Machine Learning

From Open Source Ecology
Jump to: navigation, search

Basics

  • A process by which an algorythm can produce other algorhythms, code, and even CAD based off of training data, and training
  • many methods each with their own advantages
  • One near essential application is complex problems such as autonomous navigation, problem solving, object recognition, etc
  • Being software, there are many open source tools
  • Main requirements are AI Training Data
  • Can create solutions to problems humans could never make
  • Can sometimes be hard to explain the core of the "method" it comes up with
  • See the CPG Grey Video On It
  • Can be a buzzword so keep that in mind when reading headlines etc
  • NEED EXPERT EXPLAINING

Applications

AI Machine Learning with Deep Neural Networks have been used to produce many types of content that may be valuable to Open Source Ecology.

Producing Text

In 2022, ChatGPT was publicly released. It can produce large quantities of text from a prompt.

Producing Images

In 2021 & 2022, DALL-E and Stable Diffusion were released. These tools can produce images from a prompt.

Images generated by these tools could be very useful in creating visuals in documentation.

  • "photo of a futuristic steampunk tractor on an assembly line"

  • "STL file of a light bulb"

  • "open source compressed earth block press"

Producing Code

In 2021 Microsoft released GitHub Copilot, which -- using a neural network trained by open-source software on GitHub -- is able to produce code from a prompt describing a function.

Producing CAD

TODO: see if anyone has trained a neural network with large quantities of CAD data (eg everything on theingiverse) to produce CAD diagrams (or at least OpenSCAD code) from a prompt

See Also

Useful Links

Retrieved from "https://wiki.opensourceecology.org/index.php?title=Machine_Learning&oldid=279644"