[extropy-chat] Not necessary smarter, just faster?

Samantha Atkins sjatkins at mac.com
Thu May 11 19:54:48 UTC 2006


On May 11, 2006, at 11:42 AM, Eliezer S. Yudkowsky wrote:

> Samantha Atkins wrote:
>>
>> It also seems very likely that the neocortex would optimize many
>> problems faster and more fully when run at vastly higher speeds with
>> equivalently speeded up inputs.  Such a brain would be smarter  
>> over time
>> and in much shorter time than otherwise.
>
> 1)  Didn't you mean to say vastly higher speeds *without* equivalently
> speeded up inputs?  Otherwise you have a simple isomorphism that
> wouldn't optimize any faster or more fully.

How would you optimize such a network faster with no more inputs than  
before?  I see that the internal chewing over the inputs would happen  
much faster and potentially be more extensive.  But that internal  
processing may terminate with not much better results unless more  
relevant external input is available.

>
> 2)  If you try any hack that *isn't* a simple isomorphism, the  
> neocortex
> will probably break down unless you use an extremely tricky  
> engineering
> hack to keep it running sanely.  The human brain is not
> end-user-modifiable and its parts are not individually overclockable.
> If the neocortex has any inputs that depend on spiking frequency or
> spike timing (d'you think?) then speeding up the inputs breaks the  
> API.
>

Since we are positing a human brain equivalent that runs a million  
times faster it seems to me this question is challenging  what was  
posited to start with.   Somehow we have a brain running one million  
times faster.  Now what can it likely do and not do?

- samantha




More information about the extropy-chat mailing list