NancyLebovitz comments on Open Thread: April 2010, Part 2 - Less Wrong

3 Post author: Unnamed 08 April 2010 03:09AM

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

Comments (194)

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

Comment author: NancyLebovitz 19 April 2010 03:12:27PM 0 points [-]

Computers have memory limits. They're just much higher than human limits.

WM?

Comment author: gwern 19 April 2010 04:28:28PM 3 points [-]

They're just much higher than human limits.

It's not just quantity; it's quality. Human WM is qualitatively different from RAM.

Yes, you could invent a 'dual 4-gigabyte back', and the computer would do just as well. Bits don't change in RAM. If it needs to compare 4 billion rounds back, it will compare as easily as if it were 1 round back. Computer 'attention' doesn't drift, while a human can still make mistakes on D1B. And so on.

You could cripple a computer to make mistakes like a human, but the word 'cripple' is exactly what's going on and demonstrates that the errors and problems of human WM have nothing interesting to say about the theoretical value (if any) of forgetting.

You only need to forget in DNB because you have so little WM. If you could remember 1000 items in your WM, what value would forgetting have on D10B? It would have none; forgetting is a hack, a workaround your limits, an optimization akin to Y2K.

Comment author: cupholder 19 April 2010 03:27:13PM 1 point [-]