Show HN: Thinkable Programming

1 pointdlillard6 comments

Start with taking a file and a program that executes it. Then have a neural network take its own output as its input, using the difference of the program and the average between the file and the program as its error. Finally set the program to the predictions from the neural network.

This is the algorithm for a thinking program.

View on HN

This post is very entertaining.

I would like to add:

Consider taking the set of all cli programs that you can pipe inputs and Outputs to,

Then take your entire file system tree,

Use just a few rules so that like image stuff isn't opening up sound files

Cook your CPU and maybe suffer from catastrophic forgetting?

It’s really a very simple program that I don’t see how you’re misunderstanding.

And I'm misunderstanding how you don't see my comment as an extension of your idea. I'm talking about how to automatically load, process, and train a model based off an entire file system and programs which is an extension of what you said?

Does it work?

I haven't implemented it yet.

Show HN is more for things that can be showhn and tried, take a look at https://news.ycombinator.com/showhn.html

Reply on news.ycombinator.com