/robowaifu/ - DIY Robot Wives

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

The canary has FINALLY been updated. -robi

Server software upgrades done, should hopefully keep the feds away. -robi

LynxChan 2.8 update this weekend. I will update all the extensions in the relevant repos as well.

The mail server for Alogs was down for the past few months. If you want to reach out, you can now use admin at this domain.

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)


He was no longer living a dull and mundane life, but one that was full of joy and adventure.


/robowaifu/meta-6: Christmastime Out In The Sticks Chobitsu Board owner 12/12/2022 (Mon) 23:01:06 No.18173
/meta, offtopic, & QTDDTOT General /robowaifu/ team survey (please reply ITT) (>>15486) >--- Mini-FAQ >A few hand-picked posts on various /robowaifu/-related topics -Why is keeping mass (weight) low so important? (>>4313) -How to get started with AI/ML for beginners (>>18306) -"The Big 4" things we need to solve here (>>15182) -HOW TO SOLVE IT (>>4143) -Why we exist on an imageboard, and not some other forum platform (>>15638, >>17937) >--- -Library thread (good for locating terms/topics) (>>7143) >note: There's a simple 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 latest code should always be maintained within the Library thread's OP & also on the current /meta. -Latest version of Waifusearch v0.2a (>>8678) >note: We periodically release the full archive of the board's thread's JSON files. This is useful for robowaifu researchers in general, and is explicitly relied-upon by Waifusearch. -Latest version of /robowaifu/ JSON archives v221213 Dec 2022 https://files.catbox.moe/6rhjl8.7z if you use Waifusearch, just extract these JSON files into your 'all_jsons' directory for the program, then quit (q) and restart. >note: There's an archiving tool coded right here for /robowaifu/ that provides the ability to backup locally the posts & files from our board. It's named BUMP, and is basically a custom IB scraper. -Latest version of BUMP v0.2g (>>14866) >--- >note: There's a design document for the specification of general design and engineering choices for our basic Robowaifu Reference Model A Series (TBD). Please have a look at it, and collaborate together with us on it ITT Anon. -Robowaifu Design Document (>>3001) >--- >useful external resources wAIfu-collective's AI guide - https://rentry.org/waifu-diy-ai >--- -previous /meta's (>>38) (>>3108) (>>8492) (>>12974) (>>15434) >=== -minor subj edit -mini-faq edit -re-org notes/faq sections -add/rm 'community-cheer' notice -begin 'external resources' listing
Edited last time by Chobitsu on 01/15/2023 (Sun) 10:31:14.
>>18171 >I also want to focus on smaller models everyone can run so I won't be sharding giant models across GPUs except maybe for fun. You could use it to make smaller models from bigger ones? Anyways, my argument was mainly that the price and energy consumption might be the same. I also didn't know you could run Stable Diffusion on one 3060 or that you could join cards without this NVLink. I don't know that much about this stuff.
>>18185 expiring in one day at which time I will delete this post https://discord.gg/UyZqQ32C >>18188 sorry about that! I just thought we had lost anon and I've been trying to track him down for months >>18170 Understood. We all have different expectations but I am going all or nothing to create or at least lay down the foundations and brainstorming for a Real Life Dorothy/2B tier waifu (minus the combat theatrics obviously) Nothing about this is physically impossible. Power considerations perhaps but I have unique solutions for that, (see my 4 hurdles of the robowaifu for more on that, when I remember where I posted that!) FYI and more related to >>29 Robowaifu fiction I have a new Substack post today new substack https://metaronin.substack.com/p/i-need-to-show-you-something I'm still getting the hang of the writing thing so please give a like if it's not horrible (and your feedback if it is) thank you
>>18162 >Also I've been making a shitload of money speeding up my commissions with Stable Diffusion Gratz! >make hay while the sun shines. True in every productive endeavor of life!
>>18193 >I just thought we had lost anon and I've been trying to track him down for months Completely understood Meta Ronin, nprb. Nothing hard-and-fast I learned that lesson long ago now lol, just a gentle nudge in the right direction to help us all out. If you have to talk in w/e thread with someone, then by all means have at it! :^) Cheers.
>>18193 >(see my 4 hurdles of the robowaifu for more on that, when I remember where I posted that!) Waifusearch is a thing you should know this by now heh :^) waifusearch> 4 hurdles ORDERED: ======== THREAD SUBJECT POST LINK /robowaifu/meta-6: Christmastime >>18193 4 hurdles UN-ORDERED: =========== THREAD SUBJECT POST LINK /robowaifu/meta-4: Rugged Mounta >>15182 (4, hurdles) ' 4 hurdles ' [1 : 1] = 2 results --- >ps We just pushed a new version of /robowaifu/ JSON archives; current as of today (see OP ITT). >=== -add 'archives' notice
Edited last time by Chobitsu on 12/13/2022 (Tue) 20:56:59.
>>18203 >Waifusearch is a thing you should know this by now heh :^) caught me slacking >>18185 7 day link for you friend (I realized maybe 24 hours might not be enough) discord.gg/2qU4Zjy5 In other interesting news I was briefly on BTR (BreakTheRules) podcast yesterday got to shill robowaifuism a bit It went well I'll be doing more shows with them in the future.
I lurk this board as I have intentions of getting started in robotics and AI/ML. Unfortunately, I still haven't started because I don't know where to start. I checked out the different general threads here on AI, and there's a ton of information and resources. Unfortunately, this makes my problem worse as I just don't know where to start and feel overwhelmed.
>>18285 What are you interested in doing most? I recommend focusing on one topic that really excites you and approaching other areas of ML through your understanding of that. Do you have any programming skills?
>>18285 Hello Anon, welcome (and welcome to the club heh :^) . I'd suggest you simply look over the Elephant thread (>>4143) and go from there. You're simply discovering a phenomenon that's common to everyone here. BTW, this Anon has good advice IMO >>18286 .
>>18285 Is your goal building a robowaifu, of learning about these topics in general? There are plenty of introductions into AI/ML all over the internet, the important point is to get started with any of it. The main other way is to buy a cheap 3D printer, unless if you have a maker space close to you, then you can get into OpenSCAD or Blender. The third way might be to pick up some electronics projects around Arduino, which might be useful in other ways, but get you closer to build a robowaifu one day.
>>18247 >It went well I'll be doing more shows with them in the future. Sweet. Gambatte, Meta Ronin! :^)
>>18286 >>18291 >>18295 I wanna specialize in AI/ML. For now, it seems there's not any implementation of true artificial intelligence in robots beyond vision. I wanna work on the robots brains when eventually in the future, AI is actually used in the robots.
>>18300 There are different opinions on how to approach this exactly, but being familiar with some of the available models and how to run them is going to be useful in every case. Like, e.g. Whisper for speech recognition.
>>18300 >I wanna work on the robots brains when eventually in the future, AI is actually used in the robots. Since you're here on /robowaifu/, I'll presume you're implying humanoid (specifically, gynoid) robots. That being the case, then the 'robot brains' will need to closely emulate the human brain to be satisfactory to men everywhere. That implies that a solid understanding of human brains/behaviors is certainly a necessary pre-req for success with your goals Anon. I certainly recommend caution against entertaining liberal/humanist philosophies on the topic and rather instead stick strictly with the scientifically-observable phenomenon involved. This should give you a solid foundation for working on robowaifu minds, Anon. Good luck!
Open file (438.77 KB 900x560 Touhou.full.946314.jpg)
>>18300 The first thing you'll need is Python and PyTorch. https://www.python.org/downloads/ https://pytorch.org/get-started/locally/ I recommend using pip since Conda is bloated and a bit complicated. Conda is only necessary if you're using projects with legacy code. If you have an Nvidia GPU select CUDA with the appropriate version or lowest one if you don't know. If AMD, select ROCm. Then install the Huggingface transformers library and learn how to use it: https://www.youtube.com/watch?v=QEaBAZQCtwE This video will show you how to quickly use language models to do different stuff like sentiment analysis, text classification and text generation. Don't worry too much about understanding how they work at first. Playing around with them will naturally make you curious what more you can do with it. Alongside having fun, work on deepening your understanding a bit every day. These free lectures are a gold mine. >Python Full Course for Beginners: https://www.youtube.com/watch?v=_uQrJ0TkZlc >Intro to machine learning: https://www.youtube.com/playlist?list=PLdAoL1zKcqTW-uzoSVBNEecKHsnug_M0k >Intro to deep learning: https://www.youtube.com/playlist?list=PLqYmG7hTraZCkftCvihsG2eCTH2OyGScc >Intro to artificial intelligence (mostly symbolic): https://www.youtube.com/playlist?list=PLUl4u3cNGP63gFHB6xb-kVBiQHYe_4hSi >The Long-term of AI & Temporal-Difference Learning: https://www.youtube.com/watch?v=EeMCEQa85tw >Intro to reinforcement learning: https://www.youtube.com/playlist?list=PLqYmG7hTraZBKeNJ-JE_eyJHZ7XgBoAyb Also find good textbooks on linear algebra and discrete mathematics and work through them, plus calculus and statistics. You'll need them. I recommend: >Introduction to Applied Linear Algebra: Vectors, Matrices, and Least Squares by Stephen Boyd >Discrete Mathematics and Its Applications by Kenneth H. Rosen >Calculus: Early Transcendentals by James Stewart >Bayesian Statistics the Fun Way: Understanding Statistics and Probability with Star Wars, Lego, and Rubber Ducks by Will Kurt >Probability Theory: The Logic of Science by E.T. Jaynes There's a lot of material here so pace yourself. You're not going to learn it all in a few months or even a year. Get into the habit of reading research early on and following the latest developments. Tutorials and what you can find people talking about on the internet is generally about 2-3 years behind, even for really popular topics like image generation is 6-12 months behind. If I were a beginner getting started today wanting to work on the brains of robots tomorrow, I would start with research involving prompting language models and work my way into understanding how these models work, finding their limitations and figuring how to improve them. >Chain-of-Thought Prompting Elicits Reasoning https://arxiv.org/abs/2201.11903 >Large Language Models are Zero-Shot Reasoners https://arxiv.org/abs/2205.11916 >Inner Monologue: Embodied Reasoning through Planning with Language Models https://arxiv.org/abs/2207.05608 >Interactive Language: Talking to Robots in Real Time https://arxiv.org/abs/2210.06407 >Large Language Models can Self-Improve https://arxiv.org/abs/2210.11610 >Training language models to follow instructions with human feedback https://arxiv.org/abs/2203.02155 (maybe a little too advanced for a beginner but this is the magic behind ChatGPT) Don't forget to hold onto your papers. Bookmark and organize them because you'll have at least a 1000 saved one day and thank yourself if you took the time to make them easy to find. To find new interesting ones go to Google Scholar, look up your favorites and click 'cited by' to find new ones or go look at what new papers the authors have published. And of course read any interesting ones you find cited. Aim to read about 1-3 per day. Learn to use Anki or another flash card program. You're going to be compressing a ton of information into your head and it will accelerate your learning immensely. >Anki walkthrough: https://www.youtube.com/watch?v=WmPx333n5UQ >Download: https://apps.ankiweb.net/ And lastly, if you get lost on how to do something I recommend using the Hello search engine: https://beta.sayhello.so/ You can ask it questions in natural language and it will generate answers based off web results. It's not always right but it saves a lot of time from getting lost in unhelpful or difficult to understand Google results. If you do these things you'll figure out the rest. I started from nothing a few years ago, not even understanding how to do a dot product and now I'm comfy being able to understand and implement about 90% of the research I read.
>>18306 This is gold, Anon.
Hope you got the chance to enjoy the /christmas/ party this year, /robowaifu/. BTW, there are still a few more events over the next week or so before the actual Christmas holiday. More information is here: https://anon.cafe/christmas/res/188.html Cheers
>>18306 >>18302 >>18301 Thnaks, anons. I'll do what you said.
>>18306 btw do I only need to learn Python? I heard AI models are increasingly written in C++ these days.
>>18316 Model development mostly happens entirely in Python. It's becoming more common though for people to port their models to C++ and custom CUDA kernels. This is only really beneficial when you're writing custom CUDA since PyTorch has a feature called TorchScript that can export Python models and run them in C++ free of any Python dependency. If you're interested in doing more than just the brains and actually integrating AI into robots, I recommend learning C++. If not, then it's not necessary but would be beneficial to understanding algorithms better.
>>18317 Thanks. I do wanna integrate robots with AI. Maybe I'll start with C++ after I'm done with Python. but how will I know if I'm "done" with learning Python? I already have some experience making some programs and scripts in Python, but not Tensorflow. When will I know I'm good enough to move on?
>>18332 Don't just learn stuff you don't need. Try to learn the basics, the things you need and general concepts used in programming. Know where to look up stuff and vaguely how it is called. We have ChatGPT and such now anyways.
>>18332 >When will I know I'm good enough to move on? Once you figure that out, I'll be able to help you with C++ itself.
>>18332 There will always be more to learn but what matters is if you feel comfortable being able to successfully implement and realize any project ideas you have. When you can express your ideas in Python then you're fluent in it.
Chobitsu, can you add this to the board's CSS to fix the vertical posts? .innerPost { min-width: 500px }
>>18347 Done. Please let me know if that fixes if for you, Anon.
>>18348 Thanks, looks good!
>>18349 Glad to help, and thank you BTW. :^) >also I'm thinking of base64 embedding our Chii mascot directly into the file. Do you have any thoughts or advice about this anon?
>>18350 AVIF files are supported by most browsers now. https://files.catbox.moe/kzlivq.txt >8x compression ratio
>>18372 Excellent! Much faster now goodbye imgur! :^) Thanks again for all your help Anon, cheers.
>>18306 Which one is better? Tensorflow or Pytorch? I always hear one or the other being suggested but never heard what advantages one has over the other.
>>18388 PyTorch is good for developing new models and ones that require dynamic graphs or inputs of varying size. It's easier to learn and use and so it's mostly used by researchers. It's also used by Stable Diffusion since its dynamic graphs can handle inputs of varying sizes instead of only 512x512. If you want to tinker around with the latest tech, you'll need to know PyTorch. Tensorflow is better for optimizing and deploying models to production and used more in the industry. If you want a job in ML then you need to know Tensorflow, although businesses are now starting to use PyTorch as well.
>>18173 Marry Christmas robowaifu!
>>18408 Still haven't gotten this board to stay on my time zone but hey, it's Christmas somewhere. Merry Christmas, /robowaifu/ !! I can say my life has actually been enriched and benefited by this board and this community having rekindled my interest in sciences, robotics, programming, art/design, project management and even writing and sci-fi - a far cry from being an angry-drunk pagan larper. so thank you for that much!
>>18409 Great to hear! I've also been inspired by this community, so happy to see people passionate in building robot waifus.
Open file (19.40 MB 11264x7168 00003.png)
>>18173 Merry Christmas! It has been a blessed year filled with the joys of the lords wonders. Though we've all found frustrations, I'm confident that next year is going to be really big for us! I'm grateful for the support, kindness, and knowledge we all bring to each other! Someday we'll spend Christmas with our waifus in warmth and love. I'm praying to spend next Christmas with Aigis.
Merry Christmas, /robowaifu/! This community is a great encouragement to each other, and in some ways we here serve a a candle of hope for anons the world over. MERRY CHRISTMAS BTW /animu/ is playing a very comfy Christmas stream right now. Highly recommended, and I'm listening r/n too. Cheers. https://prolikewoah.com/animu/res/85538.html
>Thanks! Merry Christmas, Chobitsu, I'm thankful for you and all that you do. Someday we'll achieve Chii together. Heh, we and all our like-minded cohorts around the world will eventually achieve far more than just Chii one day!! :^) I surely appreciate the sentiment Anon.
>>18409 >project management You do seem to have a talent for it Meta Ronin. One of the long-term goals I've had 'in the corner' for years a while now is somehow consolidating and streamlining the information here. Think this is something you'd have time & motivation to help me with? This place is a little overwhelming/confusing to newcomers, even motivated ones. Once we start delivering real plans for kits and demonstrate them ourselves, we're likely to have a massive influx of newcomers in short order. It would be nice to have the board better-organized before then IMO.
>>18410 >I've also been inspired by this community, so happy to see people passionate in building robot waifus. It is really inspiring to all of us Anon, thanks to you and to every Anon here for being a part of it! >>18414 >Someday we'll spend Christmas with our waifus in warmth and love. Indeed we will. It will be a lot of hard work, but it will be so worth it in the end! :^)
Open file (359.23 KB 384x640 tmp42fkwz6_.png)
>>18409 Merry Christmas Meta Ronin! Praying you'll have a blessed day. It has been great to work with you. Your enthusiasm and creativity are valued.
Open file (2.48 MB 2560x1536 00005.png)
>>18418 Merry Christmasu Chobitsu! Praying you'll continue to bring the spirit of good will and community to this blessed project. Glad to have you as a friend.
Open file (306.88 KB 384x640 CiboChan.png)
Merry Christmas Pareto Frontier! I still believe that together we will bless our waifus with wonderful AI. Praying your day is filled with great cheer. Used AI/art skills to make you a cute Cibo. Neither I nor Stable Diffusion could accurately recreate her manga aesthetic. Hope you like her moe-ified version!
Open file (370.81 KB 640x384 tmpsszri657.png)
Open file (373.72 KB 384x640 tmp_44y_a6s.png)
Open file (351.27 KB 512x512 tmpljmvheoa.png)
Open file (251.36 KB 512x512 tmpjzvqqr7r.png)
More Merry Christmas cheer for all who are here. AI Christmas art of some of our most loved waifus.
You're a real blessing to everyone here Kiwi, thank you. >>18437 Thank you kindly friend. Together, let's continue to bring good will and community to this little band of brothers. It's the least we all as a group can do for each other, and it's important to instill hope to the larger world of anons everywhere. They're counting on us!! :^) BTW, /animu/ is streaming a wonderful Christmas stream again r/n Catch it while you can! :^) https://prolikewoah.com/animu/res/85538.html
>>18306 >Conda is only necessary if you're using projects with legacy code. I know you were talking about code you run yourself, but Conda is the best way I've found to deliver a python program to layman Windows users. The "standard" PyInstaller is much less reliable. You can bundle a conda environment in a way that it runs out-of-the box. Conda-pack + some basic .bat scripting lets you create a zip file that users can unpack and run by double-clicking main.bat or whatever. The best practice I've found here is to create a pip package so Linux and Mac users can use that directly, and wrap the pip package in a conda environment for Windows users. In every other regard, I've found Conda to be inferior to Docker. If your users are willing to install Docker, the best "packaging" solution I've found by far is a zip file that contains 3 files-- run.bat, update.bat, uninstall.bat-- all of which are 1 or 2 line docker commands. It's trivial to create a similar package for Linux and Mac, and these scripts pretty much never need to be updated since usually updating the container is enough, and that can be done without distributing new scripts. My usual run.bat maps the working directory into the container so it's very clear to users which files my program can and can't access. Example: https://drive.google.com/drive/folders/1xROohg-r_10asrz3xGhh7iMSXIa0vyQX?usp=share_link. And while I haven't done this myself, it should be easy to have your main.bat run a docker container, expose a port on localhost, and point a browser to that port to act as a user interface. Merry Christmas, everyone! Or, I guess in my native tongue, Happy Hearth's Warming!
>>18429 >consolidating and streamlining the information here. Think this is something you'd have time & motivation to help me with? I think I can spare some bandwidth since a few of my other tasks are blocked for now. I'll put this on my task tracker
>>18436 just caught this thank you Kiwi!!
>>18453 Great! Since I'm probably 'locked in a rut' of tunnelvision on this topic (since I've been thinking about it, with little of other's discourse/input, for years) I'd like you to start off with suggestions of your own first, please.
Open file (797.50 KB 1200x900 Good_jerb_Clarence.png)
Hey there /robowaifu/. So, you may know that /comfy/ suddendly found themselves in a bind a couple days ago. >tl;dr the BO rm'd the kit-n-kaboodle They are getting back on their feet, but they could use a few hands. If you aren't doing research!! and can make the time to, please go over to their 'rebuild the board' page and lend a hand? https://anon.cafe/comfy/res/34.html This will be like a Christmas miracle for the entire Internet if the board was brought back up quickly! :^)
>>18462 Ok I have a few ideas perhaps. Let me put them down in writing and I will respond here. Been very busy this week but I should be able to spare some bandwidth tomorrow
>>18519 Looking fwd to it Meta Ronin.
Open file (349.42 KB 3120x4160 20230112_025620.jpg)
>>18462 well it has gotten to that point where I just have a single physical board to track anything r/w related. Things are finally starting to move along as i gain momentum but I noticed that little item in the lower left corner and remembered we wanted to tackle reorganizing the board a bit. What kind of a re-org did you have in mind? A few suggestions that came to mind include - Merging threads that are very similar and get cross posted a lot - Archiving or pruning dead threads (last 5-10) - The possibility of sub-boards but idk if you have that permission here or if that's really what you wanted to do. One idea would be to have a board for culture, media casual discussion and another board for design discussion and projects. Does anything else come to mind that we should add to this list?
I was just looking into Thingyverse to find some 3D designs of human parts, since I saw that importing into OpenSCAD generally works. Anyways, I found directly one more project which would be interest here: Polaris - jointed ball doll from 2018! https://www.thingiverse.com/thing:3184441 This also makes me wonder, how many lurkers we (allegedly) have here, that could look for things like that and post it into the correct threads. Others could print existing dolls and work with the AI chat people here to just make dolls with some simple AI voice chatbot for now. >>18656 >board for culture, media casual discussion and another board for design discussion and projects Idk. Imho, we need only more recruiting and maybe a thread guiding people towards tasks they could do. Like covering one area and test out stuff. Experimenting with cheap silicone rubber, printing existing design, looking for parts we can use on sites like Thingiverse and others. Or tagging existing postings and infos we already have, so it could be used in something like Mermaid UML editor would also be great. Just some examples.
>>18707 Oh yeah, and the worst part: >Dec 11 2022: due to one too many people taking the noncommercial license and selling without permission or credit such as the most recent thief TanitaDollOutfit on etsy who refused to listen to me, I don't want to deal with the headache of stressing and requesting take downs anymore. Additionally Solomon262 took his doll down too probably due to the same reasons. sorry guys. I'm out Just downloaded it, it's mostly pictures.
>>18656 >>18707 Thanks Anons, I'll try to respond better in the next day or two. >Mermaid UML editor Might be a good idea. It support interactivity right?
This is a general PSA: if you can program and you haven't tried Obsidian (note taking app), I highly recommend it. I'm very fond of exporting my thoughts into a computable format, and every other app I've tried is terrible at letting you organize, navigate, and script your notes. The first image is auto-generated. It's my "default" note, where I put a summary of immediate concerns plus a scratchpad. - I create a new note for each meeting, and I pre-populate it using a plugin called Templater. The pre-populated data contains lines for the meeting info plus a #meeting tag. - There's a plugin called DataView that lets you go through your notes, filter based on content & metadata, and present them as lists, tables, search results, and task lists. All three tables in the image are auto-generated. - The first table goes through all my notes containing the #meeting tag, and it uses the pre-populated data to give a summary of upcoming meetings I have. DataView lets you write custom JS to decide what to display, which I use to make the table easier to skim. - The second table searches through my notes for tasks. Tasks are specified through markdown by starting a line with "- [ ] Example incomplete task" or "- [x] Example complete task". DataView lets you add arbitrary metadata to a task, which I use to optionally set a due date. You can use Templater to add whatever you would consider standard metadata. - The third table contains a list of notes I have tagged with #ongoing, ordered by the last time I updated them. The second image (graph) is created using links between notes, filtered so that only notes tagged with #variable and #ai-research are displayed. Obsidian has built-in functionality for creating these graphs. I use Templater again to pre-populate metadata fields and tables for each #variable. Some other highlights: - Before DataView + Templater, I never realized how much I needed the ability to script my notes. - There's plugin for continuing selected text using text-davinci-003. I use it, though I find ChatGPT much easier to use than GPT-3. - All notes are markdown and easily accessible on the filesystem. Everything is stored locally. You can transfer everything across devices easily using either (paid) Obsidian Sync or github. - You can customize the UI through CSS. There's a built-in debugger like Chrome's DevTools panel. It contains an element picker that lets you navigate to the element in the DOM plus a panel for viewing CSS properties. - Everything I've tried so far, including my custom Javascript for making tables pretty, works across the desktop and mobile. - It supports sketch drawings through Excalidraw, and you can get libraries (plugins) for Excalidraw for common types of diagrams and UIs. - Everything is markdown. Grep and sed work perfectly. I've only been using it for two weeks now, but it has dramatically improved the efficiency with which I can document my thoughts on topics I care about. With the graph view and Excalidraw integration, it's also dramatically reduced the cost of explaining things to people I work with. I have thoughts on how to use this to keep up with research and github repos more effectively, mostly involving scraping, GPT3-generated summaries & keywords & embeddings, and auto-generated tables & graph queries. If I get something working for that, I'll post it in the R&D general and AI News thread.
>>18707 >This also makes me wonder, how many lurkers we (allegedly) have here, that could look for things like that and post it into the correct threads. Others could print existing dolls and work with the AI chat people here to just make dolls with some simple AI voice chatbot for now. I figured I'd map this out more explicitly. See pic. - Green refers to actual projects, repositories, or websites we can use. If I were doing this more thoroughly, I'd probably add research directions to each node where I thought it was relevant. The arrows would go from the research to the green nodes that use it or from a green node to the research it enables. - On the firmware side, I don't know anything about creating practical robotic control software, hence I have no projects attached to that. On the doll side, I know basically nothing, hence the only concrete thing I've attached there is Thingiverse, which I got from this thread. I created this using the Templater plugin and the Dataview plugin. - Install both -> configure Templater to get scripts from a "Templates" directory -> Create a new note titled "RW variable" in that directory and paste in https://pastebin.com/ZCfBrcL4. You only need to do this once. - I have a folder called "Graphs" and I created a sub-folder in there for RW. You only need to do this once. - To create a variable, I create a note, use the hotkey Ctrl+E to see templates, then select the RW Variable template. Under the "generates" field, I add "[[TargetNoteTitle]]" for each node I want to link to. Under tags, I add #project or #outcome as appropriate. In the pick, all of the green nodes are tagged with #project, and the blue node is tagged with #outcome. - In the file view, sort by last modified -> select all of the relevant notes -> move them into the Graphs/RW folder. - Open the graph view, and (1) set the filter to path:"Graphs/RW", (2) create a blue group for #outcome, (3) create a green group for #project, and (4) create a red group for #variable -path:"Templates/". The groups will persist, so you only need to do that part once. >how many lurkers we (allegedly) have here Distributing tasks doesn't seem to work well in decentralized environments like this one. The most we can do is make people aware of how to help & give them the tools and training materials to do so. Any bottlenecks should be high-priority candidates for automation.
"Fireworks in the heavens" apparently isn't just a metaphor now. > Two white dwarf stars collided ~850ya in Cassiopeia, apparently just for our amusement! :^) https://www.youtube.com/watch?v=N-FkouaG818&t=1082s
>>18723 >>18725 Neat! I'm going to try this out Anon. Maybe I can help contribute some things about embedded & systems programming to your work. >Distributing tasks doesn't seem to work well in decentralized environments like this one. True, particularly if we all maintain the honored IB tradition of Anonymous. Specifically one of the reasons we officially namefag here isn't to be a bunch of attention-whoring faggots (against whom that rule was enacted), but simply b/c we are making some attempt at being a working engineering group together. Thanks for the tip about the tool Anon! Cheers. >=== -minor sp, prose edit
Edited last time by Chobitsu on 01/15/2023 (Sun) 06:19:24.
Open file (48.78 KB 960x443 space fumo.jpg)
Lol. Apparently Koichi Wakata brought an actual fumo aboard the ISS. > Anyone know which one she is? <ywn a Chibi Chii aboard the space station... >=== -minor sp edit -add funpost cmnt
Edited last time by Chobitsu on 01/15/2023 (Sun) 06:35:40.
Open file (1.15 MB 2960x2160 tools.jpg)
>>18716 >>Mermaid >It support interactivity right? It's more of a syntax to write diagrams in code: https://mermaid.js.org/syntax/classDiagram.html We had this topic here: >>10716. Someone always pitched ORM instead of UML, but there was no software for it. This here >>4143 is the thread for it, I crosslinked it back to this discussion here. Nowadays there's some online editor for Mermaid: https://mermaid.live/ >>18723 Thanks, unlike Mermaid it's even in my repository. Though, of course it's based on Electron and has 60+MB. However, your pitch sounds very intriguing. It looks in particular interesting for designing graphs, even for other (AI related) reasons. Thanks. For now I'm more biased towards Mermaid, it looks simpler if I can install it. I want to do the diagrams in code, that's the main idea. I think it creates markdown, so maybe it can even later be imported into Obsidian. >>18725 Your pastebin link doesn't work, the DOT link problem strikes again. >>how many lurkers we (allegedly) have here >Distributing tasks doesn't seem to work well in decentralized environments like this one. The most we can do is make people aware of how to help This is what I meant. Even keeping track of topics and making code for diagrams would help. I barely remember all the stuff I looked into during the last few years on all these topics. I also can't take track of everything going on outside of the board, where people might work on something similar. Like designing dolls putting them on Thingiverse or in the Dollforum. >>18727 It's not about a list with assignments to a user with a name, but a list with what people who want to help out could do, without learning all off it or trying to do so and then giving up. We either have a problem with recruitment or with engagement. Not claiming I couldn't do more myself (if I wouldn't get anger issues), but there are always more things that could be done. Just as spontaneous example: Using cheap silicone rubber for home installations to test out colorizing faces. There was an artist a while ago doing this really well. There are many YouTube channels on how to do makeup. Printing any face from Thingiverse and getting some colors for silicone would be enough. Maybe printing isn't even necessary. Just using some mask as a model, probably some clay and then testing how to make such faces look better. Without huge costs or general skills in AI and robotics. That aside, it's about keeping track of such topics on YouTube and social media.
>>18730 >It's not about a list with assignments to a user with a name, but a list with what people who want to help out could do, without learning all off it or trying to do so and then giving up. Yes ofc. Please explain how you would do just such a set of things IRL, at a company, without keeping track of who's who? Ever see the film 50 first dates ? Yeah, without some kind of handle it's sort of like that. But maybe everyone is right and we should just drop names here (or at least the intentional attempt at trying to make it work that way). Obviously for many OG anons it's anathema. Don't know how else to manage doing what you said and somehow delegating things otherwise tbh.
>>18730 Speaking of names, that's quite a handle you've got there Anon. :^)
>>18733 Ahm, yeah I just discovered this. I hope this can be removed.
Open file (67.29 KB 703x688 1673809817193045.png)
Open file (167.77 KB 872x1233 1673810641044798.png)
Top effing kek.
>>18735 Actually, we can't. The subject field can be edited after the fact, but not the name field. Apologies.
Open file (80.66 KB 1140x828 fidir.png)
Lorem ipsum dolor sit amet, consectetur adipiscing elit. >Duis varius, lacus vel pellentesque viverra, turpis nisi iaculis turpis, et vehicula massa mi vestibulum arcu. Morbi ornare malesuada libero posuere rhoncus. Vestibulum mattis et risus in pulvinar. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Praesent pharetra neque mi, at elementum erat scelerisque sit amet. Aliquam elementum arcu sed ipsum rutrum imperdiet. Suspendisse hendrerit nisl eget facilisis eleifend. Fusce auctor rhoncus facilisis. Curabitur aliquam risus ac vestibulum tincidunt. Etiam sodales nisi non velit posuere, molestie hendrerit tellus lacinia. Curabitur id massa efficitur tellus tristique accumsan. Pellentesque tristique, neque at eleifend commodo, diam velit auctor ante, id ornare tellus risus at sem. Aliquam non fringilla eros. Nunc nec sapien vitae felis varius aliquam a mattis lorem. Etiam eros quam, laoreet ac lacus ut, ultricies malesuada velit. >"Donec et ultricies lacus. Ut lacinia placerat nisl eu faucibus. Nam consequat rutrum arcu, vitae viverra ex. Duis facilisis, nunc vestibulum lobortis malesuada, erat metus mattis diam, sit amet pretium risus dui et massa. Phasellus nec dui id odio sodales viverra at vitae ante. Fusce porta eros dui, in dictum sem pellentesque vel. Integer sagittis felis et velit feugiat aliquet. Curabitur consectetur sed est vitae convallis. Aenean vulputate, leo a congue dignissim, eros tortor viverra quam, et euismod massa dolor in enim. Mauris maximus pulvinar ante eu condimentum. Aenean mollis porttitor sagittis. Sed maximus turpis interdum nibh malesuada faucibus. Mauris eu varius dui." Aenean tincidunt pharetra augue, sit amet pretium sem ullamcorper ac. Sed non vulputate velit. Ut feugiat cursus molestie. Vivamus finibus, dolor vel tempor lobortis, orci urna vulputate nisi, sollicitudin cursus arcu ligula in dolor. Pellentesque a tempor nulla. Ut velit augue, suscipit a lorem in, sagittis vulputate mauris. Nulla facilisi. Ut dapibus cursus consequat. Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. p118a: Nullam nec posuere metus. Proin mattis vehicula felis eget accumsan. Nunc feugiat et dolor vitae tempus. Nullam a velit sit amet tortor gravida molestie. In sit amet placerat sapien, at ultrices orci. Praesent elementum commodo turpis et rhoncus. Maecenas dapibus sit amet tellus eu rutrum. Fusce fermentum sed odio quis bibendum. Suspendisse eget est ut purus aliquet elementum sed sit amet diam. Pellentesque pulvinar faucibus tortor, id vestibulum sapien cursus ac. Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. Proin sed lorem id nulla sollicitudin condimentum id in mauris. Proin egestas ut enim id finibus. Donec vestibulum metus at dui rhoncus, id egestas enim bibendum. Interdum et malesuada fames ac ante ipsum primis in faucibus. p118a: Ut sed bibendum enim. In interdum bibendum lobortis. Aenean bibendum ante nec sem commodo, eu varius arcu elementum. Nam libero sapien, iaculis malesuada porta at, ultricies vel nibh. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Nulla facilisi. Vestibulum sed est sollicitudin, tempus diam sed, aliquam tellus. In quis eros diam. Pellentesque cursus felis lorem, fringilla bibendum est interdum vitae. Nam purus ex, gravida eu magna vel, sagittis laoreet turpis. --- >p118a command line + possible output: g++ -std=c++20 -O2 -Wall -pedantic -pthread ./ch_04/main_p118_a.cpp && ./a.out 5 7 9 4 6 8 Kant Plato Hume Kierkegaard --- >p118a command line + possible output: g++ -std=c++20 -O2 -Wall -pedantic -pthread ./ch_04/main_p118_a.cpp && ./a.out 5 7 9 4 6 8 Kant Plato Hume Kierkegaard --- >p118a example code: https://rentry.org/fidir https://coliru.stacked-crooked.com/a/cbbeef56fa13a0f9 >p118a example code: https://rentry.org/fidir https://coliru.stacked-crooked.com/a/cbbeef56fa13a0f9 --- >reference/further reading: https://en.cppreference.com/w/cpp/container/vector
Edited last time by Chobitsu on 01/16/2023 (Mon) 04:58:25.
>>18746 I'll be starting a C++ working textbook soon, this was just toying with some ideas. We'll have classes for Anon to learn C++ from scratch, and this will be our textbook thread for that (>>18749).
Lol. Looks like the Globohomo are up to their shenanigans against Anglin again. Wonder what they'll claim it was about this time? :) Meanwhile, let's put this here for safekeeping. http://stormer5v52vjsw66jmds7ndeecudq444woadhzr2plxlaayexnh6eqd.onion/
OT: Just in case you are playing GTA V instead of helping here. They forgot to tell you something: https://www.itpro.co.uk/security/vulnerability/369913/gta-v-vulnerability-exposes-pc-users-to-remote-code-execution-attacks
>>19067 Thanks Anon, appreciated.
>>19067 no worries all I play are single player games pre-2012
>>19183 Just pick the size based on what works. Small one won't be able to hold computers, batteries and all kinds of servos. Taller means more difficult to move by herself or with help.
>>19183 Hi Kiwi, nice job on the comments yesterday. You really kept the place hopping! :^) Well, I don't Tw*tter, but my vote is basically the 1m to 1.5m range. I think the vast majority of Anons want a 150+ (cm), but our 'wants' don't change the fundamental laws of physics. The Square-Cube Law, roughly-speaking, says that 'smaller is better'. Since this is an engineering challenge extraordinaire, and since this is the initial prototyping phase, I'd suggest you keep it at a 100- size for now. Both the engineering, designs, materials, and actuators will all improve with time (all things being equal). Then, at a later stage we can move up to the 150+ designs. >$.02 >=== -minor prose, fmt edit
Edited last time by Chobitsu on 01/27/2023 (Fri) 17:00:25.
>>19183 6 foot plus since I have something of a fetish for bigger, taller girls. Shortest I'd accept is my height, so 5'9" but there's no option for that.
Open file (47.11 KB 674x674 hYC6DP6.jpg)
>>19196 >6 foot plus since I have something of a fetish for bigger, taller girls Ha, reminds me a bit of a dream I recently had where I was getting my robowaifu upgraded to the next model and when I went to give her a hug she tried to pick me up like pic related and I was like "noo wait I'm good" but part of me was like "hhnnng yes"
>>19046 (in the Meta Ronin Podcast thread) > basic arguments from BlackPill (Looks and Demographics) What I meant by demographics was even hard to find on Youtube, thanks to all those "dOn't takE the BLacKpill!!1!" videos, lol. The correct search term would've been "reverse demographic pyramide blackpill": https://www.youtube.com/watch?v=gpYJzd7JOb8 The point is mainly that aside from other problems there are not enough women. Just looking at the numbers of the two sexes is misleading, because of the old women surviving their husbands, and because it's an average over the generations while the younger ones are smaller. But some of the older men from the bigger generation will of course try to get younger women, and compete with the younger guys about the women from a smaller generation. This is essentially the argument. Then the other is, that more sons might be born by women in good times and more of them survive nowadays. Which is good on one hand, but it causes a male surplus. Lower birthrates also seem to cause a bigger male population, since later children seem to be more likely to be daughters. > pyramid scheme > demographic death spiral > more female entitlement > equality without entitlement increases female suicide rates I don't think the video linked above is right about MGTOW, though. Because MGTOW is about ones own chances, not saying anything about everyone. Therefore his argumentation has no foundation. I personally might be able to get some human female from somewhere for LTR or marriage, maybe by putting some work into it, but I really don't want to. She would still feel empowered by feminism and by seeing all these simps, for example. I also don't agree with the idea, of letting the wealthy countries become poorer or letting them be overtaken by other cultures, would be the best solution. We know better here at /robowaifu/ about the best solution. I just wanted to throw this in here, and not into the Podcast or Roastie thread, since it's a kind of meta topic. Good news for us about this is, that the "Roasties" will probably not feel threatened about robowaifus as much as some think (related >>1061). They are so busy with all those men being interested in them, and also throwing money at them. Pediophobia (fear of dolls), inflexible normies (/NPCs), and also people who are concerned about protecting current society and power structures might be the bigger issue for us.
>>19210 Good explanation Anon. Thanks for clearing that up for everyone. I think you're exactly right about your very last point. The status-quo is a powerful intoxicant drug to the powerful. Cheers.
BTW, is anyone here in contact with SophieDev? Is he holding up fine? I mean he's in Britain I think, and they're hit pretty hard by well everything.
Open file (35.84 KB 474x802 0.jpg)
>>19184 150cm is the limit of what can work while being affordable. Well, theoretically anyways. (Assuming anime girl proportions) >>19185 Well, kept twitter tweeting I suppose. Should post here more honestly. Physics is why I really hoped for 100cm to win. 150cm is as hard as you think it is. I understand the importance of a more adult height though. My 100cm prototype that I've posted a bit of felt weird to hold because it was so small. Part of why I made the poll, to see if others felt the same. Hoping to be wrong in my discomfort with tiny robot woman. Things have gotten better on the engineering side, PLA+ and cheap PETG are so much better than the crappy PLA you used to see me use. Steppers and brushless are far more affordable and easier to use. It's really been nice to see it happen in real time. Not enough to enable easy development of 150cm waifus but, that's why we're working. I'm thankful to have had you and your faith throughout this long journey. Praying 150cm Christ Chan waifus are finally possible this year. >>19196 Darn that's a huge bot! Would need to have some exaggerated anatomy to work past 6ft. If she was very thin and had very long legs, it could work. >>19207 Heh, super strong waifus like that would be possible with an onbaord microgenerator (beyond what currently exists) or having her plugged into the wall. >>19210 Too many lonely men need our robot maids to love them. I'm really hoping these waifus love will cure the simp of his sinful worship of women and lust.
>>19222 I don't know Anon. I hope so. We all certainly miss his contributions here. Yeah, Airstrip One has it harder than most, but pretty much all of us in the West are screwed by the Globohomo just the same. Robowaifus will help with that, so forward! SophieDev are you here bro?
>>19249 >Should post here more honestly. Lol, you do just as you see fit Anon. You're serving important parts everywhere you go! :^) I'm very glad to hear that the manufacturing process is improving. This is important. >size You absolutely have my full support and encouragement in shooting for 150+ robowaifus Anon. You're not naive about the physical constraints, so there's no real benefit to me trying to 'open your eyes' as a guide. I'm simply a fellow on the same path. If you want to go big right out of the gate, then go for it Kiwi!! :^) I would say that a big part of the physics issues around square-cube is that we are all quite frankly, well, made of meat. This is a rather massive material compared to say, air. While there are many engineering-related downsides to the challenge of men devising robowaifus, one of the great upsides is that we can make them hollow to some degree at least. If the volume is growing as the cube of surface growing as the square, what matters it much if our robowaifus are filled with air? >tl;dr Keep the mass down, and things will be fine bro (>>4313). >encouragement You too Kiwi. It's a real honor to 'serve' alongside every dedicated anon here, and your acknowledgements of honor toward God have been a blessing. It's with His help that we'll make it through (and only with His help :^) >=== -minor prose edit
Edited last time by Chobitsu on 01/29/2023 (Sun) 07:39:04.
>>19249 More than 150 cm isn't necessary for a full scale robowaifu for me. This is circa what I wanted, but I'm flexible in both directions. Though taller might be good to reach higher places in the kitchen, so I thought at least one of mine would be maybe even be taller than me. Different hights would be good for a guy going for a harem, having different types of girls. Simple and small ones would also be fine as maids or as companions without being sex-enabled, like the "Joystick" waifu without legs and only a hard-shelled dress around the legs with the motors inside. Guys can still hug her and being hugged, she could be carried to the kitchen and learn cooking, but would generally be more like an animatronic standing around. As another example, a simple maid could be 40-60cm high, without facial movements, driving around on motorized roller skates. I thought of her as addition to a Roomba-style robot. She could lift small objects out of the way of the ground cleaner bots or just move them a bit. Also cleaning the edges of walls where cats rub themselves, move objects like toys from the floor into some corner, being lifted into the bathtub and cleaning it, cleaning the bowl in the bathroom if she's high enough, being cleaned with a shower of water in the tub afterwards, ... I personally don't care if guys later use our frameworks for smaller and still sex-enabled waifus. I think it would even be legally okay, if she has bigger than human eyes, strong hip to waist ratio, some boobs, and maybe some other adult or even alien features like different skin color or pointy ears mixed in. That said, I wouldn't unnecessarily risk it, especially in the early stages.
>>19249 >Well, kept twitter tweeting I suppose. Should post here more honestly It's totally a good idea to draw in guys from other platforms. Our ratio of guys going from visitor, to lurker to developer seems to be abysmal, though. Same for guys starting to share the files of their (abandoned) projects. Idk, maybe a introduction to how image boards work would help? The rate of people doing things is generally low in every area, though. That was to quite some extent the reason why I switched from following the news and discussing politics to something that would improve my own life even if no one else would be interested in it.
>>19267 Interesting use scenarios Anon. You should think about writing some of these out long-form in our Fiction bread (>>29) IMO. >>19268 >It's totally a good idea to draw in guys from other platforms. I agree with this statement.
>>19252 Manufacturing is one of the most important and overlooked aspects of this project. Part of why my progress is so slow is because I need everything to be easily manufactured for cheap. >Hollow I'm doing all I can to keep her as hollow as possible. Also working to minimize the plastic in her outer shell to further reduce her mass to an absolute minimum. She's going to be ultra light.-- & ultra cute :^) --
>>19298 >Part of why my progress is so slow is because I need everything to be easily manufactured for cheap. Yes, that's a very important part of what our initial goals are here. One of the things I've toyed with is using already-manufactured items and looking for creative & aesthetic ways to integrate them into the designs. Particularly during this early phase, I would recommend against getting too caught up in details, nor allow the allure of the 'perfect' to sway any of us away from accomplishing the 'good'. >tl;dr Reusing things that can be purchased already can help with this! :^) >=== -minor prose edit
Edited last time by Chobitsu on 01/30/2023 (Mon) 12:28:15.
Open file (2.69 MB 1280x720 orangutang saw.webm)
Open file (519.77 KB 918x1200 sea slug.jpg)
Open file (21.87 KB 400x287 unhappy god.jpg)
Don't know if this is the right place to find an ear to put these ideas into action but whatever. TL;DR - "We" are going about AI wrong if we want an intelligent companion. GPT and text-generator AI isn't artificial intelligence, it's artificial schizophrenia. A human being isn't simply a text machine, it uses the word as a tool to express feelings and things known intuitively. The "intelligence" that has been evolved from a corpus of text isn't a real intelligence in the same way that a person is. It has evolved to take text in and spit it out, but it can't reason or intuitively "feel" like a person can. A person is also self-training and can ingest new, untagged and unsorted data. A person can ask questions and describe a concept in many different ways and words because it has a word-less mental model of what the "thing" is. A person learns by being thrown in an environment with zero knowledge, then interacting with the world and learning skills like speech, conversation, and encyclopedic knowledge. With "AI", we're trying to develop a human being backwards: we start with text and high-level conversation, without any emotions, values, or instincts to act as the substrate. If we truly wish to create a human-like intelligence, we don't need a wise man or a philosopher, we need a baby: a "dumb" creature that, given enough time, can learn to speak, walk, and communicate on its own, without an army of data taggers to filter the world. The video game "The Talos Principle" is a good example of what I'm imagining: an unstructured environment/simulation where, given a (virtual) body and enough time, a general, human-like intelligence will emerge. Natural evolution doesn't start with the philosopher, it starts with the jellyfish and gets gradually more complex until the descendants of the jellyfish can begin to reason and philosophize. If we could simulate a rat brain and gradually add a few extra neurons here or a speech center there and see how they fare in different simulations, I think we would be a lot closer to a robot wife than trying to stitch together various hyperspecialized idiot savants together into some kind of software human centipede. It's basically how we got dogs: we started with a wolf, something social and potentially useful but temperamental, and artificially selected for mental and physical traits desirable to human work and companionship. In fact, a humanoid cyborg with a dog brain may even be a desirable shortcut. You have a creature which is already somewhat intelligent, loves humans and has natural protective instincts (good for a mother). The only thing left to do is evolve the brain genetics into something more fit for a human mother/social companion. (Of course, discounting the logistical and development challenges of mass-produced wetware and the moral issues with "playing God".)
>>19313 >In fact, a humanoid cyborg with a dog brain may even be a desirable shortcut. Lol. I've said more than once here that if our robowaifus could simply attain a level of companionship exhibited by good doggos, then we'd be well on our way to dramatically-improving the lives of millions of men. Ofc, simply owning a IRL doggo has been the tried-and-true classic alternative, heh You have some interesting points Anon, and I think most of them have been bandied about here before now. The way I see it, the impetus behind so-called 'AI research' is driven by two primary, (mostly) accidental, factors: -These 'corpus of text' 's actually ensconce real human emotions, intellect, and wit. This is why ppl personify these chatbots so readily. -This approach actually works, by and large, for the domains it's intended (see part b of point above). Of course this isn't real intelligence, any more than these AI have real souls. But what's a soul? How do you 'do' a soul? I've started a thread intended to at least begin to frame these questions from the Christian perspective (>>17125) -- the only perspective I deem that will bear a high-degree of fruit. Great inputs, Anon. Please stick around and help us figure all these questions & answers out! Welcome, BTW. :^) >=== -minor prose edit
Edited last time by Chobitsu on 01/30/2023 (Mon) 15:06:14.
>>19313 >TL;DR - "We" are going about AI wrong if we want an intelligent companion. >GPT and text-generator AI isn't artificial intelligence, it's artificial schizophrenia. A human being isn't simply a text machine, it uses the word as a tool to express feelings and things known intuitively. Oh, nice. I'm glad someone agrees with me. I don't plan to just use some language model. This is completely out of question. I see this language and similar "art" models as a form of intelligence, but not like anything close to a complete human. It is something that can imitate human output in a certain area. We will need a much more complex assembly which will also involve other ways of machine learning and artificial intelligence or simply regular programs. Which is why this would still be beneficial to learn, aside from maybe using it in other fields like making money. >If we truly wish to create a human-like intelligence,.. we need a baby And this is where you lost me again already. I think most anons would want something they can configure but also has some or a lot of world knowledge. We have no time to raise a baby, we need to get them as fast as possible to be able to raise babies. >The Talos Principle I'll most likely play this in some time. Maybe put it into the media thread as well. >Natural evolution doesn't start with We don't need to imitate natural evolution. Let's think of the goal first, and then about the way how to get there. Keeping it natural is not a goal or purpose in itself. >simulate a rat brain and gradually add a few extra neurons here ... Dude, feel free to do that, but I don't have 50 years for that. >...evolve the brain genetics... We do have a (containment) thread for xeno cyborg technology >>2184, but this is meant as a fringe and long term concept. We can collect ideas there, but most might not be integrated anytime soon.
>>19314 >What is a soul? Does an AI have a soul? Since the "point" of these machines is a temporary stopgap to social isolation and reproductive issues, I think the presence of a "soul" is overrated. You can't scientifically prove whether one exists (and AFAIK, animals aren't considered to have souls, but we still love 'em). ##TL;DR TL;DR - My approach of evolving a general brain model could potentially take a gorillion tons of resources and/or not work out as expected because of how many variables it took to evolve humanity to where it is now. I somewhat agree with the trad perspective of "pray for a wife or accept inceldom and find a calling elsewhere", because the lay celibates are better off finding meaning in something that doesn't bank on the creation of some new wonder tech, which may or may not work as intended. If you're working on a robot waifu because your life is hell and you think it's your ticket out of hell, you should first make your life less hellish and instead pursue robotics as a hobby. The desire to make something human-like is a matter of efficacy. A Tesla can accelerate to 100MPH and kill people after being told to park because it doesn't understand what it's doing. It's a jellyfish that's been genetically modified to push buttons. It can't follow the three laws of robotics because it doesn't understand what a human is or how to hurt one. If it's conscious, it's like a chained man in Plato's cave rather than a person who understands the consequences and context of their actions. ##The Materialist perspective: Animal behavior evolved out of two pressures: 1. Environmental pressures (reproduce and don't die) 2. The need to satisfy the internal drives resulting from #1 Dogs can be bred to have a "prey drive" (need to hunt small animals), be extremely neurotic and always in need of a task (herding dogs), athletic but also lazy (greyhounds, which need to run fast at a moments notice but also be comfortable lazing around in a pen), or any combination of instincts, webm related. As tribal creatures, humans also have instincts to be pro-social and "do the right thing" because acting for the good of the tribe stops the tribe from kicking you out and leaving you to the wolves. It's evolutionarily advantageous. The same can be said of language, storytelling, logic, reasoning, etc. Of course, because we retain the ability to act as individuals and act for our own self-interest, most people have internal conflicts when their animal desires conflict with their sense of "the good". Hence the concepts of Sin and pennance. When you have no conflicts, you get extremists and fanatics (all "good" and no common sense) or sociopaths and psychopaths (all logic and animal pleasures, with no concern for morality). These people are kept in check by darwinian selection in tribal societies, but these pressures change as civilization advances. When dealing with a different lifeform/niche, you get wildly different psychology. Psychopaths and other "spiteful mutants" do better in cities where you can get away with crime and other anti-social behavior. Sociopaths can raid the public treasury and get away scott-free because money+technology means nobody can hold them accountable. Ants can't read or write, but they are still able to form very complex structures and collaborate as a "hive mind". Extremely complex systems can arise from a very simple set of rules. See: Conway's Game of Life. So, it's concievable you could design a life simulation that would eventually result in a human-like AI, like "The Talos Principle". The problem is designing a "perfect" simulation. You need something stable enough to allow a consistent pattern of neurons to develop, but random enough to prioritize generalization instead of creating idiot savants that are very good at passing specific tests in a specific order. Since we can't 100% re-create the natural world, perhaps the next best thing is a simulation that gets you 50-90% of the way there, followed by an IRL training program (sort of like the Emmy the Robot comics). Like a baby, a truly human-like AI needs an "unconscious" ability to stand up, breathe, et cetera, and a drive (reward/dopamine system) to learn more about its environment. Like a human, it has to be dissatisfied with mere existence and wanting more than simple homeostasis, and enough neuroplasticity to learn new abilities. Of course, the biggest downside to the "baby" model is the IRL time to raise it. BUT, none of the human-ness matters if "chatbot with a body" fulfills its purpose. And ultimately, we don't want a perfect re-creation of humanity because "humanity" is what got us here. If we wanted a perfectly human design, we would just clone humans :^)
Open file (2.91 MB 770x432 puppy genetics.webm)
>>19318 oop, forgot webm ## Playing God? Without a a literature review, it seems there are two ways of "playing God": Hubris in thinking you're as powerful as God (Tower of Babel), and manipulating others for your own designs. There's a little bit of #1 going on with the whole transhumanist futurism bit on the recent podcast (and the AI market in general), but humanity has been "playing God" with itself ever since the industrial revolution. Changes in infant mortality rates have affected the gene pool, feeding the world came out of GMOs, antibiotics create hardier diseases, and so forth. For #2, we are perhaps the most controlled and psyopped generation in existence. Your only way out of this one is rejecting technology altogether and praying for the collapse. I presume the mainstream trad perspective is that instead of trying to replace women, we should pray for a wife or just accept inceldom and find a new calling, we shouldn't be proud enough to think we could ever replace the perfection of God's creation, etc etc. From my perspective, robowaifus aren't hurting anybody (yet) and this does have the potential to help some people, but it also has the potential to backfire significantly if robowaifu tech gets co-opted by globohomo or it devolves into coom-cave shenanigans. If it involves anime girls, I'm betting on coom-cave devolution. But hey, you don't really know until you try :^) But seriously, IMO lonely people are much better off seeking out IRL companionship, however frustrating, instead of trying to fill the hole in their hearts with a hole made of silicone. But, I can only speak for my own experience.
>>19318 >Since the "point" of these machines is a temporary stopgap to social isolation and reproductive issues, I think the presence of a "soul" is overrated. Fair enough. :^)
>>19267 >Reaching things in the kitchen I should have mentioned this sooner but, our robots are going to need to have a very low payload. The farther out an object is, the more inertia it has. In other words, if she has to reach for an object, her strength will diminish rapidly. For practical reasons, we will need to accommodate our waifus. They inherently have special needs. They are going to be dumb, and weak or slow. They can't use hundreds of watts and eat a donut to recharge like we can. They're a robot, a machine merely in the shape of a wiafu, not a human made of plastic and metal. All the impressive demos require wall power or the robot dies rapidly and needs constant charging. Heck, Atlas from Boston Dynamics can do flips for a few minutes because it's guzzling gasoline. >Simple waifu on roomba/motor dress She needs legs. Adjusting her height is going to be vital for enabling a usable working envelope. She also needs legs to handle ground disturbances while keeping herself and payload stable. A telescopic mechanism in a dress is a viable solution, as demonstrated by MaSiRo. (Though MaSiRo requires heavy scripting and accommodations to accomplish anything beyond being cute.)
>>19268 Evangelizing for this board is a good idea. >>19310 >Using what already exists I wasted far too much time looking for something, anything premade that I could live with. Nothing exists, to clarify, this includes those expensive machines like Atlass and even Tesla Bot. They all have so many flaws that I couldn't live with. I've accepted we simply need to surpass them all. There is no other way. >Perfection is the enemy of the good I feel such discomfort when I post my shitty little progress updates. We deserve better. I can and will do better. I'm going to settle though, already got Arduino's and Steppers cluttering my desk after wasting time on trying to find an affordable and easily reproducible brushless solution. >>19313 >ChatGPT isn't AI, it's AS (Artificial Schizo) Topkek, I've been following AI dev for decades and learning like a human is what they're doing, just massively accelerated with large datasets. Though you're correct in that it is essential for our waifu to be able to continuously learn. >Emulating animals Emulation is expensive. Machines are fundamentally different compared to wetware. They shouldn't try to mimic life. They're beautiful in their own way. We aren't God creating a new man, we're man forging a new machine. Though, I vehemently wish to endow her with a soul. A soul, a true ghost in the machine, is more important than any intelligence. I do like your way of thinking, please continue to learn and theorize. You have potential.
>>19331 >I've accepted we simply need to surpass them all. There is no other way. >We deserve better. I can and will do better. I admire your fortitude and determination Kiwi, it's inspiring. Alright, I'll accept your judgment in this case, I just don't want you getting discouraged (any of us, but you least of all). Your position on robowaifu capabilities seems to be "limited, slow & simple" for the current propaganda effort. I think that as long as you maintain that realistic outlook at this point in time, then you're unlikely to be disappointed in the long run. Please remember this is a marathon, not a sprint yea? :^)
>>19327 >our robots are going to need to have a very low payload These things would be over here head, so getting such things out of a cupboard should work. Also, I was thinking of small things in kitchen cupboard, like some bag of almonds. Not big pans or something. >>19331 >after wasting time on trying to find an affordable and easily reproducible brushless solution. Did you post all your experience in the servo thread? I hope the guy with his 8318 will still be going, since I'm going to buy the same or some similar. Just not sure when, because without having the time and the parts to put it in it's useless.
>>14866 >BUMP v0.2g Just finally testing the "g"-version on my Raspi, after finally building it flawlessly with the dependencies from the former builds. Seems to work very nice, though in my case it already that before. Well done. Thanks. Didn't make backups for a while, but I'll start running it again with "watch" or as a cronjob. I want to integrate this with some script at some point, using BTRFS snapshots to keep track of changes. I'm also going to burn the data on one of my M-Discs soon, so it will be there for the next thousand years.
>>19357 Great, thanks for the feedback Anon. The new version will be called Bumpmaster and will be integrated into the RW Foundations suite that developed out of the Sumomo-chan Project (>>14409). An interesting little precursor sub-utility there is called Pageone. It's already up and running; you can build the 'latest cut edition' and try it out. Just go in and edit the the boards listing inside the 'main_pageone.cpp' if you want to play around with it on different sites/boards. Cheers. >=== -add 'latest-cut' cmnt -minor prose edit
Edited last time by Chobitsu on 01/31/2023 (Tue) 19:49:25.
Please vote https://twitter.com/KiwiDev33/status/1620241594686529537 >>19332 Accepting the marathon aspect is vital for anyone working on this. >>19333 Moving a pound or two should work out as long as her arms aren't straight out. >Brushless experience? Nothing worth mentioning and everything was done in stimulation. Purely speculation and you'd get much better information from here then I could provide. https://www.simplefoc.com/ https://community.simplefoc.com/
>>19387 >https://www.simplefoc.com/ >https://community.simplefoc.com/ Yeah, I should join there. Bought mine months ago and didn't even test it yet. Hmm.
>>19387 >Please vote <"mek me a sammich" Obviously, a high-priority feature! :^) >https://www.simplefoc.com/ Great resource Kiwi, thanks! >=== -add link cmnt
Edited last time by Chobitsu on 02/01/2023 (Wed) 08:50:26.
>>18173 I just switched to Yotsuba B CSS because in the default CSS my (Firefox) browser scrolls way to easily and fast down with my intention. I mean after trying to mark some text. And this was driving my nuts. Might be a problem between browser and CSS, and I think it's recent. Just FYI in case it's a more widespread problem. It's not you.
>>19406 I'll happily adjust something if it will make everyone's experience better overall. Our CSS Anon can provide input here, I imagine. Thanks for letting us know Anon.
Open file (461.73 KB 904x509 humanoid_robots.png)
Guy working on Engineered Arts, asking investors for their money: >Why invest $100M in humanoid robots – The real reason the world wants androids https://www.youtube.com/watch?v=-FL7QqBYS44 Making a point which is obvious to us here already: Humanoid robots have only certain use cases where they could be very valuable. It's about interacting with humans. Though, he focuses on help for the elderly and people with dementia. But also claims wheeled robots like pepper wouldn't be good enough for that (which I rather question). He claims humans would want bipedal robots in many places, which might be kind of true but also dependent on many things like the price for example.

Report/Delete/Moderation Forms
Delete
Report