to afford anything outside of food/water/shelter. I've been wanting to really focus on programming and trying to "catch up" to what seems to be 3 decades worth of knowledge, but instead I have been setting up a farm making sure I can at the very least feed myself if things get bad. I'm almost there, I can produce my own electricity, collect my own water, and soon have a stable way to produce my own food. It's going to be shitty if society falls apart before we can make significant progress on robowaifus.
>>11701 its been a concern of mine too but I don't see us "collapsing" into something irreversibly pre-industrial. If anything it would still be like the poorer parts of Star Wars where they still have technology it's just old, re-purposed and fixed up with gum and duct tape. That being said, is there any reason why we can't have IRL collaboration? It would show a level of seriousness if some of us had actual workshops and space dedicated to this cause. I'd be the first if I was at a better place in my life. (It is getting there, but going through a divorce at the worst possible time, and then Covid right as I was getting back on my feet) Idk what I have for a vision of this other than maybe a shared or accessible space with spare parts, 3d printers, resources, etc. A Robowaifu Research Institute (and Laboratory) in so many words.
>>11704 That's called a hackspace with 3D printers and maybe other machines, which might be run by non-profit organization or as a business. There are podcasts and video talks on how to set something like this up, if you search for it.
Hi guys Looking for introductory robotic hardware resources. The very babbiest of stuff. Words like "arduino" and "DC motor" just fly over my head currently. I guess I could start with some kids craft projects just to dip my feet in.
>>11723 Then search for videos on YouTube on those topics like "dc motor explained" or " introduction to Arduino".
>>11723 Hello Anon, welcome. If you actually want to try your hand at it, then you might have a look at anon's OSRM thread for ideas (>>11446). You could also look into the Waifu Materials thread for inexpensive materials ideas ( >>154). If you're simply trying to gain some basic understanding overall, then I'd suggest you look into various children's robotic toys, particularly the humanoid ones, and familiarize yourself with how they work internally. There are still a wide variety of videos available for them on YouTube at present. BTW, you can feel free to ask any 'silly' questions ITT, we welcome newcomers on our board.
Open file (51.78 KB 600x740 44824818_p11.jpg)
>Also, any doll owners or prospective doll owners, I would recommend you go to the gym and start lifting. It makes moving the doll around, especially laying her down MUCH easier. You don't even have to do a crazy amount. Just start with a basic bench press, deadlift, row, and arm routine. I find it hilarious imagining a bunch of buff classy men being robofuckers, probably rich too from pioneering robowaifus. I'm going to start lifting for my waifu.
>>11729 Good idea. I did that for a 3DPD, so why not much moreso for an actually-classy companion instead? Anyone have some robowaifu-themed DO IT FOR HER's?
Open file (445.95 KB 1400x1000 aigis.jpg)
>>11733 It will also increase bloodflow and productivity.
>>11744 Perfect, thanks bro. Aegis is top-shelf robowaifu materials for sure. B2 or Chii would also be great too. Now I want to figure out how to arrange having this picture where I can workout again.! :^)
Open file (678.56 KB 2300x1900 Humanity is saved.png)
>>11704 >its been a concern of mine too but I don't see us "collapsing" into something irreversibly pre-industrial. I think we will, as the USA doesn't have any real "industrial" plants as is. Once the inner cities can't get their gibs, it will devolve into absolute chaos and maybe even cannibalism. >That being said, is there any reason why we can't have IRL collaboration? Fear and paranoia. There is the possibility, of working together to move to a red county and out-numbering the current inhabitants to fill the positions of the county government. We could stop a lot of the "woke" laws and make it where it is hard for the welfare policies to creep in. This is already happening, but some people have seemingly completely bought up certain counties, but their intent isn't robowaifus lul. I've talked to some CS/CE engineers already IRL, and I am sure they lurk here now, but they absolutely hate talking about it IRL and are also paranoid. Personally, I think it might be better to keep it digital right now, but progress will be slow and secretive. I think we may see a "game" released that will double as an app that works as a type of secretary bot. Although, the other issue is the current fact that we don't make certain electronic parts (chips/semi-conductors) state-side anymore. Which inhibits any near future mass production of a simple physical bot. While I am hung up on making a chatbot, I think perhaps making a bot that can do household chores would be a huge hit and be a good source of funding for any future endeavors. We will either see some (((big company))) come in and try to high jack the idea of robowaifus, or we will see a random "leader" anon that will set us on an opensource path to greatness. Although, we can assume there WILL be push back once women realize where it's leading.
Open file (1.98 MB 1522x1588 IMG_20210728_014524.jpg)
can someone explain the technology of how sentence is built to convey or answer or tell something? I am trying to write for fiction but i find myself lost and inconfident about the content so i m going to complete basic just to know if the stuf will work please dont say "find something fun or interesting" as those dont work well for me so i hope there d be something more basic or conceptual about this
>>11826 It's a really good question Anon, but not one that bears it's own thread. I'm going to move this into the QTDDTOT thread. Maybe you'd want to expand it into a more detailed inquiry later?
>>11779 > I think perhaps making a bot that can do household chores would be a huge hit and be a good source of funding for any future endeavors. I think you are exactly right Anon, and it will be far more than just a 'good source'. I'd predict the company that capitalizes on this will have many billionaires among it's founders within five years or so.
>>11725 >>11727 thanks guys i've been looking into it are there any free simulators i can use? kind of like the bridge software we did in highschool? i think maybe that would be better than just dropping money on stuff out the gate.
>>11848 Sounds good. Not sure of exactly what kind of simulator you're thinking of Anon. There's a robotics simulator mentioned (>>10531, >>10536). There's also a general electronics 'game', and an online simulator for the 6502 chip that I can think of.
>>11826 I'm no expert on natural language processing (NLP) Anon, but I think one reasonably-sensible sounding option is to break a sentence down into the parts of speech each word is (noun, verb, adjective), and then chain those together in a correct way to extract out basic phrases or statements in sentences. So "See Spot Run" most basic fact is Spot Run (noun, verb relationship). Does that kinda make sense?
>>11704 I wouldn't mind IRL collaboration with someone I trust but I live in the middle of nowhere off the grid. It's just not practical. Also building a lab too soon would just be burning money. I estimate I'll need at least $50k to make a full-size robowaifu and $2k to comfortably start printing a body and prototype, so I'm focusing on making big fuck you money first. Without money there's no chance anyone is going to build a robowaifu. Without a positive cash flow whatever workshop is created will shutdown.
>>11826 I don't understand how this is question was meant. You won't get around doing some reading and watching explainer videos. There are prescripted responses with the option of changing a part of the response (AIML). Then there are text generators based on statistics like GPT-3. Then sentences could also be generated by a program based on data in a database. >>11848 Simulators for what? Arduino?
>>11827 ok, thanks alot. i hope it gets answered
>>11852 Words of wisdom IMO.
I have to admit, everything going on in this board is a bit hard for me to follow. This is just a suggestion, and I have neither the skills nor the time to be doing this but; what do you guys think of creating a news letter? Like the one smug/a/ is creating. It could be useful to keep track and updated on the progress Anons have made in all these different threads all in one contained news letter. Still, its just a thought and I doubt I can contribute to it but maybe there are some Anons here who can. Or maybe I'm just retarded and the only one who can't follow the board.
>>11873 >Or maybe I'm just retarded and the only one who can't follow the board. No, that's definitely not just you Anon. It's a vast arena of different domains, and it's always been a 'glorious' chaos using an imageboard to try and keep try and keep track of it all. As anon pointed out this can be quite beneficial to the creative process: >The advantage of this is you can develop ideas in a chaotic way that cannot be replicated with hierarchical ordered cultures (>>11355) I myself tried to overreach trying to convince anons to keep things orderly and finally did a >tableflip.exe over it. I'm over that now and accept that's just how it's going to be, insofar as the board /robowaifu/ itself is concerned. A newsletter is a great idea IMO, and maybe you can present something along this line to the OP of the robowaifu podcast thread? (>>11364) I personally can't add that to my plate ATM.
>>11873 I'd suggest thinking about also setting up an RSS feed for your online newsletter if you do this.

Report/Delete/Moderation Forms

no cookies?