Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

jhuffman comments on For progress to be by accumulation and not by random walk, read great books - Less Wrong

35 Post author: MichaelVassar 02 March 2010 08:11AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (102)

You are viewing a single comment's thread. Show more comments above.

Comment author: jhuffman 05 March 2010 06:19:15PM *  0 points [-]

Not if you think what Karl mentions above. The problem is that the amount of thought that you can hold in your head at one time is finite and differs significantly from one person to another.

In other words: algorithms need working memory, which is not boundless.

Comment author: JamesAndrix 05 March 2010 07:45:41PM 0 points [-]

Well first off, I was assuming pencil and paper were allowable augmentations.

I would be surprised if it were the case that our brain process that finds big insights with N 'bits of working memory' couldn't be serialized to find the same big insights as a sequence of small insights produced by a brain running a similar process but with only N/2 available 'bits'.

Comment author: jhuffman 05 March 2010 08:35:43PM *  0 points [-]

Imagine yourself studying a 4 megapixel digital image only by looking at it one pixel at a time. Yes, you can look at it, and then even write down what color it was. Later you can refer back to this list and see what color a particular pixel was. Its hard to remember more than a few dozen at once though, so how will you ever have a complete picture of it in your head?

Comment author: JamesAndrix 06 March 2010 02:07:55AM 1 point [-]

I could find and write down a set of instructions that would allow you to determine if there was a face in the image. If you were immortal and I were smarter, I could write down a set of instructions that might enable you to derive the physics of the photographed universe given a few frames.

At this level it's like the Chinese room.

But I don't think the ratio between Einstein's working memory and a normal person's working memory is 100,000 to 1.

It would be EASY to make instructions to find faces even if someone could only see and remember 1/16th of the image at a time. You get tons of image processing for free. "Is there a dark circle surrounded by a color?"

A human runnable algorithm to turn data into concepts would be different in structure, but not in kind.