On 13 Jan 2014, at 20:27, meekerdb wrote:
On 1/13/2014 7:17 AM, Gabriel Bodeen wrote:
On Friday, January 10, 2014 8:17:13 PM UTC-6, Brent wrote:
On 1/10/2014 10:49 AM, Gabriel Bodeen wrote:
On Tuesday, December 31, 2013 4:25:04 PM UTC-6, Brent wrote:
As you've explained it above your theory makes a rock just as
conscious as a brain. I'm
sure you must have a more subtle theory than that, so I'll ask you
the same thing I asked
Bruno, if I make a robot what do I have to do make it conscious or
not conscious?
Brent
Did you receive any interesting answers?
Hm, should I take that as a negative answer, or merely as a skipped
question?
I didn't get any answer from Mr. Owen. Bruno's answer is that the
robot has to be Lobian, i.e. can do proofs by transfinite induction.
Normal induction is enough.
You can even limit induction on the decidable (sigma_0) formula, but
then you have to add the exponential axioms:
x^0 = 1
x^s(y) = x *(x^y)
Those exponential axioms are not provable from addition and
multiplication without induction on at least all semi-decidable (RE,
sigma_1) formula.
I have adequate background in neuroscience but I'm mostly ignorant
of AI math, robotics work, and philosophy of
mind, so excuse my rampant speculation. This is what I'd try in
design of a robotic brain to switch on and off consciousness and
test for its presence: First, I'd give the robot brain modules to
interpret its sensory inputs in an associative manner analogous to
human sensory associative regions. All these sensory inputs would
feed into the decision-making module (DMM). One of the first
steps taken by the DMM is determining how important each sensory
signal is for its current objectives. It decides to pay attention
to a subset of those signals.
So is it conscious of those signals? How does it decide?
1: As described in the next two sentences of the original
paragraph, no.
2: The choice of function used to select the subset is unimportant
to the experiment, but if we were aiming for biomimicry then each
sensory module would report a degree of stimulation, and attention
function would block all signals but the most stimulated 1 to 7.
Second, I'd put a switch on another input to make it part of the
attention subset or not:
What other input would you put a switch on? What inputs are there
besides sensory? I think you've assumed "conscious" = "self
aware". Is one conscious when one is "lost in thought"?
1: The switch would go on the signals described in the second half
of the sentence that you hastily cut in half. :D
2: Inputs besides sensory associations are important to a
functioning robot but not, I predict, to a robot designed only to
test for consciousness.
3: I chose to address the specific matter of qualia rather than all
of what people mean by "conscious", as described in the "I predict
this because..." sentence of the original paragraph. :D
4: I suspect that the human experience of being lost in thought
differs between specific cases. Most times for me that I'd call
"lost in thought" I can still operate (drive, walk, eat) on "auto-
pilot" which undoubtedly requires my senses to be engaged, but
afterwards the only things I can recall experiencing are the
thoughts I was lost in. Introspective evidence and memory being as
bad as they are, that shouldn't be taken as a necessarily correct
description. But if it is a correct description, then by my
definitions in the original paragraph, I'd say that I was
conscious. But if what you mean by "conscious" includes awareness
of surroundings, then no, I was not conscious under that definition.
Yes, it seems there are different levels and kinds of consciousness:
perception of the external world, perception of one's body, modeling
one's place in the external world, being aware of one's thoughts
(although I think this is over rated), feelings of empathy,...
the attention's choice of signals would also an input to the DMM,
and I could turn on or off whether that attentional choice was
itself let pass through to the next processing stages. I would
predict that, with the switch turned off, the robot would be not
conscious (i.e. it would have no experience of qualia), but that
with the switch turned on, the robot would be conscious (i.e. it
would experience qualia corresponding to the signals it is paying
attention to). I predict this because it seems to me that the
experience of qualia can be described as being simultaneously
aware of a sensory datum and (recursively) aware of being aware of
it. If the robot AI was sufficiently advanced that we
could program it to talk about its experiences,
the test of my prediction would be that, with the switch off, the
robot would talk about what it sees and hears, and that with the
switch on, the robot would also talk about fact that it knew it
was seeing and hearing things.
So is a Mars Rover conscious because it processes video from it's
camera to send to JPL, AND it senses that its camera is powered and
working and that its transmitter is working AND it reports those
internal status variables to JPL too.
If there are two separate inputs to the transmitter, "the video
feed" and "the camera is functional", then this does not satisfy
the relationship I described and consequently I would predict no
consciousness (of the video feed by the Mars Rover). However, that
should be possible to change. The Mars Rover is, I think, semi-
autonomous, meaning it is programmed to make certain decisions on
its own. I'll suppose a scenario in which JPL instructs the Rover
to advance toward a nifty-looking rock, but leaves the details of
that operation to the Rover's programming. Then the Rover examines
the video feed, identifies the pertinent rock in the video feed,
and advances toward it. As it does so, it uses the video feed and
the part of the video image identified as rock to continually
recalculate and adjust which part of the video feed it is
identifying as the rock. That scenario matches the one I described
previously so I would predict that the Rover would then be
conscious (of the rock).
I think that's the way it works, and it is also aware of a lot of
parameters describing itself, e.g. its position on the surface,
temperature of modules, power supply voltages and currents,...which
one could equate to a feeling of health.
The Rover would still not be self-conscious (i.e. conscious of its
self) in that scenario. If we wanted to build that kind of
consciousness, then I predict we'd need a different set-up. A
robot programmed to move so as to prevent anything from touching
its robot body would need to be given a definition of what counts
as its body. Then I think it would count as self-conscious.
However, if you want something still deeper, like psychological
self-consciousness (i.e. consciousness of its own psychological
state), then you might have to build a robot and program it using
quining or something like that -- I'm not sure, as this ventures
far enough into AI math that I know my intuitions are a very bad
guide.
That's where Bruno would invoke the ability to prove that not all
true statements are provable.
Hmm... OK. (for now :)
Bruno
http://iridia.ulb.ac.be/~marchal/
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/groups/opt_out.
