thechnology: optimists and pessimists

Carl Feynman (carlf@atg.com)
Tue, 01 Jul 1997 16:31:31 -0400


At 11:34 AM 6/25/97 -0400, Michael Lorrey wrote:
>Carl Feynman wrote:
>> There have been other
>> periods of exponential growth in technological capacity, and they all came
>> to an end, usually without anyone realizing in advance that they were about
>> to.
>
>Usually it was politically imposed stagnation by reactionaries holding
>progress hostage to force change or to merely maintain the status quo.

No, usually it was economically imposed saturation of a market or a rise in
costs due to diminishing returns to scale. The mileage of canals in the
U.S. grew exponentially from 1790 to 1836, the mileage of railways from 1830
to 1890, the mileage of paved roads from 1900 to 1950. After that the
growth leveled off, and in the case of canals and railways mileage
eventually began to decline. I suppose according to your theory, if it
weren't for those nasty reactionaries, Americans would enjoy billions of
miles of canals and railroads.

>Technology has no limits save c.

And what people are willing to pay for.

>>
>> Commercial air transport improved from 1903 to 1970 at an exponential clip,
>> and has made essentially no progress since. Nobody had any idea in 1970
>> that progress was over. If we were still on that curve, we'd have
>> commercial trips to the moon. The movie '2001' was a plausible
>> extrapolation at the time.
>
>And if commercial supersonic flight were not illegal over continental
>landmasses, and the fuel embargo had not occured, we would still be on
>that curve. It was not a matter of technology.

It doesn't matter if it was, according to your standards, a "matter of
technology" or not. The point is that exponential growth does not continue
indefinitely, and it often ends by running into unexpected barriers. In
1970, who would have thought noise, pollution, or a bunch of Arab
politicians could hve stopped the mighty juggernaut of American aerospace
progess? In fact, political forces tried to continue the exponential curve
past the point of feasibility, producing the money-losing Concorde.

>In fact, if you count our
>autonomous spaceflight technology,

I don't. I chose my words advisedly: commercial air transport. I would
include commercial manned space transport, if there were any such thing.

>> The maximum speed of a Pentium might be 500 MHz for
>> decades, until someone develops a breakthrough technology.
>
>Sure there may by a maximum speed for that particular computer
>technology, but there is no reason other technologies will not surpass
>the Pentium chip as it becomes outmoded by economics.

Well, as long as we're arguing on the basis of "no reason why not": there is
also no reason why those new technologies should not appear either faster or
slower than Mooore's law suggests.

The recent exponential increase in the cost of fabs to build each generation
of chips is an alarming sign. Each generation costs twice as much capital
as the one before, and it is only the matching exponential growth in sales
that has kept the price per unit manufactured down. If this continues, in
three generations, the cost of a single fab will exceed the quantity of
capital that can be feasibly assembled into a single commercial venture
(about 10 billion dollars). That's why I picked 500 MHz: it's two fab
generations in the future. It will be perfectly feasible to build faster
chips than that, and Intel probably will. It just won't be feasible to
build them at a profit. A 1000 MHz Pentium can't cost more than three
times a 500 MHz Pentium, or nobody will buy it, because it will be cheaper
to use a multiprocessor machine.

If this happens, I suspect the PowerPC or other RISC chip will eventually
become the microprocessor of choice, because, given a fixed fab technology,
RISC chips are about 50% faster than Pentiums, ande in the absence of rapid
growth, siwtching will make sense. So in 2001, when the financing for
Intel's $12 billion new fab falls apart, my advice is, buy IBM and Motorola.

--CarlF