As AI and manufacturing come together through 3D printing, we coders will come in charge of how things are actually made
Blockchain and Dunning-Krugerrands will probably make less of an impact on the world than 3D printing and AI, especially for developers. We get paid a lot to push bits around the internet, but ultimately people want those bits to result in things happening.
As AI and manufacturing come together through 3D printing, we coders will come in charge of how things are actually made. In fact, this is already happening.
Here are a few examples:
- Testing: GE is using computer vision to detect flaws in its 3D-printed parts before the component is assembled.
- Model correction: Most models are either scanned or created using CAD software or vector instructions for a slicing machine. By using machine learning, you can create correct more efficient models or instructions before anything is printed or sliced.
- 3D extrapolation: A whole lot more people can sketch than use CAD software, and there are a whole lot more 2D sketches in the world than 3D models. You can use machine learning to extrapolate a 3D model from a 2D sketch.
- Presorting powder: Most 3D printing is an additive process that melts powder into place; for metallic parts, that means a metal powder. Nothing is perfect, but the right qualities are needed in a powder especially if you’re going to create medical devices and other critical devices. Computer vision is being used to sort the primordial powder before it even becomes a part.
- Voice search: A lot more people know how to talk than even sketch, and voice and search are making it possible for people to describe what they need to a printer.
How can you as a developer get started in this? Well, having a 3D printer and knowing how to use it is a good start. A 3D scanner is also helpful, but now there are apps for that (granted, with lower quality than a real scanner). The next trick is to learn more about machine learning specifically and AI generally.
Experimentation is, however, critical to understanding. Facebook just released its object detection framework for its deep learning toolkit. And Thingiverse is where you can find a lot of predesigned models ready to print. Can you find the flaws in what you’ve printed? Can you maybe even figure out how much infill to do based on your requirements? These are manufacturing problems that code can solve and why the manufacturer of the future is a software developer.
