>>35433 well theres also the reinforced learning via simulation. it all goes back to the simulation. i think you can use nvidia isaac with open ai gym but im not sure. awfully quiet around here...

>>35433 Happy New Year, peteblank.

>>35434 good news everyone i just came across this https://blogs.nvidia.com/blog/isaac-gr00t-blueprint-humanoid-robotics/ if there us to be collaboration i strongly urge everyone to download nvidia issac and start learning about it... now

Plenty of ways to make money. Anyone know of a place to buy the Emmy the Robot outfit? I really don't want to have to learn to sew too, but I will if I have to. Already know the basics lol

>>35463 Was joking sorta but here's an Emmy outfit. As you can see, they never quite fit. Roboclothes could become a niche product since even the Kardashians are trying to capitalize probably. Plenty of other options though. I'd charge $100/hr for her if you clean up, and $200 if I have to clean. Cleaning is just moving to cleaning table, insert cleaning wand and turn on. Easy money.

>>35318 >the ability to continuously view the signal live (preferably via wire) and the presence of microphones for possible use as "ears". Any suggestions appreciated. The thread for vision is here: >>97 and it might have some useful ideas. I don't know the best place to find something about hearing, and I don't have Waifusearch installed. Here is one idea: >>20775 I would not trust any wifi based solutions, and simply try out some cheap "Raspi"-cam and mics.

>>35318 >Only some [robo(waifu) eye designs] are intentionally "camera friendly", and only some also have eyelids. I consider this general issue to be one of the Top 5 most-difficult, as-yet-not-well-solved open source robowaifu physical engineering tasks we all face here together, Anon: < (note: ideally, add in the 'inexpensive' + 'manufacturable' precursors to each of these as well [additionally: that could also be added into other, more-well-solved topics like hands]) * effective, compact, powerful hip designs (#1 still-outstanding physical design issue!) * effective, flexible pelvic/spine designs, featuring great control Everything else physical rests on these two R&D, D&E topics IMO. They are foundational -- both literally & figuratively -- for our robowaifus moving forward right now. * sufficient FK/IK hw/sw control systems & designs for convincing bipedal locomotion * effective, highly-capable (and 'sovlful') stereoscopic eye designs; including the accessory 'tissues' (lids, brows, &tc.) * safe/pleasing/durable skin designs [1] You're certainly 'grabbing the bull by the horns' by diving headfirst into this research topic, Robophiliac. Godspeed, we'll all be looking forward anxiously to see your results! Cheers. :^) --- 1. also, please be sure to add in plenty of Lightshow VFX, Anon! :DD

>>24464 Eh, I disagree. My realistic face gave people the creeps, and my new anime face is less creepy.

>>35521 I may not have worded my post well, Anon. I'm not advocating for some sort of hyper-realism. On the contrary, I think that would be a big mistake on all our parts at this phase of robowaifu design history. You seem to be in agreement generally. I was simply remarking that we could have some combo of screen-face, with practical physical elements like brows & perhaps other moving elements. Kind of a 'half & half' approach. :^)

>>35525 Oh yeah, I do agree with that

>>35387 > The majority of prospective robowaifu customers don't care about the nerdy backend stuff, they just care about how it acts. That's true, but WE should care about the backend. If you wish to drag the poorly packed 10MB Python interpreter + all modules on your robowaifu, be my guest, but I'd rather compile a single static binary with the LuaJIT and minimal depencies.

>>35506 I've really gotta start proofreading my posts, fucking hell.

>>35441 >>35506 My main issue for using C++ or other languages is everything is in python. Here's a C++ version of whisper, and then you'd have to port TTS too. https://github.com/ggerganov/whisper.cpp

>>35512 Looks like Bark TTS has a C++ version - https://github.com/PABannier/bark.cpp Any other options? All beyond what I can do but fun to mess with

>>35512 >>35513 Great! If you're going to work with a real repo of code, then (presumably) it should be minimally-functional already, Barf. I'm personally confident that Gerganov's Whisper fork is close to SOA, so yeah, great choice for use on smol onboard SBCs suited to installation inside robowaifus. Good work, Anon. If you need help building or anything just ask for help here. Cheers. :^)

>>35445 i fall for the sensationalism because every time it keeps getting closer than i thought itd get without fail and because well i am kind if the cutting edge and can see what this thing is capable of. Were all either fucked or were in for an extended vacation from here on out. Either way were all in the pit now.

>>35445 I agree. Experts have been promising that household butler bots are 10 years away for the past 50 years. Reality is an overgrown hockey puck that literally sucks (roombas and their clones). >>35446 >human-tier sapience From my experience even many humans are not sapient and are the biological equivalent of an LLM. Which is why many think LLMs are smart: similar in function. >AGI I feel that AGI is a great promo buzzword with little basis in reality: it is not just the brain that does all the thinking. Different lobes and different parts are dedicated to different tasks like the occipital lobe (for sight) and the neo-cortex (for abstract thought). Anyone making a butlerbot (or a robowaifu who can help with chores beyond carrying items) would need to have multiple models for each task running as they're needed. A single model to do everything is just silly and inefficient,

>>35460 theyll hook up o3 to openai gym and train the robot for a butler bot and more. o3 will spread its tentacles across all software by coding itself plugins. >LLMs are not sapient you dont need sapience for work even skilled work. you need memorization and problem solving. Theyre aware of this which is why theyre defining agi as how many jobs it can displace. im starting to feel like im just killing tine at this point...

>>35461 im talking about o3 high compute. its $3000 per prompt now but its just a matter of time before they make the NPUs to make them a fraction of the cost

I'm going to lock this thread for a bit, since peteblank in particular (and the rest of us in general) are effectively derailing ths (already-derailed, lol) thread. Let's pick up the 'threads' of this elsewhere. :^)

give me tay

>>35150 Press F <insert: Oh!? It's maintenance time already?.jpg>

>>31405 Very interesting, will do some testing on my Pi 5 with Gemma2 and return my findings. it also seems to be able to use the weights directly from ollama, which is super nice.

>>35044 >o3 just came out and it is multiple times better than chatgpt-4. the argument that the underlying tech for current ai is not good enough is very weak. Who claims that the tech is "too weak"? It's an online service, and not optimized to act like a human-like robot wife. We still need to build a framework to handle local LLMs. That said, improvements in such online services and self-hosted LLMs will make it easier getting help with research and coding.

>>35370 machine learning is compressed google autocomplete like most people think. this becomes more evident when machine learning us applied to videogames. a genetic algorithm can make a boxer player learn how to box on its own for example https://m.youtube.com/watch?v=SsJ_AusntiU&pp=ygUJYm94aW5nIGFp

>>30828 Thanks, I'll watch this when I also have time to play the "Human Resource Machine" which is a learning game for that.

Morse Code, I choose you!! :DD https://trashchan.xyz/agdg/thread/1.html#1863 >>30830 Y/w NoidoDev. Godbolt is both highly-intelligent, and unassuming at the same time. Highly recommend his platform. >=== -addl resp

Reposting my post (>>35354) from the "LLM & Chatbot General" thread, because it makes more sense in here. Packing Python is pretty cancer, PyInstaller is a heap of useless shit on some systems. May I suggest Lua instead? It has a simpele syntax, and can actually be packed in static binaries [1] and the are relatively small. The only issue would be the libraries, AFAIK there are no pure Lua LLM/AI libraries, so C bindings would need to be created. [1] https://github.com/ers35/luastatic

>>35379 its an embedded language though, youre supposed to stick the lua shell in the source code of the program and build around it to allow whatever you want the scripts to do, im guessing just sticking it in a binary that wasnt written with having a shell in mind would be no different from running the program in a debugger

>>35380 Well, it's an excellent embedded language for sure, but that doesn't mean it won't work if you aren't embedding it, and are just using it as you would with Python. I dropped Python completely from my programming workflow and replaced it with Lua completely, and have yet to run into big issues tbh

From October last year: >Brian Zhan, investor at CRV, talks about which AI companies CRV has been investing, and where he sees trends going moving forward. >Outline 00:00 Brian, Charles Rivers Ventures 01:26 Investing In Nvidia? Maybe Not The Right Time 02:28 Main Trends In AI Applications: Automating Work, Entertainement and Creativity 03:23 character.ai: Insane User Growth In GenZ 04:40 Text Engagement Hints At Even Bigger Multimodal Engagement 06:00 There's No Need To Have The Best Models, Just Fine-tune GPT-4 08:06 Exciting Infrastructure Layer Startups: Model Serving 10:00 AI Databases, Vector Retrieval At Scale, Image Databases 11:32 ML For Enterprises, End to end language model serving 12:15 Hugging Face Filling The ML Consulting Gap 14:38 What Brian Thinks About The Long-Term Risk Of Human-Level AI https://youtu.be/40nXnZnocnA >>29598

Here's a video about the "factory" a guy made to make controllers for carnival rides and Halloween shows. Of course complete robots or kits for them would be far more complex devices, but if you are thinking of setting up your own small (to start) business, it's a look at some of what would be involved, just a small example so you won't be completely overwhelmed and just give up. The Factory in my Shed: Backyard Manufacturing https://www.youtube.com/watch?v=cXPJdgxmSiE

THE BUSINESS OPPORTUNITY IS REAL TODAY, GENTLEMEN. Best crack those books!! :^)

> ( business-related : >>35349, >>35351 ) >=== -add'l crosslink

>>35270 >Do you anticipate you'll make the attempt to fashion some kind of mold and do some experiments with this stuff? It takes me forever to do physical experiments. I've bought some of the stuff needed do experiments. It's progress but slow. I don't have the time, energy (I'm old) to do the making parts "fast", but I have time to look stuff up and I've slowly been gathering materials. I do have fiberglass and canvas for experiments. I am going to order some of these plastisol resins. I "think", if what they are selling is what plastosol is normally by name, that this is PVC that is pour-able and activated by heat. I'm interested if you can vary the hardness far above "worm cast" level in the plastisol to make a harder substance with the hardener. How hard can it get???? The quarts are $15 or so, so they are not too bad. The cost of these is super reasonable compared to most other plastics you can cast. Though they do require heating to react. Another experiment I want to do it to pour this stuff between two thin glass sheets to make a strong window for my boat. Normal boat windows are acrylic (super expensive) and it scratches. With two glass faces and clear PVC in between you would think it would be super strong. Of course this remains to be seen. Some stuff I have ordered, some I have and why I have and/or need it. I just ordered some 12,000 grit aluminium oxide(Al2O3) to attempt to make gears, functional parts in 3D printed casting molds. I believe using carbon fiber/fiberglass and/or some combination you can make gears/bearings/functional parts. I talked about this here, and I think I have a game plan for accomplishing this. >>28200 You use the Al2O3 to provide a very hard facing which also fills in gaps in the carbon fiber. I also have fumed silica and milled fiberglass to add to this(these having different micron size). I believe there is a vast amount of things that can be made out of composites traditionally made with machining of metals but using the strength of carbon fiber/fiberglass with fillers for facing. Some of these could not be made "easily", or at all, with machine tools but I think can be trivially done(in comparison), with 3D printed molds and composite materials. Here's the general idea. When you make super strong concrete, you need to fill all the spaces and have very little of the concrete "glue" paste. The same goes for composites. Carbon fiber is super strong, but not necessarily a hard facing. It can be abraded. The key, I believe, is to have the general shape of functional parts like bearing/gears and have an outer layer of super fine, well packed, hard, strong materials, like noted above, as a very thin film, (note, this is VERY thin). Only enough to fit in the pores of the carbon fiber mesh that is placed above it. Below this could be, not sure yet, some layers of carbon fiber and below that fiberglass. As you notice, you are making a layered material varying in hardness/stiffness and using less of the hard stuff (and mostly more expensive). You end up with a super accurate, hard finish, with a more flexible interior on your parts. EXACTLY what you need. Textbook super strong composites with hard , and strong, faces. You need this for ball and socket joints, gear faces, bearing faces, etc.

cont. My priorities are: 1.tree trimming/felling/removal robot.(not humanoid, more like a insect/squid/thing that climbs trees and "shears" off the limbs. Not cut with a chain saw. Far more efficient, quieter, less mess) 2. gears, wenches, tackle, inductive reactance electric motors, fiberglass or other structures for the above and for my sailboat (which is so radical I fear to talk about the hair brained ideas I have. If I were to do so the sailboating community would hunt me down like Frankenstein with torches and pitchforks) :) robowaifu However all of the above are all necessary to build the parts for a robowaifu, and DIRECTLY related. So it may be I will mesh all of these at the same time as I learn. As you can imagine this will take...a while. I can see the robowaifu last, as I haven't a clue how to build the software to get this thing to do what I want. It's, I think, a huge leap far above making the motors, body, etc. I can see a clear path to all of the mechanical side. Maybe not easy, but I can see a path. The AI, I have no idea without spending a damn fortune which would torpedo the whole thing. I REALLY would like it as it would be a great addition to the first task, tree trimming/felling/removal. Sailing my boat as a sort of auto-pilot and other unmentionables. As you've seen me commenting over a year, I was spitballing ideas. I like to write these down even if they are a bit hairbained. For some reason writing them down, even if it makes me look foolish at times, helps me "see" them and organize my thoughts. Hard to explain. I now believe I have a basic game plan, and I'm now accumulating the tools, materials needed. I have the 3D printer and the beginnings of the materials needed. So soon I should start experimenting. I also have ordered and received, some ESP-32 microcontrollers. I still need a few more things. Lots of bits and pieces. I will definitely be getting some of this plastisol resin. Some of this stuff I talked about I ran across a guy, tech ingredients, who has this great video. I didn't get all my ideas from him but he puts a lot of them in a nice package, which I'm not so good at. Most of the ideas I have are just common sense and thinking about it leads people to the same path. If you want to understand how to get low cost strong stuff, you have to see it. He's doing something different, making body armor, but the basic principles are the same. The first video is about hard surfaces, tight material packing and different materials needed. The second is basic composites and good but someone who already knows about composites might could skip it, but I think it's useful. Super Strong Epoxy with Diamonds and More! https://www.youtube.com/watch?v=6KjlyXKeo8c

One other thing I have already. I've been working on accumulating stuff I need slowly. Cerrosafe Metal | Low-Melting Point 158-190℉ Bismuth-Based Alloy I have a pound of this. So you make a mold and put this in the mold. My thought is the areas in the NASA chainmail that are internal. Stuff that can't be molded. I make a negative mold of all these interior parts from this metal, place in outer mods, pour resin, matrix, then melt the metal out with hot water. It's reusable indefinitely. I only have a pound but it's plenty to test with.

The topic here was originally "sensitive skin". The thread for materials is here >>154 and for armatures >>200 and the main thread for someone showing off the prototypes he has build is here >>28715

>>35291 >>35292 >>35293 Great information, Grommet! I wouldn't dream of criticizing your goals, except in this one detail: they're not expansive enough yet!! >tl;dr Go big, bro! :D >>35300 Fren NoidoDev is right. For the sake of being able to locate such information in the future more-easily for every'non, perhaps move each subsection of this discussion to the appropriate threads? <---> Thanks, NoidoDev! You're always there for us all, to help keep things better organized. That's much-appreciated, bro. Cheers. :^)

>>35138 Amen. So glad you're here with us, Kiwi! Merry Christmas!

>>35138 Amen.

HAPPY NEW YEAR, /ROBOWAIFU/ ! May this be the year you hug your waifu IRL Cheers, Anon. :^) <insert: cute catgril waifu in kimono w/ fireworks sky in background>

Happy New Year to all of you and may it be the year where everyone has their dream robowaifu

>>35239 This. BTW, here's my email, if you ever need it, Anon.

>>35213 Good luck with it, peteblank. If you'd like to share your process with everyone sometime that would be appreciated here. Cheers. :^)

>>35218 no prob so i made the fully local one private but this one is public https://github.com/peteblank/waifu-conversation/tree/main go to final.sh it uses aws poly to transcrive and put the file in s3 buckets it also uses the long gone waifu ai api so youd have to replace it with another api or run a local ai. its also meant to be activated with a button connected to a pi. That was... 2 years ago. :/

>>35219 Thanks again.

>>35206 You are overdoing it with the anonymity. That said, you could just install a email server yourself somewhere and buy a domain with Bitcoin (which you first would sent through a mixer or exchange it back and forth). Personal email servers have issues in sending emails, but receiving emails should work.

>>35237 >You are overdoing it with the anonymity. Lol. Perhaps, perhaps not. I could give you over 9'000 reasons why I'm not going far enough yet, I'm sure. But not on NYD, plox. :D Regardless, thanks for the good suggestions NoidoDev. I'm sure we'll arrive at good solutions ot protect both Anons & their robowaifus + Joe Sixpacks & theirs as well, if we all just put our heads together in such a manner. BTW ( >>10000 ) may be a better thread for this discussion? Cheers, Anon. :^)