Interpretations, subjectivity

From: Hans Moravec <>
Date: Thu, 08 Jul 1999 01:01:32 -0400

Wei Dai:
> I'm not a fan of the "all possible interpretations" approach. First the
> existence of all possible objects can be deduced from a simplicity
> assumption, so the "all possible interpretations" approach is not
> necessary. Second I don't think the subjective experience of an observer is
> necessarily open to interpretation. We don't know what it feels like to be
> someone else, but that doesn't mean there is no matter of fact about what
> it feels like to be that person.

Thinking about interpretations brought me to the all-universes position!
You may remember the early stages of the rumination in in Mind Children,
where I discussed possible beings in Fourier and other transforms of
normal space, each transform being a different interpretation and an
independent universe, for which the others were merely background noise.
The interpretation route connects a lot of possible worlds to our
present experience, and opens the possibility of seeing some of
them. I imagined a gadget that, as you turned a knob, scanned through
a large number of possible transforms, say of the sun. Most of the
setting would show noise, but just possibly you find a transform where
intelligent beings had evolved like the ones I postulated in
the Fourier domain. (Visual and audio interfaces to simulators and
video games already do this in a premeditated way, allowing
us to interpret confusing electron motions in a handful of silicon bits
as the exciting life of the Super Mario Brothers.)
Attribution of subjective experience is a festering issue for those
of us who presume to assemble messes of inanimate components into
and then wish to claim the results have a subjective life on the basis
of internal models that allow the robots to discuss their experiences,
goals, likes and dislikes just like conventionally grown persons.

On the one hand, an engineer who completely understands a robot has no
need to attribute subjectivity to it, since she can fully explain, and
even predict, everything the robot does from a mechanistic model
of the designed interaction of its parts.

On the other hand, normal folks interacting with the robot, which is
built to act like normal people, will instinctively attribute
feelings to it, especially when the robot discusses how it feels
(and asks them how they feel, updating its models of them on the basis
of the conversations). And in its own mind (when it is interpreted
as having one) it believes it has feelings, and for that matter that
it has beliefs. Just ask it.

If you're so inclined, you can laugh this anthropomorphic
"subjective life" interpretation of the robot's behavior, which is
explained more accurately by the engineer's purely mechanical
interpretation. Many AI critics say just that about any possible
computer AI.

Now suppose an alien super-engineer lands near you. It takes out its
ultra volume-scanning camera, and gets an atomic-res snapshot of all
your many parts, inside and out. From the scan, and its knowledge of
physics, it deduces a precise mechanical model of your behavior.
>From analysing the model, it deduces that directing the odd
sequence of noises at you: "please bring me that refreshing glass of
water if you would be so kind" will make you, by a predictable
mechanical chain of cause and effect among your parts, supply
fusion fuel for its vehicle. It would laugh at anyone who suggested
an unnecessary and sloppy subjective interpretation to your actions,
which are explained much more precisely by its totally mechanical
interpretation of what you are.

The only satisfactory way I've found out of this and related
confusions is to accept subjectivity as an interpretation or
attribution like beauty or value (or computation or belief). But,
when one attributes subjectivity to an entity one may grant it the
ability to make its own attributions. And then it can attribute
subjectivity to itself. It may not really think, but it thinks it

By the same reasoning, beings in the many universes may not
really exist, but they think they do.

i.e. existence itself is an attribution.
Received on Wed Jul 07 1999 - 22:03:10 PDT

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:06 PST