@ebastler
You say "what I am missing in your approach is some basic estimates on the required forces and velocities for all your "muscles", followed by estimates of the required volume, weight and electrical power consumption (based on commercially available motors)." --- I have done this calculation work privately but did not post it on my project public updates. I have only done very in depth calculations for finger joint actuation so far but the other muscles of the human body are proportional in size to strength ratio with the finger muscles so just knowing the size of motors that work for fingers, you can estimate the size of motors needed for the other muscles accordingly with significant accuracy in my opinion. In other words, say a human muscle is the size of a thick black permanent marker and another human muscle is twice that size. If you calculated the motor that matches the first marker, in order to select a motor to match the muscle that is twice that size you'd merely need to find a motor that outputs twice as much as the first motor and generally that means that motor is about twice the size of the first motor in can size volume along with the increased specs being about twice. Granted it's good to do the detailed math eventually to make sure but this will get you close to estimate when ordering motors initially.
I will post shortly a full math in detail that I've done and never thought to post (did not think there'd be interest to see it). In fact, I still am pretty sure there is zero interest in seeing the math I've done on this other than the interest to point out it was not shown in an effort to discredit me and make me look amateurish to poke fun at me frankly. But regardless the reasons I'll post it shortly.
You say "You are aiming extremely high here with your goal of replicating all degrees of freedom of the human body, in the same form factor and with human-like agility and fluidity of motion." --- I agree it is a high aim but I also am certain this will be the bear minimum standard for humanoid robotics in the future and aiming for anything shy of it now would ensure my project is outdated by the time it's done and would be considered sub par and total trash by people in the future were it not to achieve these minimum benchmarks you just mentioned.
You say "It looks like you started to cobble together pulley drives etc. right away, without at least rough estimates of the performance requirements?" --- that's utterly false. I first had rough estimates before I began even designing the pulleys and then I had hard specific numbers before I began the physical building of them.
You say "You really need to take that engineering approach of starting with top-down estimates, otherwise you will design yourself into a corner." --- I have done so and so I agree with this advice and practice it already.
@steve
You say "You think you can write some basic code to kick it off, which will then write the rest of the code for you. But this is so far away from what AI is about and can do that I think you will be disappointed." --- I assume by this you are referring to popular industry standard AI - like LLMs and their limitations. I agree with you on that. Which is one reason why I have zero interest in using or developing LLMs and have no interest in following ANY AI done by ANYBODY else. I am making my own custom AI that can do this type of thing in question and thrive at it (in theory, ideally).
You say "The concept is tempting, but I need to explain. Before anyone can write the code - be it AI or human - someone has to specify the functionality, otherwise it'll just grow uncontrollably in random, purposeless directions." --- I agree. I already have laid out for the AI its chores, its roles socially, its "family" aka my own family, etc. Then its research and self learning and the accompanying code it makes stemming from its research and self learning will be goal oriented toward enabling it to fulfill its chores and tasks as well as its social functions. This will prevent it from coding irrelevant things that go into too many rabbit holes and wasted time for it.
Also, I'm not going to release it randomly to just start coding randomly. I will be directly supervising and revising and overseeing all the coding it does on a regular basis to ensure quality and relevance. If I see it going into things I deem irrelevant I can tell it to stop that and get back on track for example. Your notion of it growing uncontrollably like I'm going to set it and forget it is not at all reflecting the very structured way I intend for it to learn and grow and be supervised in its learning. I will have fine control of the whole process.
You say "And specifying the functionality is what requires 99% of the work, not the coding." --- I disagree. Not sure really what you even mean by this. I'd say its the very opposite of what you just said here.
You say "Coding is what you delegate to coders: human or otherwise." --- coding an AI that is custom and world class and one of a kind and completely innovative is not something I would delegate to others. That is a recipe for failure. I must do that myself and then hand the baton to the robot itself to assist me in furthering the development. The core AI that enables an AI to code AI is one AI - I call it the seed. Once the seed is made, that seed can be planted and grow into a tree. I'd be the husbandman of that tree, trimming it, watering it, protecting it from bugs, overseeing its progress, and harvesting fruit from it. I cannot delegate the seed making since that is something never before done to my knowledge and HARD. I will only delegate the tree growth to the seed.
You say "So there's no chance you can set it going and just sit back. How will it know what its god (you) wants?" --- I already specified what I want like 100 pages of this type of stuff from ethics to rules of social engagements to task scheduling and priorities to daily social interactions to trust systems for sources of new information to what types of research I want it to do etc. But just as my children IRL don't learn unsupervised, but require constant adult direction, so also it will be as my robot learns and codes itself like my children code themselves with adults supervising. It is not a sitting back thing at all I agree.
You say "The idea that we can take a little bit of AI code and tell it "build me a word processor", "build me a graphics editor"" --- that's really not quite how this works in my plans though frankly. The seed (little bit of AI enabling it to code) just gets me to a more collaborative coding arrangement with the AI where it begins self learning with direct supervision and will most likely be quite error prone at first and amateurish - this would be like a child in infancy just learning to learn. At this phase it requires very strict supervision and is not particularly productive and the goal is just getting it better at coding and giving it more and more rules to clean up its code production and improve that on simple coding tasks. Only YEARS later as it continues to collaboratively learn with me and gets more and more advanced would it be getting into being able to make independent code for outside software projects. There's a difference between it coding itself in my own custom coding language for it to code itself and for it to code in C++ or w/e an outside software project for end users to use which you are proposing here. That's apples and oranges. I will say that after years of slow growth and correction, the type of requests you posted here may be on the table as possibilities. But you suggesting that RIGHT OFF THE JUMP from the seed phase it taking on "build me a word processor" and doing that successfully is just a obvious non-starter but I never claimed or intended that understanding. That type of complex job done by it is at full grown mature tree phase, not seed phase, and that type of job would be mature fruit bearing. A seedling does not bear fruit yet. So you are jumping the gun by decades on that type of task ask. I never was suggesting that.
You say "I predict the software will be the downfall of this project." --- well most predict the software, and every aspect of the hardware will be the downfall or rather the downfall has already happened and I will never even pass the starting line. So at least you are making it sound like there's a chance on the other stuff! LOL.
You say "The OP has an answer to every point raised by others, but I don't believe he is actually thinking about and considering these points." --- I disagree. I have thought about them for many years and already addressed them all years ago. This is mostly all rehash.
You say "I don't believe the OP has done anything at all to familiarize himself with current state of the art in robotics." --- that's not true. Every youtube channel even relevant to the field I've seen most of the videos available in even tons of adjacent fields. Thousands and thousands of hours of study. Every robot related podcast I've devoured that I could get my hands on as well. Even lectures given at universities on robotics I've devoured. My notes are enough to produce books.
You say "People have already made robots with string and pulleys - over 200 years ago!" --- I have not gone into 200+ year old robotics projects research admittedly. That source of information I have not really come across nor do I think its very relevant necessarily to modern materials and techniques available.
You say "I don't think this is fundamentally different from any other automaton, is it? No new technology that I can see, here." --- there are dozens of innovations I've made from the means of cooling the motors to the means of cooling the air and the manner of water cooling systems and ice cooling systems and peeing systems and the manner in which I attach the motors and the bone sleeves idea and the use of a medical skeleton and the use of bldc motors with external downgearing located physically significantly far away from the motor within the robot, etc etc these are all significant innovations never before done to my knowledge. So no, there's TONS of new technological breakthroughs I'm bringing here.
You say "It seems too much like lots of other automata (especially the 19th century ones)" --- I know little about that but am pretty sure they had a single motor and the whole thing was like a clock with countless gears and cogs. Mine is nothing like that. Not even sure why you make this comparison. Bizarre. And they had no software so were not robots. They were just clocks in a way - well clockwork devices, not time pieces.
You say "And yes, uncannily lifelike partial robots can be seen all over the place online." --- and so what? None of them can do useful work so someone needs to make them able to do useful work like industrial humanoid robots do while also having the uncannily lifelike robot appearance. The best of both worlds has not been done and I hope to do it. Others will do it too and I'm surprised none have even attempted it YET but they will shortly.
You say "The OP is severely hampered by his Christian belief, because it prevents him from recognizing that the human form is imperfect ("God made Man in his own image", so we must be perfect, too) . That puts major and unnecessary constraints on the scope of his ideas." --- I disagree. I decided its perfect not because the Bible says it is nor because I assumed God intended it to be. I decided its perfect based on the evidence as I studied it. It is my assessment of it based on the merits I saw during my study, not because of my Christian ideologies as you falsely suppose. Also, it is not hampering anything. If anything my ideas need scope limitations and constraints to prevent me from going for even more unrealistic goals than I already have (which cause so many to complain as is). You think I've aimed too low and literally everyone else is saying I aimed too high already. Ironic.
You say "I don't want the OP to respond to this post, because we are already going in circles." --- well I disagree. I should respond to anything I deem in need of response. It's actually foolish to not want me to respond to specific points made in your post that are not previously answered and have valid responses readily available. Also I disagree that we've been going in circles.
You say "one thing for sure, the OP is NO engineer." --- not by formal definition. But I am doing work engineers do. So informally I am an informal engineer in the eyes of many. In fact a colleague who is a university trained engineer and roboticist called me a very talented and creative engineer. That's a fact.
@Wobbly
You say that when I claim 100% of the text in this thread posted by me is written by me is evasive. I disagree. Let me clarify then. 100% of the text in this thread is posted by me, a human being typing with human fingers using a human brain to come up with it and no AI is involved in any way in the text production. Good luck finding loopholes in that.
You say "It doesn't preclude you being an LLM yourself (perhaps one set up to try to pass the Turing test)." --- I am not an LLM myself typing this and BTW you FAR overestimate LLMs. They cannot come even remotely close to the level of my responses.
You say "Kudos on your typing speed though, I've never known anyone to type that fast and make so few mistakes, very impressive." --- LLMs don't type so you just outed yourself by pretending you think I'm an LLM typing this while acknowledging its being typed on a keyboard. Also, I proof read before I submit so that's part of why there's not many mistakes. That and a spell checker on my browser.
You say "Doesn't pass the Turing test in my opinion though" --- I'm a man typing at my computer and I don't pass the turing test. Then the test is flawed. I mean what is so robotic about my typing that people think I can't be human. Very odd.
Home
Help
Search
Login
Register
