[ExI] What's Wrong With Academic Futurists?

[Keith Henson]

On Tue, Jan 28, 2014 at 11:19 AM,  Robin D Hanson <rhanson at gmu.edu> wrote:

snip hkh comments

> You seem to think that only predictions made with confidence are relevant.
> If you had assigned a 5% chance to the event that happened, while others
> assigned a 1% chance, then your prediction was more accurate, even if it
> was far from confident.

I see what you are getting at, for example, Dr. Shen's approach,
spectral optimal averaging method for estimating uncertainty in
climate change. "Rather than presenting their findings as absolutes,
the IPCC provides estimates of certainty for their findings in an
effort to provide to the public an indirect measure of confidence in
their conclusions."

It may be useful to climate professionals, but I am not sure how
useful this is to the public. My suspicion is between not very useful
and not useful at all. Climate is not like predicting eclipses, at
least not yet. You also have an infinite regression problem of how
confident are you in your estimation of uncertainty.

(Not that I am a denier, I just don't think it matters. The solution
to the more serious energy crisis also takes care of climate to
*whatever* extent CO2 affects climate.)

But climate prediction is relatively easy compared to the class of
future affecting disruptions such as AI, nanotech, or some unknown and
unpredicted wild card that trashes our common baseline assumptions
about the future.

Not that I have not tried, "The Clinic Seed" being my attempt to
project a gentle, if still species ending, interaction of AIs with
humans.  And you know of my concerns about some of us experiencing 50
million subjective years before the end of this calendar century.

> One can also score a pool of conditional predictions for accuracy. So even if
> you only have opinions on what might happen if some other things happen,
> we can still talk about the general accuracy of your predictions.

You completely lost me here.  If you make a prediction for ten years
out, the only way I can see to rate it for accuracy is to wait ten
years and see what has happened.  If you have another idea, I would be
really interested in it.

Keith