# PREDICTION: Naive extrapolation

Gregory Sullivan (sullivan@blaze.cs.jhu.edu)
Mon, 02 Dec 96 16:25:40 EST

I suspect many longtime list members have been exposed to a variety of
predictions based on naive extrapolations of present trends. "Moore's Law"
has been one of the more successful heuristics, yet it is also known to be
inaccurate. For example, the original statement of "Moore's Law" claimed
that there would be a doubling of elements on a chip every 12 months - not
every eighteen months as claimed nowadays.

I highly recommend that anyone interested in thinking about the future read
one of the classic early articles of prediction based on mathematical
extrapolation.

Science Fiction is Too Conservative
by G. Harry Stine
May 1961
Analog

Many of the key ideas behind the notion of a "singularity" are contained in
this article from 35 years ago. The article also illustrates how simplistic
extrapolation can lead to wildly inaccurate projections.

Examples:

`manned vehicles will be able to achieve near-infinite speeds by 1982.'

This prediction came from graphing the speeds of vehicles over time. The
graph looks great - dramatic - it is superexponential. The prediction,
however, was deeply flawed.

`Life expectancy is increasing, and this trend curve indicates that
anyone born after the year 2000 A.D. lives forever, barring accidents'.

This is an intriguing projection - especially for a writer in 1961. If
nanotechnology succeeds `quickly' it might even turn out to be largely
accurate (if you include events such as `war' and `terrorism' in the
category `accidents'.)

`Population is rising rapidly, and early in the Twenty-first Century there
isn't enough room on the planet Earth for everybody. This curve shows no
more signs of leveling off than the other trend curves do, so we cannot
take the easy way out via starvation, birth control, or mass destruction,
because those things are apparently not in the cards when other trend
curves are also considered.'

Population curves have leveled off in many countries. There is negative
population growth in some Western countries. Current projections by
organizations such as the UN foresee a leveling off of the population.
(Note, I think the UN projections as just as bogus as Stine's in this area
but my point is that it is difficult to make projections.)

`By 1981, this trend curve shows that a single man will have available
under his control the amount of energy equivalent to that generated by
the entire sun.'

Where is my super-atomic blaster ray? If we build Von-Neumann space probes,
for example, then massive amounts of energy may, arguably, be controlled in
the future.

`The number of circuits in cybernetic devices is increasing on the
familiar trend curve. The human brain has an estimated four billion
neural circuits. By 1970, computer engineers may have achieved the same
number of circuits in a digital computer; they may do this by building
one large computer or by slaving many smaller computers together by data
links as they have already started to do. ... Will these machines think?'

Interesting projection. The notion of `four billion neural circuits' is pure
guesswork. We still do not know how brains work - although we are rapidly
making progress, I would claim. Certainly, there were no human-level AI
systems in 1970, and there are none today - even the world's chess champion
is still a human. However, Stine did foresee the exponential increase in
computer power that is oft cited by latter day prognosticators.

The article includes several pictures.

Caption for the picture of lunar colony:

The lunar colony...a drawing from a report on colonization done by the
Martin Company in 1959. American free enterprise has become imaginative and
conceptual in its thinking while science fiction plods its weary, well-worn
squirrel cage of old ideas.

End caption

The drawing of the lunar colony is from the perspective of a couple who are
dining and overlooking a swimming pool, a fountain, a lake, a zoo and a golf
area. They are in a bubble structure constructed with hexagonal plates and
another bubble structure is visible through the transparent walls of the
bubble.

No prediction year is given for this lunar colony. Good move.

In conclusion, Stine's article is must reading for anyone who likes to make
predictions and who likes to think about the `singularity', in my opinion. I
especially recommend it to any person who likes to give dates with his or her
predictions.

I am not trying to make fun of Stine. In fact, in overall outline, he might
be more right than wrong. Nor, do I wish to discourage thinking about the
future - I do it all the time - I have done it for decades. But, perhaps it
is possible to learn from previous errors.

Gregory Sullivan