Alan,

Your posting ignited an intense feeling of deja vu. After posting,
presenting at conferences, etc., about how to construct general purpose
fault-tolerant wafer-scale computers, now THIS, that shows every evidence
of having "missed the message".

Transport back to the 1960s/1970s, when I and others were promoting various
proposals for the first monolithic microprocessor. My design for a serial
architecture using bipolar technology could have been built years earlier.
The design Intel adopted was by FAR the worst of the lot, because it had
been selected for "political" reasons.

It is possible to build super-reliable processors from unreliable
components - but this is SO counterintuitive that few can wrap their
heads around it.

And, yes, there is even a way to recognize and survive in-process component
failures - as Tandem Computers implemented.

Once past that, you can put the pieces together using configurable logic -
to make almost ANYTHING, rather than being stuck with a fixed architecture
- like 400k tiny processors.

Most AI is adaptable to data-chaining, which can be done MUCH faster than
on conventional processors.

I fear these guys will further blacken the prospects for useful wafer scale
devices - a process started at Trilogy.

Depressing...

Steve

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Taefa5aa6c0a39ae4-M3f098347cbca91f6b8a765f7
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to