Day 1: starting from scratch, wherever scratch is

A humanoid robot — sounds incredible, doesn’t it? Ok, you’re right — nobody’s impressed anymore. Robots these days run, jump, win marathons, and generally give us all plenty of reasons to wonder about our prospects for the next umpteen years. But for me it sounds like I’m finally doing the thing I’ve been thinking about for years — and yes, to me that’s incredible. Today is a big day, because today I finally started.

A few years ago I started thinking about building my own humanoid robot. The point was never to make it human-like in the anatomical sense — what matters is that it has its own intelligence and the ability to learn. And yes, obviously, this is a from-scratch project. Though at the moment I can’t even define where exactly “scratch” is, given my complete lack of any specialized knowledge on the subject.

Luckily, I have a rich technical background from the software industry, mostly around robotics and semiconductors. But — and this is a fact, and facts are the most stubborn things in the world — neither the background nor the knowledge helps me one bit with where to start when the goal is “build a humanoid.” Modeling, electronics, learning… I’m sure there are dozens of other aspects I don’t even suspect exist yet. One big mountain of questions. And not a single answer. But if life taught me anything, it’s this: if you want to do something, stop thinking about the fact that it needs to be done — it won’t start by itself. Just start doing it. Start small, with a result you can actually reach, and let it grow from there. But first, it has to be started.

Not long ago I bought a 3D printer. The motivation was entirely unrelated, but I have to admit: somewhere in a far corner of my mind there was a barely audible voice saying “Oh! That’s exactly what we need — the robot will have to be built out of something, won’t it?” Today that purchase counts as a definite plus to motivation.

With all of that in mind — today I started, and invited a couple of AI assistants to work with me as a team. Working in a team is more fun and more productive anyway, especially on projects of the “go I know not where, bring back I know not what” variety.

First discovery: there’s a free version of Autodesk Fusion for individual use, which goes straight into our toolkit for modeling the future robot’s parts. Better yet, it turns out the free version has its own built-in AI you can hand tasks to — incredible news. Our ranks are growing.

And we created our first model:

Doesn’t it look like a humanoid robot? No? Ok, fine… I know — it’s missing the muscles (motors) and the circulatory system (wiring)… and possibly a few other subsystems. But it has one joint, and I’d call that a big step in the right direction — and, for day 1, a magnificent result.

What does a robot with exactly one joint do all day? Good question. That’s tomorrow’s problem.