Arguably even RAM are useless for realtime, since you need
~second to stream through a node's memory. And that's just
one iteration. And it's not random-access, so it can get
some 10-20 times as slow.
Now if you could process the entire ~GByte chunk in ~1 us.. 100 ns
steps, that'd be interesting.
My intuition (and experience in non-AGI software, lacking AGI as yet) for what it's worth is that if you can process all your memory in much less than a second, you should spend more of your money on memory rather than CPU/bandwidth.
Though if I remember correctly, you think more highly of the prospects for brute force evolutionary programming than I do, so I can see where you're coming from on that one.
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
