Re: [sl4] Is there a model for RSI?

From: Mark Nuzzolilo (nuzz604@gmail.com)
Date: Mon Jun 16 2008 - 16:23:01 MDT


Now we are getting into the practical details which is a departure from the
original question which (I assume) was "Is RSI possible?". If the answer to
that question is yes, then a method of reaching it exists, and we can then
worry about finding it (a separate problem from building sub-human level
AI). My example didn't have anything to do with the details of achieving
RSI, it was only to answer the above question.

I agree that lots of experimentation and slow progress is ahead, and that is
why Strong AI probably won't be here for another 25 years or so (and even
that is assuming that the funding situation will improve soon). You have to
get there with incremental steps.

Mark Nuzzolilo

On Mon, Jun 16, 2008 at 11:42 AM, William Pearson <wil.pearson@gmail.com>
wrote:

> We tend to need lots of experimentation (read as getting information
> from the environment on how to improve themselves), to create things a
> lot simpler than ourselves. Lock a group of scientist in a room with
> no test data or robot body and no way to generate any and you will
> find it hard for that team to create anything more intelligent than
> itself.
>
> You have to formalise the constraint on getting information from the
> environment to get anywhere with this type of discussion.
>
> Will Pearson
>



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT