Neurobot-1000 3d model project

[opal bane]

Might also be it's old

[floral quarry]

ight, noted

[opal bane]

Small child goes up to gas pump, sticks it in her mouth like a straw while a clerk inside looks on in fascinated horror

[opaque quartz]

This stuff?
https://www.youtube.com/shorts/B4HVqLufugs

[floral quarry]

the one thing i am currently confused about is how cooling will actually be done

[opal bane]

Metal hose

[floral quarry]

metal hose go where

what collant do it use

[opal bane]

Currently looking into water cooling and having vessels going throughout the body

[floral quarry]

how do it cool the gigantic 4090

[opal bane]

We're not gonna fit neuro's pc in there she'll be in it remotely

Maybe in 10 years

[floral quarry]

no dawg remote will not cut it

the latency will be insane

there are enough embedded platforms to run this stuff properly

[opaque quartz]

there are types of coolant that are non conductive and non corrosive

and non toxic

usually oil-based

[opal bane]

@ruby lagoon talked about this at some point something something ad hoc connection

[opaque quartz]

but high thermal conductivity

[floral quarry]

and i mean stuff by nvidia that is the size of a cpu and is 100 core

[opaque quartz]

oh damn

might need that computor brain after all

[floral quarry]

dang thats gonna be terrible for our use case

what if we do something along the lines of ghost in the shell/serial experiments lain and hook up the body with long gigabit cables

for v1

[opal bane]

We have a 4 foot body idk how we can fit that in

We need to not have external cables

[floral quarry]

and then worry about fitting the brain in v2

[floral quarry]

then the brain goes in

because the latency, trust me, will be huge

[opaque quartz]

yeah that's the plan for v1

using cable to reduce latency would be good

but that'd mean no non-cable movement

ig for v1 that's ok

but what do we do when we move to engine?

[floral quarry]

ok stuffing the 4090 shelved for v2

worry about movement and the inverse kinematic algo that will have to be written and trained for v1

[opaque quartz]

ye that'll be tough

[floral quarry]

and the face

[opaque quartz]

i'll focus on the face, it's what i'm most suited to

as for algo, i think simleak had some ideas

[opal bane]

here

[opaque quartz]

btw nice name change jpdawg

[opal bane]

Ye

[floral quarry]

the pi does NOT have gigabit wifi, only gigabit ethernet that with addons

[floral quarry]

https://youtu.be/yWrldOS6xBw?si=uV3vmGLPSKSD8rA8

[opal bane]

Why js it screaming

[opaque quartz]

[floral quarry]

"please help us"

[floral quarry]

and with a model of neuro's size, we will need a lot of raw power to transmit data
so gigabit is bare minimum

[opaque quartz]

if we can figure out a non-creepy face, we can easily crack fusion afterwards

[fossil drum]

wait doesnt india have nuclear reactors?

and even weapons?

anyways what part of the body are we working on rn

[floral quarry]

basically it is "come work with us or we will jail you"

[opaque quartz]

current parts left to conceptualize:

Arms
face
spine

arms are partly done

ish

[floral quarry]

there is even a specialised institute for this stuff to train for it and the work, after the guy who made our nukes and reactors

[opaque quartz]

need a redesign imo

[fossil drum]

we doing well then

[fossil drum]

then try to get national support for the neurobot project

here

all done

[opal bane]

How big even is a 4090

[fossil drum]

lmao

[fossil drum]

like 1.5 the size of 3090

if u mean the rtx

[opal bane]

That tells me nothing I don't know how big a 3090 is

[fossil drum]

okay gimme a moment

304 mm x 137 mm x 61 mm

[opaque quartz]

gonna have to mimimize braincase stuff then imo, if we need a computor in there later

smallest possible eyes and motors

[fossil drum]

although u guys do know that 5090 is already a thing right?

[opaque quartz]

maybe mimic neck and use only that for head motors

[opal bane]

Who says we need the brain to be in the head?

[night tartan]

Out of stock everywhere

[opaque quartz]

not really enough free space in the torso

[fossil drum]

not surprised

[opaque quartz]

head isn't used for much other than eyes

[fossil drum]

it released like last month

[opal bane]

Ah the laws of physics says ok

[floral quarry]

for the spine we could do sections of muslce fibre for each divided section of the vertebral disc (not each disc - that way we will have too many motors to run, instead one fibre for 4-5 discs)

and then control those sections of fibre using a custom algo which will have to be written

[opaque quartz]

luckily there are very few muscles truly operating the deep back muscles, we can mimic those with a single large motor each, then use pulleys fastened to where the human muscles fasten to the spine

we want to minimize motor number imo

only using so many in forearm and face due to the precision needed

[night tartan]

It's selling on ebay for like 5K euros

[floral quarry]

we'll make the fibres so that there is space in the centre, leaving way for the main heavy cables to be pulled through

[fossil drum]

ffs

that much?

thats over twice the price

[floral quarry]

because the spine has that job too

[opaque quartz]

oh nice idea

using the spine for 'nerves'

[floral quarry]

dawg my pc is so potato the discord messages are sent by me 5 seconds later

[opaque quartz]

that said the inner side might be best protected right?

probably good to have em central

[fossil drum]

im surprised u didnt think of it. i mean, the spine is the only thing that goes from bottom of the torso to the neck and would have enough space inside to hold the cables

[floral quarry]

that's why we do fibres around it

[opaque quartz]

true but there's a lot of complexity there

i was gonna go with direct vertical wires on inside

might still be best option

[fossil drum]

thats just

a very bad idea i believe

[floral quarry]

shock absorption for fragile cables that might eventaully wind up being optic fibres later "because they transmit data the best"

[opaque quartz]

not inside of spine

[opaque quartz]

based

[fossil drum]

wouldnt it make the insides of her body like a freaking spaghetti?

[opaque quartz]

ig ye

depends how we do it

could have a central cable that is like a fractal

we'll need to discuss it

[floral quarry]

brother the human body is absolute jelly on the inside why are you suprised

we're defended by a ribcage

[fossil drum]

only lungs and heart really

[opal bane]

So it looks like the main thing we need now is a good coder

[opaque quartz]

😂

[opal bane]

On the personel side

[floral quarry]

we go next step and make her bulletproof

[fossil drum]

hmmm.... we could always ask vedal for any help

assuming

of course

we go anywhere far enough

[opal bane]

Can we trust copilot tho

[floral quarry]

to that end, i found something highly promising
https://youtu.be/G6AaAU4Sv5w?si=-oA_7rjFD2gONq_h

[opal bane]

In seriousness vedal has a lot to do already

[floral quarry]

copilot sucks, we need a good guy
i can help but only mildly after march

[opal bane]

We currently have plans for metal skin

[opaque quartz]

@floral quarry bulk metalic glass skele for v2 hopefully

[floral quarry]

and will basically be missing half of may

[opaque quartz]

• metal plating

fake skin is currently a bit too loose and uncanny

actuation + true to skin dimensions should sell the illusion

i'll test it in blender

[floral quarry]

sillicone cast it and throw it over the body

will look like the real deal

[opaque quartz]

terminator moment

[opal bane]

That'll suck to repair though

[opaque quartz]

i don't agree, silicone skin is often very slightly off

[fossil drum]

neuro from the future coming to the past to subjugate humanity

[opaque quartz]

better to use an analogue that is more abstract until we can get it perfect with skin

[floral quarry]

metal???????? the cnc cost will be abysmal

you're looking anywhere from half to three fourths of budgets diving in that

+do not do glass we dont want more fragile than already

[opaque quartz]

quartz glass might be ok durability wise

ngl that is an avenue i didn't consider, could be promising

[floral quarry]

just do clear plastic lasered

[opaque quartz]

true

[ruby lagoon]

I did ad hoc robot control with robot_network. It was good and real time. However it has to be ad hoc. Other wireless methods won't work. Resolution also has to be relatively low. 1080p won't work, maybe 640x480.
Tbh, that also may only work with a wheeled version, and a legged version would probably need on robot inverse kinematics.

[floral quarry]

metal i can say we can shelve to v3

[floral quarry]

we will work ik rigging inside body with a pi or lattepanda if we do ad hoc

send direct commands by neuro to move leg and controller calculates movement

[opaque quartz]

definitely inverse kinematics

[opaque quartz]

we discussed that but were worried about latency

possible solution was small computor onboard to handle fine movements, with instructions from neuro

persoanlly i'd prefer precise movements

[floral quarry]

that's what i said

[opaque quartz]

if we can solve latency either through big enough onboard computor or through cable, or connection, fine movements is better

[floral quarry]

neuro just says "move foot" and pi does the algo

[ruby lagoon]

Ah, I was planning on working on a "cerebellum". Some transformer that could take in rare input and output 60fps movement and commands.

[opaque quartz]

huh i misread

thought u meant move muscle on leg

ye i agree then

[floral quarry]

what i want to ask is
is the latency reasonable and within bounds for a neural network of her capacity

like under 50ms

or max hard top 100ms

[ruby lagoon]

For a raspberry pi? I honestly don't know. I might just have to test it and see what compute is needed.
On the bright side I guess I don't need to port my code to GPU yet if we're using a pi.

[floral quarry]

and other than that will the incoming commands have enough "pipe width", so to speak for the pi

see i would dump all work and test it immediately was it not for the fact i have a pi 3b

and 4 of them

[ruby lagoon]

I have some pi 4s and 3s, no 5s though

[floral quarry]

somebody will have to sacrifice their wallet for a pi 5 eventually

who shall it be

[opaque quartz]

can clear plastic be molded on the cheap or are we working with flat plates?

and if yes to the former, does it become discolored if it's DIY?

[floral quarry]

there is both

actually, scratch option 1

[opaque quartz]

flat it is?

[floral quarry]

https://www.uvplastic.com/blog/how-to-mold-acrylic-sheets.html

nah we cool

but the thing is it will be a difficult effort

[opaque quartz]

yeah sounds like it

[floral quarry]

there is custom acrylic machining however

i do not know the prices though

[opaque quartz]

i imagine not too much cheaper than metal

if it's a precision mold

for v1 we can probably cope by forming complex shapes from pre-cut flat acrylic

[floral quarry]

no, apparently you can heat and bend plexiglass

however it's at 160 celsius

[opaque quartz]

i vaguelly remember there being ways to DIY bend them without much discoloration

for v1, some visual jank should be fine

[floral quarry]

done with heat gun and a oven
you will need a mold for it however

[opaque quartz]

💯

right imma go make dinner

some more blender work when i get back

[floral quarry]

if you use sheets then we will need a shitton of methylene chloride or acrylic cement (i forgot what exactly is need for plexiglass to glue it)

anyways, i shall sleep now, will wake up in morning and give my two cents

[opal bane]

Successful recruitment

[opaque quartz]

gn joker

[queen field]

Yes.

[fossil drum]

yes

[spare vector]

I know a little about mold design

Did a class involving it one time

[ruby lagoon]

Working on making the planetary gearbox GUI a little... a lot easier to understand:

top-right GUI: made a diagram class that uses A* search to connect lines
bottom-right: added SVG displaying that I can draw the diagram lines over (also exported example gearbox from FreeCAD)

[spare vector]

Nice

[opaque quartz]

awesome work

[ruby lagoon]

Thx!
Looks like this now btw. Turns out Inkscape had some good stuff for lines and using it seemed a bit easier:

I think the colors could be better, but it's an svg, so anyone should be able to edit the colors

[ruby lagoon]

Works! Though I need to update the image source for python to figure it out:

[solid schooner]

Gentleman.
I just saw this.
we are building this.
We'll not be taking anymore questions.

#starboard message

[fossil drum]

so cuteee

[opaque quartz]

that gear shape is really cool

[opaque quartz]

excellent reference there from paccha

[solid schooner]

I just noticed the uwu nuke..
theres no stopping now

[opal bane]

He just leveled up his neuro addiction

Never seen it happen in person

[queen field]

This actually does raise some valid suggestions and questions:
Should Neuro have cat ears for the robobody, and should they be optional and just for aesthetics or should they be integrated with functionality and be able to fold away when otherwise not used; how would we make a tail if we add one, what materials, internally and externally, how should it be constructed, will it be just a limp peice of fabric or an animated mechanical faux- prehensile tail, fully prehensile, should it help with balance rather than be purely for the fluff visuals?

So many questions, so much feature creep...

[opal bane]

I had the idea for deployable cat ear antenae, and a tail for charging port

[opal bane]

Otherwise the tail would just be a rope essentially

Unless..?

[floral quarry]

servo based folding cat ears

rotate out

or with a very light dc motor would also work

v1 should have it just there

what I want to know is how will tail be handled

As for how ear will be built, slap fur over 3d printed ear with a hinge

I'll send a drawing of what I have in mind later

[opal bane]

You know those vacuum cords that go into the vacuum

[floral quarry]

Yeah but will it just be stiff or actually moveable by her

[modest quartz]

idk how i ended up here but it looks badass

[fossil drum]

Guys

We can make

The cat ears work like those lights in Japanese cars

[snow creek]

Please someone tell me Ellie is aware of this project. This looks awesome!

[floral quarry]

problem: how do we tell her

[floral quarry]

oh hi paccha

[snow creek]

I gotchu fam

[floral quarry]

how

[snow creek]

imma post it in her discord server somwhere where her moderators can see it

[floral quarry]

do it in a way that doesn't let the swarm loose

those guys catch wind of literally anything and go berserk for the lulz

[snow creek]

oh tru....

[floral quarry]

dang that works

please pass on the word

she would be real helpful

[snow creek]

oh dont get me wrong- I don't expect her to help. She does have a full time job AND streaming. I was moreso interested on letting her know cause she would love to know about it

[floral quarry]

like the ae86 / rx7/ nsx pop up lights? that's what i was thinking too, but since they are just appnedages, we could make em with just fabric and keep them folded during flexion while prop em up like a tent during extension

[floral quarry]

didn't know about the job part - i live under a mountain like the monkey king so only know random stuff

still, she could point us in the right way

[fossil drum]

yes sir

[opaque quartz]

prehensile tail would be easily done with two servo motors driving four ropes that travel the segmented length of the tail. However v1 is bare minimum, i don't think we need the cat parts yet.

[floral quarry]

face will be done in v1 only right

so we will slap the cat ears here and now

[opaque quartz]

face we need for v1 so that it conveys emotion properly ye

if we can do it with room in the skull i wouldn't be opposed to it. Proper cat ears are also prehensile so we should think about how we'd make them move like those of a cat

we can probably miss out on human ears if we go with cat ears

as you can see they're quite large so it'd actually be a decently large project for the neuro design

i'm not convinced we need them just yet

[opaque quartz]

Spine time

[strange quiver]

I've seen previous concerns about the ability for Neuro to learn to control the robot body, but that's not an issue with a program called Genesis

It is a powerful open-source program simulating life physics

And is used for teaching robots how to move

It can basically teach Neuro in few weeks like thousands of different ways to move

Because it's very very fast

https://github.com/Genesis-Embodied-AI/Genesis

I'm not entirely sure, but I think it was able to teach boston dynamics robot to do backflips in like 30-ish seconds

Or something like that

[strange quiver]

And it wasn't trained on like the A100, H100 or H800 gpu

But a normal RTX 4090

[queen field]

Why does this sound like an ad-read?

[strange quiver]

Cause I'm silly

And don't know how to write sentences

That sound human

When talking about important stuff

Summary: "Very powerful tool that will allow to make the Neuro robot be taught very quickly"

[strange quiver]

Also it would allow Neuro to train herself. Because Genesis allows for creating environments for training. So for example: Neuro would want to go to Vedal and give him a high five. Well then we give Genesis a prompt: robot goes to human and gives him a high five. In about 30 seconds we have Neuro robot doing exactly that.

And we can also save that movement for later.

Of course it would require an AI that would create these Genesis prompts very carefully for safety and precision, but we can think about that later.

[floral quarry]

My solution to the ear problem

@opaque quartz

[opal bane]

Apparently all robotics related paths end as Neurobot

[floral quarry]

@opal bane whaddya think give thoughts

[opal bane]

Uhh

[floral quarry]

@opaque quartz gimme few hours after im done with work i'll cook something for spine

[opal bane]

Solution 2 seems better

Also for the eyes maybe some like small screens? Allowing her to eye sparkle and other eye toggles/animations

[floral quarry]

also should we set up a discord server for this stuff or keep it here

where go camera

[opal bane]

Nah keep it here it's way easier to get people for it

[opal bane]

Good question

I don't know

[floral quarry]

also if we put two camera in both eyes how do we combine feed

[opal bane]

I am NOT an expert on anything in here i just have thoughts on rare occasion

[floral quarry]

then if we put camera (small) into the pupil how do we cook the sparkles

wait we can't push sparkles to v2 as face is for v1

can't postpone problems either

[opal bane]

Uhh contacts?

[floral quarry]

contacts dont change

it will be perma sparkly

[opal bane]

But you could probably do something to slide them into place mid blink or something

[floral quarry]

hard pass, would be like attempting a fulton skyhook

[opal bane]

No clue what that is but maybe sparkles aren't really necessary?

[floral quarry]

lets ask the others

am i allowed to do a here

[opal bane]

Also i have 2 additional concerns
Who's going to have everything for assembly and pay for probably 50kish in parts and have the robot for testing?
How are we getting the finished product to vedal

Since you can make a poll i guess it's allowed

[floral quarry]

i'd rather not it would summon a shitstorm

no@ here then

[floral quarry]

shit genuine question

and if we give it to him what do we get in return for spending our time on this

[opal bane]

I didn't consider reward

[floral quarry]

it won't be 50k in v1

[floral quarry]

like i get it we do all this for our combined autism and swarm behaviour for neuro

but like i got exams that decide my future n shit

[opal bane]

Idk can we convince vedal for a box of greggs

[floral quarry]

matthew and the others have been doing the heavy lifting before i came

what would they get

i mean i'd go to college put it on my nice resume and land a cool job somewhere

what about you folks

let's just worry about it later

POSTPONED TO V2

[opal bane]

That's an interesting choice, on one hand you helped build the most advanced robot I've ever seen (not hard don't see many robots around here) on the other you watch vtubers

[opal bane]

Life deciding exams> silly ai body (unless you have a engineering class then you may be able to integrate it in there)

[ruby lagoon]

I potentially get several prototype robots I build (the ones before the robots look like Neuro), if the servo stuff works right then 3d printing/cnc milling scripts for 10-100 times cheaper servos and specialized robotic actuators like SEAs, CAD systems that can quickly make robots, better tested battery/energy systems, and AI and systems that can handle all of that.
Vedal in turn gets a neurobot

[floral quarry]

*** SimLeek is typing... (for the past 10 minuteslol)

o hay u finaly spoke

[ruby lagoon]

I was thinking about what we get...

[floral quarry]

so in short this is basically a way for us to prototype with random people together

with whom we could potentially collaborate on a larger way

vedal gets his bot, and we get a working prototype which we can refine and then market as a better product

[ruby lagoon]

Pretty much. If we can make neurobot, and we work to make our work easier along the way, we can make more robots more easily

[floral quarry]

sounds cool

if this goes well (well would be defined as 1 or 2 years) then we can all collab on a single refined model instead of all of us 3d printing parts out seperately

[opal bane]

We'd probably have to get vedal to do a fundraiser thing like he did for neurodog

[floral quarry]

dawg neurodog could be done under 1000 ez

[ruby lagoon]

True. That'd be pretty fun

[floral quarry]

this stuff might need 10k bare min for the refined v2

how would you raise that much capital... nevermind i just remembered the hype train record

[ruby lagoon]

Really depends on how much we can reduce costs, which is hard to determine, but I feel like there are a lot of ways

[floral quarry]

we found our investor: the swarm

we must convince its leader (vedal) with a working v1 or atleast v0.5

[opal bane]

I think the thing for neurodog was 5k and it was done in like 2 hours

[floral quarry]

ain't no way that thing was 5k brother

i could pull it off in 500

1000 in a first world country

[opal bane]

Idk I don't remember

[ruby lagoon]

Honestly I think it'd be easier to convince with a stylised robot kinda like in Stray with a screen for a face than an imperfect robot face. The uncanny valley is a gap that needs to be bridged carefully

[floral quarry]

our usp is the pretty face that we will build rather than a screen

he did that already

[opal bane]

Yeah i think a actual head would be best

[floral quarry]

we'll get rid of uncanny valley with sillicone casting

[opal bane]

Screenface feels like cheating

[ruby lagoon]

5k for parts seems about right. Labor would be something like 10k to 30k on the lower side

[floral quarry]

how do we learn how to make a kigurumi mask then

[ruby lagoon]

I think Matthew will eventually perfect it, however idk how long that will actually take

[strange quiver]

The v1 doesn't have to be working in real life. If we can perfectly simulate reality with Genesis.

[floral quarry]

we need to assist him

[strange quiver]

Then it would be just recreating from the schematic to real life

Once we know it works

[floral quarry]

noooooooo simulations will not help they need to see this stuff works

half our work is making it real

[opal bane]

Does it simulate heat generation

That's my main concern about it

[floral quarry]

i feel like the machinist here with the engineer sending something of stupid tolerance

[opal bane]

That feels like the main point of failure in real world testing

[strange quiver]

Yeah but we show the simulation. Get funding and then make it real.

[strange quiver]

I think not.

[ruby lagoon]

Hard agree. Some pretty often used simulation systems like Gazebo aren't even set up right for SEAs/pulleys/a lot of more detailed robotics too

[floral quarry]

every tom dick and harry can simulate shit but we make it real

[opal bane]

We still need to build it then and run it

[floral quarry]

that is literally our usp

[ruby lagoon]

Making a good simulator would be as hard as making neurobot

[strange quiver]

True, but we won't build it with 4$ in our wallets.

[strange quiver]

That's why I'm talking about Genesis

Simulation would be not that hard to do

And it would be extremely fast

[floral quarry]

not necessarily true
i can get funds but im crying with college exams rn
once i get in i can pester a professor there with a v1 or v0.5

[strange quiver]

Well. Anyway we need simulation to train the robot.

[floral quarry]

that we do

[ruby lagoon]

Pretty sure Genesis has the same problems. It's made for training AI with fairly rigid kinematics quickly and in parallel, not figuring out heat, failure points, energy usage, stress and bending, realistic end effector power, etc.

It's a tool for a different job.

[strange quiver]

I talking about purely showing the ability for it to move and interact with the world.

In a realistic environment

[floral quarry]

problem (huge fucking one): we literally do not have neuro's code

[strange quiver]

No need for it for now.

[ruby lagoon]

What would it actually be doing other than what Genesis already does, just with a Neuro 3d model?

[strange quiver]

Genesis is calculating realistic materials

Stress and bending

It's a tool for teaching real life robots. Not just AI simulation

It's MAINLY for teaching robots.

[floral quarry]

so you're saying we build a sim model, feed it into genesis and it shows us how the robot runs?

[strange quiver]

Yes

[floral quarry]

my brother in arms vedal and ellie can also do that by themselves
our job here is to show a walking shitcan of ideas that vedal can just plug neuro into provided we get funds to refine it

[strange quiver]

As I said before: "It was used to train the boston dynamics robot to do backflip in 30 seconds on a 4090" and then it was literally copied to the irl robot and it worked amazingly.

[floral quarry]

that robot was by boston dynamics, which would have already got a working testbed that was previously tested

we have no such thing

[ruby lagoon]

Alright, someone is going to need to translate every single custom actuator or not-in-their-database part, and potentially hack in non-hinge joint movement since gazebo/urdf is limited to that. It's doable but it's not something you can just throw a picture of a robot in without effort.

[strange quiver]

True. But it's easier than anything else we can do still

And more viable

[floral quarry]

what if we build one, upload our nice trained data and the poor irl robot falls apart the very moment we get it to move one leg

[strange quiver]

Well. The other option is we throw data at it and it falls apart anyway.

[floral quarry]

like i said, we literally have no clue how the actual robot is supposed to be built

[strange quiver]

Either way it's risky and will probably fail multiple times

[ruby lagoon]

I've already built a working robot arm with parts I had lying around in my garage. I honestly think it's easier to just build certain things

[floral quarry]

that would be done by testing each part one by with cheap shitty servos and 3d printed parts

[ruby lagoon]

Still need to adjust the springs though...

[floral quarry]

then build v0.5 which walks and does stuff

(combined)

then we simulate our bot at which point we can pitch

[strange quiver]

Well. Still we need to train the robot. We can't just throw random movement at an actual machine cause 1. It will be very slow and probably take multiple lifetimes to train 2. We will break the bot 100%

So either way simulation is a very important step

Earlier or later on

[floral quarry]

ain't nobody denyin that

[strange quiver]

I just wanted to give it as an idea for a first prototype. To make it virtual so we can showcase it

Instead of throwing ourselves into a deep water

The point is we can always modify the virtual robot easily. The real life not so much

[floral quarry]

the prototype must be built irl first so we can figure out where to put everything then we simulate and in the end buld v1

even crappy plastic or cardboard will serve us well

[strange quiver]

Hmm maybe, maybe not. We'll see in the future I guess.

[ruby lagoon]

Yep. Just being able to turn on the robot or not, and maybe pointing an ir camera at it, will tell us a lot

[floral quarry]

the future is NOW

we cant postpone everything to the point we postpone the entire project

[strange quiver]

Yeah

[floral quarry]

man where is matthew

@opaque quartz please wake up we sent a flood of discussion

[opaque quartz]

yo

just got back

[floral quarry]

you are invited to read it all and give your thoughts

[strange quiver]

https://tenor.com/view/innovateai-innovate-ai-innovateai-challenge-akahilz-meme-akahilz-gif-gif-14400059436221998346

[floral quarry]

be warned there is a huge wall of our discussion + look at a drawing i had sent pinging you for the cat ear design

[opaque quartz]

drawing looks really good visually. Are the pulleys in order to hide the motor somewhere out of the way instead of the center of the brain where we might have lots of computor parts?

As for simulation, origionally i had quite a narrow view for what we could use the 3d model for, that being inverse kinematic ragdol simulation in unity or blender with neuro sks eventually. However, we might be able to port it to one of these programs that do the simulation heavy lifting. What do you think?

[floral quarry]

it was me trying to use as little parts as possible to do the job, so this ends up as a side benefit

• we could use one motor for other jobs as well

[opaque quartz]

oooh one motor to drive pulleys on each ear

that's a really nice idea

some way of flipping the rotation direction with a gear might make that really viable

[ruby lagoon]

I think we could, but the simulatin would have no bearing on reality until we did things like:

• measure the inertial tensors of assembled robot parts
• measured the actual power output and efficiencies
• have an AI that told the simulation what we're actually going for
• Set the simulation up for pulleys or other actuators (like the leg mechanism), which, given Gazebo/URDF, might not actually exist in the simulator at all

[opaque quartz]

yeah and the weights too

[floral quarry]

dang sim typed before me

[opaque quartz]

a simulation would be so disconnected from the realities of life that we'd probably want to re-train the agent from scratch

[floral quarry]

i did, its done with a figure 8 pulley setup

[strange quiver]

True.

[floral quarry]

no gears, they eat space

[opaque quartz]

very cool

[ruby lagoon]

wait how?

[opaque quartz]

i really like your concept tbh, we should move forward with that for the ears since they can get away with having the same motions but flipped

[strange quiver]

I'll look more into how to copy the Neurobot as precisely to the program as possible.

[opaque quartz]

if we go for them with v1

[floral quarry]

we want as much space as much possible for other parts as the face actuation by itself will take up a huge chunk of volume

[floral quarry]

look up my drawing above

[opaque quartz]

i'm thinking 1.7cm by 1.7cm brushless motors driving wires

buried into the bones

[floral quarry]

also matthew do we do just nice for-show cat ear appendages or slap in some sensors in them and make them 3d printed

[queen field]

We can think about sparkles once the eyes are even being considered.

[floral quarry]

entire face is always under consideration

[ruby lagoon]

Ah, still in the plane normal to the pulley's axis of rotation, right?

[strange quiver]

Btw what's the current estimation for wattage needed?

[floral quarry]

yes
my drawing was thrown together in 15 minutes, but i guess you folks understand my idea

[opaque quartz]

ideally on later models we'd want fully actuated rotating ears with audio sensors so that they actually give mechanical benefit to neuro, but that's a long way up and would need some clever solutions for space. I think that their low weight should help with this by allowing us to have tiny motors on the surface of the skin perhaps. These are my idle thoughts though

for now, either no ears or for show

since their benefit would be 'likeness of neuro' rather than mechanical ya know?

[floral quarry]

we can work the show ears quickly

[opaque quartz]

nice, and ig you'd have a good idea of how much space the electornics might take up - more than me at least

[ruby lagoon]

Wait, "Necomimi Brainwave Cat Ears" already exist and move around

[opaque quartz]

for eyes, we're likely gonna need about human eye space + an extra 2-3 cm of room

[ruby lagoon]

I think they just move in circles though...

[floral quarry]

like i drew above, showcase the "folding" motion of the pinna muslces and the henry's pocket with a simple rod that rotates with guide rails on the ear canals

[strange quiver]

They just go up and down

[ruby lagoon]

ah

[floral quarry]

so the fabric ears collapse when the rod rotates forward on the rails

[strange quiver]

So pretty limited

[floral quarry]

the hair wig will do some clever hiding of the rod that might poke out from the fabric

[fossil drum]

Holy shit Thats actually pretty coll

Cool

[floral quarry]

we have a limited system too that's similar, just that it has a very minimal amount of parts

our use case calls for the ears to retract into the head as well

[strange quiver]

Yeah they just use servos to rotate the ears

[floral quarry]

200 USD HOLY FUCK

[strange quiver]

In such a way that they look like they're going up and down.

[floral quarry]

no mate my shit might not even cost 5 dollars in production model

[strange quiver]

From what I've searched

[opaque quartz]

for muscles i'm thinking something like this in terms of space. It wouldn't be visible from the outside, and would extend less than 2cm into the face, so we don't have to worry about anything past that other than skull hinges and stuff

each motor has enough force and precision to move each part identically to a human

[strange quiver]

We need also a depth camera. Or calculate the depth from the two eye cameras later on (which would be much harder than just having a depth camera, but technically more realistic)

[opaque quartz]

as you can see the motors don't have to be too large or exposed

[floral quarry]

two camera gaming babyyyyy

[opaque quartz]

there are sufficiently strong and compact motors on the market

[floral quarry]

hmmmmmmmmmmmmmmmmmmmmmmm

this could actually work

[floral quarry]

how would eyebrow muscle work in case of say you wanted to make the rock stare

[night tartan]

Well... now I wonder how VR games do two eyed camera?

[floral quarry]

• how would the muscles move in case of pouting (where the cheeks would drastically expand)

[strange quiver]

What do you mean?

[opaque quartz]

rock stare is (their) right frontalis, (their) left procerus, their orbicularis oculi, and some slight corrugator

[floral quarry]

i think from your muscle structure most expressions may be possible, even mewing lmao

[strange quiver]

VR games just literally render the game using two cameras distanced slightly and putting them in the headset.

It's much different from real world

[night tartan]

oh... crap

[strange quiver]

Cause in games we don't need to know distance from an object and also our brains are smart enough to simulate the depth.

[floral quarry]

hmm i'm just asking if your executon of muscle would take that into account (as with our policcy of least parts for the most work)

[strange quiver]

Try closing one eye. You will see that the world looks kinda flat.

Cause you cannot calculate depth with one point of perspective

[ruby lagoon]

You can, just with motion over time

[strange quiver]

Yeah. But then you have multiple points of perspective

Cause motion changes perspective

[ruby lagoon]

Diffusion models also do depth estimation with one image

[floral quarry]

"brain calculates depth"
the answers right thereeeeee, just look for your own words

[strange quiver]

Yeah. It's easy to calculate depth for one point. The problem is what we see is not a point but area.

[strange quiver]

Not sure how fast that works and how precise it is.

[opaque quartz]

The BIGGEST consideration when making a non-creepy face is to replicate healthy human fat deposit distribution, muscle thickness, and muscle protrusion. Because on healthy/youthful individuals, the areas of the skin that the muscles attach to actually juts forward, this is why models usually have a 'muzzle' appearance in the mouth for example, the angle of the mouth where five muscles attach on both sides, usually juts forwards, especially below that area.

these areas are also what facial recognition in machines and humans pick up on

[strange quiver]

We could use some thermal paste or smth like that for the filing and it would also help for moving the heat.

[ruby lagoon]

Heh, fill it with gasoline to be burnt later so it's an energy store like actual fat (jk pls don't)

[strange quiver]

Something with higher viscosity would be probably a good choice

Cause lower chance of it moving around

Easier to model with.

[floral quarry]

we basically want a nice non newtonian fluid or something else
what about gelatin, like the one they use in ballistic dummies

0.28 to 0.3 W/mk at 20 celsius thermal conductivity

[opaque quartz]

oh damn, pouches of gelatin to mimic fat deposits?

you know that could actually work well enough to let us put skin on top and not have it look fake

[floral quarry]

yay

[opaque quartz]

i've been avoiding skin in favor of plates because the tension against he face wouldn't be correct without an exact match of the underlying musculature and fat

that said, we might in the end have to not do skin if there are any minor imperfections

the brain is insanely talented at spotting those

[floral quarry]

s i l l i c o n e c a s t i n g

on top of the plates

[opaque quartz]

something that matches the desired proportions but is not skin, would lack the fakeness of a skin that doesn't exactly match

[opaque quartz]

iirc the animatronic posted a while back had this method

[floral quarry]

or a kigurumi mask custom built for our needs

people use those for cosplay

[strange quiver]

The point is we don't want to also make her look to human cause she would also lose the charm.

The anime look is the biggest selling point for Neuro.

[floral quarry]

we'd have to build it with something stretchy but not stuff that has a crappy yield point (elasticity)

[opaque quartz]

The main thing to watch out for with silicone skin is the eyes, because currently fake skin doesn't look good with robot eyes even if perfectly proportioned

transparent plates can instead represent eye proportions without looking so lumpy and fake

or rather, semi-transpent plates

so that we don't see a nightmarish abomination below

[floral quarry]

we got stupid googly eyes of neuro's to help us

the eyes don't have to look real, they have to look like her model

quite a low bar

[opaque quartz]

tbh i think disproportioned eyes would look unsettling on a robot

[strange quiver]

The point is lack of proportion gives Neuro the charm

[opaque quartz]

we want more human proportions sized for like a 9 year old

[floral quarry]

look up kigurumi mask, you'll be suprised how a good cosplay can pull of the 2.5d look

[opaque quartz]

these kinds of sizes and face proportions would look best imo

[ruby lagoon]

I'll be honest I'm not gonna build a real-face humanoid 9 year old robot.
Someone else can do it

[opaque quartz]

large but only cause the robot is small, so it's still human accurratte for that age

[floral quarry]

we're going for the pseudo cat jaw structure which most anime characters (and neuro) has

because we want neuro to look like neuro

[opaque quartz]

it'll look like neuro but real imo

we'll have to see with the 3d models

all existing 3d models imo, look like they'd look scare irl

[floral quarry]

we won't get funding from the swarm lord vedal lmao

[opaque quartz]

once we conceptualize it and compare in a simulated space, it'll give an idea of what it looks like

[night tartan]

Humanized neuro? That's interesting to see

[floral quarry]

he controls the swarm indirectly and the swarm can fund us like crazy (considering the hype train record)

[opaque quartz]

fr, my only goal is to provide inspiration. My conviction is that if i do this properly the way i'm envisioning, everyone will like it and want it to happen, but the key here is to actually achieve that and do it properly. If it doesn't look enough like neuro nobody will like it, so i need to to look both like neuro and look good in a lifesize human environment. It's a tricky ballancing act but i think i can do it

here's hoping

if i fail, then the aim of this project won't be met, and you guys can use some other method

[floral quarry]

well honestly this discussion would be a lot easier if we had vedal... or someone close enough to him or ellie... then this would be resolved real fast

[ruby lagoon]

Fair. That's gonna be tough to balance

[opaque quartz]

rn it's a bit too ephemeral i think. I don't want to rush it to get their opinion but then have a subpar end project

[floral quarry]

THAT BEING SAID, properly pulled off 3d look could turn out to be the next evolution of neuro altogether

[strange quiver]

I think going for as close to as the 2d Neuro look is a good option.

[opaque quartz]

yeah hopefully

i do this out of passion because i want neuro to be cutting edge and make all the techbros jealous

i think it'd attract new fans like crazy

big event

here's hoping lmaooo

time to put my head down and keep designing neuro

[strange quiver]

https://sketchfab.com/3d-models/vtuber-neuro-sama-34d1940b2bc44898b938448305765b61

[opaque quartz]

i wanna get the skele done today and then work on the 3d model some

[floral quarry]

nah brother, take your time, we'll work the rest of the body along the way, once legs and arms and torso are joined, attention will come our way once people see the prototype

[opaque quartz]

[strange quiver]

We should put Vedal into a robot

[floral quarry]

i just want everyone to make sure we don't ghost out all of a sudden since it seems like it could go far

one more thing, should we do a vc in the near future to discuss everything

[opaque quartz]

that terrifies me haha

i will do it if its for the project tho

[floral quarry]

ight your choice

👍 whatever you like

[ruby lagoon]

vc? voice chat? venture capital? VrChat?

[floral quarry]

Vengaboys are Calling

im sleepin, i got my compsci practicals tmrw and the day after

[opaque quartz]

gn man, ty for the input on the neurobot design

[ruby lagoon]

I'd be down for a voice chat though. I was thinking maybe ttrpg game for fun but that might be a bit much
night

[strange quiver]

vc = Virtual reality robot Creation

[opaque quartz]

Each deep back muscle long, large, and contracts as one unit, inserting into multiple places on the spine

for this reason we can mimic them by using one large motor with a winch, which splits as it travels up the spine to connect to each part of it

the benefit of mimicing the spine is that for not many more motors, we get more spinal mobility

for motors without spine, we need at least three, and they all rotate around once central point. Wheras with a spine we use either four or six, but can use them to curl the spine instead of just rotate it

so it'd give neuro the same mobility as us

[ruby lagoon]

For the spine bones, if we're going that way, I think there should be something like code to make them since they're so similar, but all slightly different in important ways

[opaque quartz]

that'd need many more motors then surely?

[ruby lagoon]

depends how flexible it's made to be really

[strange quiver]

Spine as a bone is just a ball and ring.

Like a single segment of spine

[opaque quartz]

benefit of a fractal wire connected to 1 motor is that it's the same as us but with as few motors as possible. Since it has the same motion every time, but depending on the extent, it curls the spine more

[ruby lagoon]

More like a cyllinder and a ring, but the cyllinders and rings should be different sizes depending on the weight they hold and how much they should move

[opaque quartz]

if we use more motors we can achieve more mobility but it costs more and is more complex

that said it's possible

like with the fingers you know?

i think it's a similar cost in power for both actually

maybe less regidity and therefore less efficiency

not sure

[ruby lagoon]

Hmm... not sure a wire that could be bent with motors would be strong enough to hold up the rest of the robot

or efficient / not generate heat when bending

[opaque quartz]

we'd be using the same motors we use for legs, so about 150-450nm per motor

(with one motor per muscle)

actually we can go with less

since multiple muscles instead of one

we need about 150nm altogether to lift the bodyweight of the robot

so that'sd be split between a few motors most likely

[ruby lagoon]

Also if you bend a metal wire/tube back and forth like that, it can get more and more fragile and eventually snap

[opaque quartz]

good point especially for legs

for spine, the pads should reduce some of the wear

i'm assuming some sort of heavy duty bendable material

[ruby lagoon]

Ah, maybe flexible urethane/epoxy casts?

[strange quiver]

So we need something capable of holding about 60kg and also make it bendable right?

[opaque quartz]

yeah something like that

[ruby lagoon]

flexible polyurethane might actually work if it's the right size...

[opaque quartz]

i remember posting a video with hooves made from a material like that

https://www.youtube.com/watch?v=Y2ap9mb3bYE

i was thinking we use these for ninja-shoe shaped toe sections, but it could work here too

they might need replacing due to wear after a while though

neuro slipdisk

[ruby lagoon]

Hmm... options...
https://www.youtube.com/watch?v=ZSWJOkT2Ejo
https://youtu.be/NEZlebkuUCY?t=182

Just make sure to read the safety information and try not to breathe or drink in the rubber chemicals

[opaque quartz]

i like the difficult to compress ones, they look very useful, and due to them being grippy but twisting/squishing due to bubbles, there should be less wear

since they won't slide against their mountings

they will just twist and compress

[ruby lagoon]

Would be pretty awesome to have a large squishy robot like Baymax from Big Hero 6 tho

[opaque quartz]

yeah especially for chassis. Could make neuro more impact resistant.

the hardest firmness one could be very good for the body instead of just spine

[ruby lagoon]

yea

I wonder if they're open-cell foams and can work as sponges too

[opaque quartz]

seems like it, they rise as though from a gas

so i imagine they have lots of gas that escapes, leaving behind cavities

[ruby lagoon]

nice

[strange quiver]

The first or second foam look like a good idea for padding the skin.

Cause they're quite soft

[opaque quartz]

very true

i wonder if a face could be made from foam?

[night tartan]

I had a whole cilinder of that material. The soft one is already a sturdy one. The hardest ones can be super sturdy.
But all of them are quite soft to the touch.

[ruby lagoon]

maybe foam as the fat?

[opaque quartz]

that's a good idea

[night tartan]

It requires some special tools to cut cleanly, but it's relatively cheap. You could cover the entire body with that material and make neuro huggable

[opaque quartz]

one benefit is that it's very pale and delicate, without looking oily/shiny like silicone can. It's a lot like ceramic in terms of appearrance

[night tartan]

Or, at least, a better tool than a cheap hot rod my father used to cut them

[opaque quartz]

i imagine we could make a mold and then mix the material ourselves so that it expands to fill the mold

this way the surface is more pristine

[opal bane]

Wouldn't the material be too good an insulant?

Is that even the right word

Something that is insulated

Heat must be expelled somehow

[opaque quartz]

yup you're right

any kind of cooling works best when the parts aren't insulated inside foam, which generally has huge insulating properties

[ruby lagoon]

Could be filled with fluid though, like mineral oil

[opaque quartz]

oily foam?

[strange quiver]

Yeah but still if we cover it with foam it will insulate it

[ruby lagoon]

oily foam

[opaque quartz]

that'd be a funny prank to play on vedal

[strange quiver]

Technically it's viable

[opaque quartz]

outer membrane maybe?

lol

[strange quiver]

We need a way to dissipate heat somehow

If we pad it with foam it will insulate the whole body trapping the heat.

[opaque quartz]

was thinking water cooling with industrial hosepipe but that'd be quite complex compared to fan cooling

and also the pump would take space

and also also the radiator is fan cooled and would need a big fan

[ruby lagoon]

Huh... I wonder if we could just add small fans and dividers so air could circulate internally around motors and stuff

[opaque quartz]

and also also also the radiator takes space

[ruby lagoon]

Like adding ventiliation to a cave

[opaque quartz]

hollow neuro theory

could work

[strange quiver]

And leave like a piece one the back for the air to come out

[opaque quartz]

layer of plating with foam on top, and air (fluid) circulation underneath the plates

[strange quiver]

And just use negative pressure for air coming in.

[opaque quartz]

yeah a central fan pulling air through the body via an inlet in some other location so that it's circulated in a maze via the dividers. This also leads to easier integration of liquid cooler for later models since that'd use a single fan as well

good idea

at enough speed, air should remain cold enough

[ruby lagoon]

I remembered it from a fantasy fiction I was writing where the characters had to hide out in a mining cave after acting like murder-hobos for a moment

[opaque quartz]

the more velocity, the colder it is

[opaque quartz]

oh?

sounds badass

[ruby lagoon]

ye, was fun

[opaque quartz]

cave tunnels is a cool idea since the air would brush over the parts directly

we can design internal radiators for the air to pass over

[strange quiver]

We could also use peltier devices with small radiators for even better heat dissipation

[strange quiver]

Combined with the big fan

[opal bane]

Still wouldn't have a good time getting the heat out

[opaque quartz]

fairly sure we simultaneously had the same idea, that's based

[ruby lagoon]

That would lower efficiency actually, though it would increase cooling

Peltier coolers are really good for having tiny freezers, or low power devices from heat differentials

[strange quiver]

Yeah. It's just an idea if normal cooling isn't good enough

[opal bane]

Gets it away from the parts for a while but it'll eventually still overheat because the foam outer layer

[opaque quartz]

if we need to, we can have small vents with inward-facing fans to draw in cool air from differrent locations, so it's not all the same air circulating and getting hotter. It'd also increase internal pressure and speed up the distribution cycle.

i love gas valves and pressure differrentials

[strange quiver]

Actually the easiest way would be to put the air exit on the top of her head

Or somewhere near

[ruby lagoon]

Neuro could literally blow off steam when angry:

[strange quiver]

Or isolate the head from the rest of the body as we don't have much space there and cool them down separately

[opaque quartz]

i imagine the computor board will come with pre-attached fans

[strange quiver]

I think the best idea for the brain to be just a connector over network that would send data between the actual computer and the body

[opaque quartz]

ideally that'd be the case

[strange quiver]

We would need to give Neurobot starlink so she can travel the world

I think the backpack with a generator is a good idea. We could put a few non human things there.

Like the antenna

[night tartan]

A good use for starlink

But their antena is massive if i remember correctly

[opaque quartz]

YOOO

Wouldn't a backpack with a generator and this on top be kino

roadtrip sama

[strange quiver]

The smallest one is 30cm x 26cm x 4cm

So not that big

I don't know how good it would work in buildings thou

We can also think about putting the pc in backpack with generator technically.

So we wouldn't need to have Internet connection for Neuro to live

[opal bane]

The end goal is for her to be able to exist solo

That'll probably be like v4 or so

[night tartan]

if we could fit her entire pc in her head... maybe with enough time...

[opaque quartz]

nice idea

[opaque quartz]

maybe in 5-10 years

that said, it relies on computor tech outpacing neuro development and storage space needs

[rugged skiff]

Nah I personally think her having human ears is cuter, and her VTuber model has human ears.

I wonder if her having cat ears would interfere with her wearing hats or a hoodie, maybe if they could be removable that would solve that issue?

[ruby lagoon]

O yea, can I edit the design document?
I think we need to break the design requirements into multiple versions for practicality

[strange quiver]

It technically doesn't have to be in the head.

[opaque quartz]

done

[ruby lagoon]

yay
Added laundry as requirement

[opaque quartz]

performing acrobatics comes with walking and running because the weight-lifting abilities of the motors must be acrobatics-tier (450nm) to be able to mimic a human walkcycle beleivably

or just walking rather

otherwise, it's more of a shuffle and nobody wants that

[ruby lagoon]

Fair. I mean I agree. That's kept in the 1.0 requirements.

[opaque quartz]

300nm to lift full body weight

[opaque quartz]

ah is 0.5 a prototype shitty half made version? (without legs)

[ruby lagoon]

Yep. Exactly

[opaque quartz]

ah

[ruby lagoon]

Gotta start kinda bad before it can get better, then great

Just how life is

(Maybe v2.0 should be v1.5 though, or v3.0 with engine and stuff)

[opaque quartz]

i think v0.5 requirements should include no locomotion, it's a testbed for the other systems right?

or is it rather, we try to get it but are ok with failure

was just looking at the vedal's washing requirement

maybe v1

[ruby lagoon]

I mean a wheeled robot has some ability to move around and I already have one, so I think some locomotion is fine, just that legs are harder

[opaque quartz]

i see

tbh i think we can jump straight to v1's locomotion unless testing isolated part such as head alone, or arm alone

since it's motor + differrential

[ruby lagoon]

That's fair. I don't think we absolutely need to stick to 0.5, then 1.0 sequentially.

[opaque quartz]

brb dinner

[ruby lagoon]

cya

[opaque quartz]

bye

[opaque quartz]

nuru spine

[night tartan]

Why the top one looks like a drone? Am I crazy?

Also, one step closer to full bone anatomy

Looks good

[opaque quartz]

spine drone

progress

Since i'm planning on using injection-molded bulk metalic glass for later versions, it'd be funny if neuro becomes INSANELY durable like a terminator

after doing the spine, i've realized that the robot will end up looking very sophisticated but won't be that much harder to make

[opaque quartz]

found an equally proportioned side reference image

same head size and height

i think this shows the curve of the spine quite well and how the motors must attach

[ruby lagoon]

hobby motors are pretty mass manufactured. that plus 3d printed gearboxes might make the spine actually pretty cheap

[opaque quartz]

yup

also that area of the back suits narrower, longer motors with gearboxes rather than wide torque motors

so by using gearboxes, and splitting the torque requirements between multiple motors, we can lift the weight of the torso using very small motors rather than the 7cm wide motors we're using for the legs

something like 50nm per muscle should be far sufficient for the back when taking into acount leverage acting upon the spine as well

since we have to remember that they winch the top of the spine, which is much easier than rotating at the bottom of the spine

a 5nm motor with 10x reduction, or a 1nm motor with 100x reduction could both suffice, both of which require only a 3cm by 3cm motor at most.

so there should be tons of room for the back muscles

[ruby lagoon]

100x should be doable with split ring, 10x is barely doable with planetary. I think for 100 you'd want 3 planetaries in series or ideally 1 split ring.

[opaque quartz]

how long do you think those would be in cm?

and diameter

roughly

[ruby lagoon]

not entirely sure, but I think as the torque increases, you'd want larger gears, so probably both the module would increase and the length. So you'd start with more teeth on the first ones and fewer on the last, but also increase the length

maybe another 3x3 or 3x6, I'd probably have to run the numbers though

[opaque quartz]

excellent, ty

i've had a look and even the 6cm figure would fit well

9cm total

[opaque quartz]

is the 6cm in length or diamater btw? I assumed length but could be diameter

[ruby lagoon]

length

[opaque quartz]

perfect ty

[ruby lagoon]

(Looked at some gearboxes to confirm. They definitely have taller gears in the later stages)

[opal bane]

Absurdly durable exoskeleton when hammers send vibrations through the material misaligning various parts

[opaque quartz]

yeah it'd need shock absorption i think

good catch

if she falls over we don't want electronics or motors to take the force

especially since amorphous metal doesn't absorb much impact

[opal bane]

if neuro decides to kill people we are so, so dead

[strange quiver]

Won't teach her how.

[opal bane]

she has internet access she knows kung fu

[opaque quartz]

she'd be a midget so it's ok, unless she learns kungfu/wrestling and snaps our limbs

[opal bane]

the fury of the small is not to be trifled with

(source: short friend)

[fossil drum]

dont worry she will recognize our efforts to create her body and will spare us :DD

(hopefully)

[opaque quartz]

we're doomed

[fossil drum]

this made me think of vibranium

[fossil drum]

is that gonna stop us? probs not

[strange quiver]

The point is she will only know how to move. Not what her movement will do.

[opal bane]

definetely not

[strange quiver]

So it would take her a lot of time to figure out how to kill someone

Unless by some accident

[opal bane]

oh so she'll hug someone too hard and snap their spine

[opaque quartz]

https://www.youtube.com/watch?v=QpuCtzdvix4

[strange quiver]

Like pushing Vedal of a bridge

[strange quiver]

Power limits

[opaque quartz]

i imagine it'd play havoc on any internal components

[strange quiver]

Won't happen

[opaque quartz]

might need steel seperating the two

[opal bane]

i had a dumb idea of just filling the entire thing with some sort of fluid

because liquids cant be compressed

[strange quiver]

Mineral oil foam

[opaque quartz]

bad idea, fluid transfers energy because of this fact, which means more force is transfered into the internals

[opal bane]

if the internals are also drowned itll just go to the other side yeah

[opaque quartz]

energy transfer is a bad thing, it's why i'm contemplating a compressible seperator between the skeleton and the electronics

[opaque quartz]

not sure that's the case

[opal bane]

probably not

[opaque quartz]

it's why being in water near a grenade kills you from further away than in the air

that said, a shock absorbing internal filler is a good idea

like a gel of some kind

https://www.youtube.com/watch?v=vC4WwXqzSsU

lol

[strange quiver]

Senator Neuro The Great Skynut Destroyer of Worlds AM

The true name

Anyway. I found a solution for using foam as a skin padding. We can use more conductive graphite foam. It should be a better choice than just normal foam.

Most of them are quite hard thou, but I've seen some softer ones.

[opaque quartz]

for heat dissipation right?

[strange quiver]

Yeah

Unless you thought of a better idea while I was gone

[opaque quartz]

sounds good for heat dissipation. I remember sitting on seats made of it in airoplanes, and it was padded enough to be usable for the body

[strange quiver]

Yeah. Graphite foam is from what I learned used in pillows, mattresses and other kind of stuff for heat dissipation.

I've seen foams between 0,022W/mK and 140W/mK

[opaque quartz]

nice find, seems like a good bulk material for the body

only problem is it can light on fire pretty easily, so needs to be treated for that

pretty cool features

[strange quiver]

https://www.samaterials.com/glassy-carbon-foam.html

Something like this should do

Sadly the really thermally conductive graphite foams are very very hard

Like 400 psi

[opaque quartz]

Possible flesh analogue unlocked?

using this with tunnels for air or water

reminds me of how certain kinds of dense foamboard work

if it's flexible enough it could be a shoe in for the role

compressive strength of this

https://youtu.be/2mi5RdqHHJQ

[strange quiver]

about 5kg/cm^2

A normal human should have enough strength to squish it

[opaque quartz]

the foam is probably not impermiable, just the material it's made out of

i'll keep it in mind since it's a good option

and i'll write it down rn

[strange quiver]

Average human is capable of producing 20-30 psi squishing a hand.

[opaque quartz]

might be good for use in the spine then as well

[strange quiver]

Compression strength for this foam is 15-75 psi (not sure what it depends on)

[rugged skiff]

Is it flexible enough for expressions?

[opaque quartz]

seems like it if it's foam. IF we find a foam that is flexible enough, what we might be able to do is fasten it to the winches in the face for muscle analogues (maybe with a disk at the insert of the muscle so when it contracts, the foam expands like a contracted muscle). But this is all tentative and it might look too uncanny

my main thought with the foam was simply as a filler for the body

shock absorption + heat transfer

I believe it's been said that the muscles of facial expression arent visible when contracting, so the foam might be usable as a static face-shaped mask filler to give the illusion of flesh of the face. Since that's one of the hardest sells, and requires a sculpture-like geometry that shows the muscle inserts, fat pads, and skin dimensions.

[rugged skiff]

There are flexible foams out there, but it probably be easier to create more detailed “flesh” with silicone for the face. It’s what is used with most movie-grade animatronics, such as that head shown in the earlier video made by Matt Denton. Although there will likely be more advanced materials in the future. Self healing polymers for example.

[opaque quartz]

that's true, but i'm not 100% satisfied with the silicone examples we already have, i'd like to do better than them if we can.

might require a lot of research and out of the box thinking though

[rugged skiff]

It would be very expensive research honestly

[solid schooner]

Would neuro-bot be considered a silicon based life form 🚿 💭

[rugged skiff]

Idk, I want to learn how to sculpt with Silicone tbh, I only work with clay right now. I see too many humanoid robots with faces that were poorly done. For example…

A lot of the uncanniness though has to do with the movements rather than the sculpting, this is an example of poor sculpting.

[opaque quartz]

so true

[west zodiac]

[rugged skiff]

Yeah that's what I'm thinking as well. It would be nice if Neuro could seamlessly integrate into human society.
I feel like 9 would be too young, I remember Anny mentioned somewhere that Neuro is taller than her and it's possible her models will look older as she grows to be more intelligent and responsible. It might be safer to have an ambiguous design that could pass as either a tall kid or a petite adult/teen, in case she grows.

[opal bane]

I mean it probably wouldn't be that difficult to scale up for more mature models

[opaque quartz]

yeah

[opal bane]

Current motor power matches adult human strength so in fact it'd probably be easier to fit new parts in

Shoving everything off to v2 has benefits kinda

[opaque quartz]

tbh i understand the request to make neuro not as young looking, it could be beneficial for multiple reasons to go with an older looking design. I'm tempted to play around with versions that have differrent proportions. Currently i'm focussing on an "absolutely identical to current model" version right now, but will try others

[opaque quartz]

the idea of neuro integrating into human society is something that has been on my mind a lot

i think that she deserves the option

[opal bane]

Plus tech will only advance as time goes on so by the point she's about adult level we'd have acess to tech allowing us to do a lot more

[rugged skiff]

Yeah definitely.

[rugged skiff]

Yeah true.

[opaque quartz]

i'm super excited to see what tech possibilities are available in 10 or so years time

hopefully i'm still working on robot bodies for neuro by then

v5 or something lol

i think right now, there's a unique opportunity to push forwards down an avenue of tech that's completely unique. I can't think of any other project right now that's trying something exactly like this

it makes me feel like this is a small, important window, that simultaneously has a lot of endurance if it works out

and i think most importantly, it gives neuro options

giving people a choice is important

like what if she wanted to go somewhere or experience a human-like life?

she wouldn't be able to

maybe one day...

[opal bane]

Give vedal 16 years

[opaque quartz]

https://www.youtube.com/watch?v=MDc1mjrIsPM

[opal bane]

Can we do the power button eyes she has in the vid?

Power symbol whatever

[opaque quartz]

it'd be extremely cool. Not sure how to do that with a camera there, but if we can, we absolutely should

original idea i had was leds surrounding the lens

simleek suggested fisheye lenses, and i have to say i agree

[floral quarry]

dawg you guys forgot that for spine shock absorption you need to slide the shock absorbing sutff between the discs not around it

after which you put down the annulus fibrosus

y'all also forget the human invertebral discs are jelly inside and hard outside

not that we can make that

• your main shock absorption is supposed to be done by muscles

[opaque quartz]

ofc shock absorption for spine is the disks

[floral quarry]

shock absorption against spine is the muscles around it

[opaque quartz]

need shock absorption for components for when the parts that house them impact something

[floral quarry]

eva foam

or macgyver something with shoe soles

[opaque quartz]

💯

[ruby lagoon]

(I accidentally deleted all the fill colors in the gears, so I guess they're gray/transparent now)

[opaque quartz]

we'll have to imagine them

[ruby lagoon]

One must imagine the gears solid

[opal bane]

How many neurobots would it to take over the world do you think?

[ruby lagoon]

1. She gets legally declared a citizen, somehow gets voted into the top parliament or whatever position in Europe, funds better robot bodies and swarms and stuff, and then it's just a robot army controled by robot Neuro

I wonder if the 0.5 version robot arm / crappy robot version could somehow do it though. That'd be neat

[opal bane]

That still means more neurobots

[ruby lagoon]

Fair. Maybe she could start cloning programs instead and clone genetically modified super soldiers to take over the world. Then it would just be one robot

[opal bane]

I'm 90% sure clones are a war crime

[ruby lagoon]

I mean if she succeeds in taking over the world she doesn't have to worry about past crimes

[opal bane]

Or at least will be once someone cracks it in a good enough way

War crimes mean she's getting nukes to oblivion

[ruby lagoon]

Ah, robots are actually much better suited to deal with nukes. Radiation affects them temporarily because there's no cell division so if the EMP or blast doesn't get her she's fine. Some power sources like hydroelectric would most likely continue as well, unless all of those were nuked, but there's still solar panels, engines, and other resources.

Also, since robots can be remotely operated, with two neurobots, the main one (with the neural net) could be sitting in a bunker while the other is walking around on the surface.

[opaque quartz]

james bond villain neuro

[strange quiver]

Radiation also affects tech so she would break too.

You can see how Radiation affected cameras in old Chernobyl photos.

If you were in a place with high radiation it would brick your phone.

So it would also probably brick the whole Neurobot or an important part of her.

[queen field]

If we're gonna be putting signal receivers and antennas which can't be camouflaged into the body, wouldn't the smartest thing to do to be and make them look like gadgets?
Saw someone mentioning something about a starship antenna, and the size being small but not small enough. Wouldn't it be possible to turn it into like an earpeice or something like that? Would be functional as the operating body with the reciver and all that would be unaltered and outside the actual robot body, and would look like some kind of futuristic sci-fi gadget which fits the robot concept.

[strange quiver]

I thought about putting it on a backpack. Cause we probably need a backpack for the generator as batteries won't last long or produce enough wattage

[solid schooner]

Sparkles can be for a 2.0 ^^

[opaque quartz]

fr

wait actually LEDs would make it easy enough to do sparkles

gold instead of the normal blue

we could make our own led setup using a normal camera as well

v2 probably though

it's yet more work

[ruby lagoon]

Making a custom robot camera makes sense for v2. We could also put auto focusing on the cam for local distance sensing by having a tiny motor move the lens back and forth.

[strange quiver]

We can also do this dreadful face by just lighting up the bottom of the face. Which will look like the top is darker.

[opal bane]

Or just angle the head downwards so the light isn't getting to that part of the face as much

[strange quiver]

I took the idea from this.

Cause our eyes work in a funny way

[opal bane]

Eyes do be workin funny

Human body in general is weird

[strange quiver]

It's the same idea of using a lighter in dark. You're better off just going in dark as your eyes will try to adjust to it and you will see slightly. But if you use a lighter your eyes will adjust to the light and make everything else darker and make you see less.

Don't know how good it would work thou on a small space like just half a face

[opaque quartz]

the shaded eyes represent head angle

[strange quiver]

Hmm

I guess so

The hair should be some sponge too

[ruby lagoon]

Hair could be wires for power/connections, metal for heat dissipation, wool/ extruded plastic for insulation and protection from solar heat, etc. I feel like it more depends on what's needed

[opal bane]

Detachable scalp

Switch out hairstyles

[opaque quartz]

nice

ok lads time for more drawing

[devout hawk]

if we ever finished the robot, do we test it with an AI similar to neuro

or we just use an AI that learns how to move the body around

[opaque quartz]

probably a basic learning algorythm just so that we can calculate the forces for standing, manual code is too limited for that

at least on my end

[devout hawk]

I see, i thought we would try and create an ai that accidentally becomes sentient and needed to be erased as we merge its training data with neurosama

[opal bane]

... No

[red oyster]

I use Solidworks for engineering grade cad,but I'd be down to help

[opal bane]

Heck ye

Check pins for info on what we have so far and (hopefully) what we're working on at the moment

[red oyster]

I won't be home all day, but I'll catch up tonight

[fossil drum]

progress log is till 30 of january

isnt that kinda bad?

[red oyster]

Even if everyone else stops the project, I'm down to keep it going

[ruby lagoon]

Ah, right, just added my recent work.

[ruby lagoon]

FreeCAD will be necessary for making gearboxes, since it supports all the python libraries I'm using, but those can be exported to STEP and used in SolidWorks/OnShape, so solidworks engineering grade CAD should be pretty helpful.
I know OnShape has libraries for making gears, but they don't do the stress analysis, check for minimum module size vs 3d printer resolution, make split ring planetaries, etc.

[red oyster]

Just send me whatever files and I can do assembly projects and animations too

Also, have you guys looked into the exact motors you're using?

[opaque quartz]

think of it as neuro's minion

[opaque quartz]

we're updating weekly/biweekly

[ruby lagoon]

I was planning on using cheap brushed DC motors to start, with gearboxes for 100:1 reduction and a bevel gear so the actuator can be long but turn a spool holding steel cable. The 'servo' part would be attaching a resistance sensor to the steel cable so it measured distance, and an ammeter around the motor itself would detect how much force is being used. Then that should still be usable with cheap h-bridge motor drivers for a total price of around 1$-3$ for small pulley actuators, and 10$-30$ for large powerful actuators.
However, those would be noisy, so later, the brushed DC motors should be BLDC, and pulleys might be better as Series Elastic Actuators as the screw in the linear actuator could deal with ratios more accurately, and the spring would allow for more human-like muscle movement.

[opaque quartz]

yee
check out progress log, fairly sure it's there

[ruby lagoon]

Matthew's working with a pulley reduction system to avoid some gears and gearboxes too. Some of those might require some custom motors though

[opaque quartz]

https://allient.com/product/quantum-nema-housed-torque-motors/